®
AMD-K6 -2E
Embedded Processor
Data Sheet
Publication # 22529 Rev: B Amendment/0
Issue Date: Jan 2000
Preliminary Information
TM
© 2000 Advanced Micro Devices, Inc. All rights reserved.
The contents of this document are provided in connection with Advanced Micro
Devices, Inc. ("AMD") products. AMD makes no representations or warranties
with respect to the accuracy or completeness of the contents of this publication
and reserves the right to make changes to specifications and product
descriptions at any time without notice. No license, whether express, implied,
arising by estoppel or otherwise, to any intellectual property rights is granted by
this publication. Except as set forth in AMD’s Standard Terms and Conditions
of Sale, AMD assumes no liability whatsoever, and disclaims any express or
implied warranty, relating to its products including, but not limited to, the
implied warranty of merchantability, fitness for a particular purpose, or
infringement of any intellectual property right.
AMD’s products are not designed, intended, authorized or warranted for use as
components in systems intended for surgical implant into the body, or in other
applications intended to support or sustain life, or in any other application in
which the failure of AMD’s product could create a situation where personal
injury, death, or severe property or environmental damage may occur. AMD
reserves the right to discontinue or make changes to its products at any time
without notice.
Trademarks
AMD, the AMD logo, K6, AMD-K6, 3DNow!, and combinations thereof, K86, Super7, and E86 are trademarks; RISC86
is a registered trademark; and Fusion E86 is a service mark of Advanced Micro Devices, Inc.
MMX is a trademark and Pentium is a registered trademark of Intel Corporation.
Microsoft, Windows, and Windows NT are registered trademarks of Microsoft Corporation.
Other product names used in this publication are for identification purposes only and may be trademarks of their
respective companies.
iii
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
IF YOU HAVE QUESTIONS, WE’RE HERE TO HELP YOU.
The AMD customer service network includes U.S. offices, international offices, and a
customer training center. Expert technical assistance is available from the AMD
worldwide staff of field application engineers and factory support staff to answer
E86™ family hardware and software development questions.
Frequently accessed numbers are listed below. Additional contact information is
listed on the back of this manual. AMD’s WWW site lists the latest phone numbers.
Technical Support
Answers to technical questions are available online, through e-mail, and by telephone.
Go to AMD’s home page at www.amd.com and follow the Support link for the latest
AMD technical support phone numbers, software, and Frequently Asked Questions.
For technical support questions on all E86 products, send e-mail to
epd.support@amd.com (in the US and Canada) or
euro.tech@amd.com (in Europe and the UK).
You can also call the AMD Corporate Applications Hotline at:
(800) 222-9323 Toll-free for U.S. and Canada
44-(0) 1276-803-299 U.K. and Europe hotline
WWW Support
For specific information on E86 products, access the AMD home page at
www.amd.com and follow the Embedded Processors link. These pages provide
information on upcoming product releases, overviews of existing products,
information on product support and tools, and a list of technical documentation.
Support tools include online benchmarking tools and CodeKit software—tested
source code example applications. Many of the technical documents are available
online in PDF form.
Questions, requests, and input concerning AMD’s WWW pages can be sent via e-mail
to web.feedback@amd.com.
Documentation and Literature Support
Data books, user’s manuals, data sheets, application notes, and product CDs are free
with a simple phone call. Internationally, contact your local AMD sales office for
product literature.
iv
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
To order literature:
Web: www.amd.com/support/literature.html
U.S. and Canada: (800) 222-9323
Third-Party Support
AMD FusionE86SM partners provide an array of products designed to meet critical time-to-
market needs. Products and solutions available include chipsets, emulators, hardware and
software debuggers, board-level products, and software development tools, among others.
The WWW site and the E86™ Family Products Development Tools CD, order #21058,
describe these solutions. In addition, mature development tools and applications for
the x86 platform are widely available in the general marketplace.
Contents v
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Contents
Revision History.......................................................................... xv
About this Data Sheet...............................................................xvii
Overview................................................................................. xvii
1AMD-K6™-2E Processor .............................................................. 1
1.1 AMD-K6™-2E Embedded Processor Features......................... 2
1.2 Process Technology..................................................................... 4
1.3 Super7™ Platform Initiative ..................................................... 4
2 Internal Architecture .................................................................. 7
2.1 AMD-K6™-2E Processor Microarchitecture Overview ........... 7
2.2 Cache, Instruction Prefetch, and Predecode Bits.................. 12
2.3 Instruction Fetch and Decode ................................................. 13
2.4 Centralized Scheduler.............................................................. 17
2.5 Execution Units......................................................................... 18
2.6 Branch-Prediction Logic........................................................... 20
3 Software Environment ............................................................... 23
3.1 Registers .................................................................................... 23
3.2 Model-Specific Registers (MSR) ............................................. 40
3.3 Memory Management Registers.............................................. 47
3.4 Paging......................................................................................... 49
3.5 Descriptors and Gates .............................................................. 52
3.6 Exceptions and Interrupts ....................................................... 55
3.7 Instructions Supported by the AMD-K6™-2E Processor ...... 56
4 Logic Symbol Diagram ............................................................... 83
5 Signal Descriptions .................................................................... 85
5.1 Signal Terminology................................................................... 85
5.2 A20M# (Address Bit 20 Mask) ................................................. 86
5.3 A[31:3] (Address Bus)............................................................... 87
5.4 ADS# (Address Strobe) ............................................................ 88
5.5 ADSC# (Address Strobe Copy)................................................ 88
5.6 AHOLD (Address Hold) ........................................................... 89
5.7 AP (Address Parity).................................................................. 90
5.8 APCHK# (Address Parity Check)............................................ 91
5.9 BE[7:0]# (Byte Enables) ........................................................... 92
5.10 BF[2:0] (Bus Frequency) .......................................................... 93
5.11 BOFF# (Backoff) ....................................................................... 94
5.12 BRDY# (Burst Ready)............................................................... 95
5.13 BRDYC# (Burst Ready Copy) .................................................. 96
5.14 BREQ (Bus Request)................................................................. 96
5.15 CACHE# (Cacheable Access) .................................................. 97
5.16 CLK (Clock)............................................................................... 97
5.17 D/C# (Data/Code) ...................................................................... 98
5.18 D[63:0] (Data Bus)..................................................................... 99
5.19 DP[7:0] (Data Parity).............................................................. 100
5.20 EADS# (External Address Strobe)........................................ 101
vi Contents
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.21 EWBE# (External Write Buffer Empty) ............................... 102
5.22 FERR# (Floating-Point Error)............................................... 103
5.23 FLUSH# (Cache Flush) .......................................................... 104
5.24 HIT# (Inquire Cycle Hit)........................................................ 105
5.25 HITM# (Inquire Cycle Hit To Modified Line)...................... 105
5.26 HLDA (Hold Acknowledge)................................................... 106
5.27 HOLD (Bus Hold Request)..................................................... 107
5.28 IGNNE# (Ignore Numeric Exception)................................... 108
5.29 INIT (Initialization) ................................................................ 109
5.30 INTR (Maskable Interrupt).................................................... 110
5.31 INV (Invalidation Request) ................................................... 110
5.32 KEN# (Cache Enable)............................................................. 111
5.33 LOCK# (Bus Lock) .................................................................. 112
5.34 M/IO# (Memory or I/O) ........................................................... 113
5.35 NA# (Next Address)................................................................ 114
5.36 NMI (Non-Maskable Interrupt) ............................................. 114
5.37 PCD (Page Cache Disable)..................................................... 115
5.38 PCHK# (Parity Check) ........................................................... 116
5.39 PWT (Page Writethrough) ..................................................... 117
5.40 RESET (Reset) ........................................................................ 118
5.41 RSVD (Reserved).................................................................... 119
5.42 SCYC (Split Cycle).................................................................. 120
5.43 SMI# (System Management Interrupt)................................. 121
5.44 SMIACT# (System Management Interrupt Active)............. 122
5.45 STPCLK# (Stop Clock) ........................................................... 123
5.46 TCK (Test Clock)..................................................................... 124
5.47 TDI (Test Data Input)............................................................. 124
5.48 TDO (Test Data Output)......................................................... 124
5.49 TMS (Test Mode Select)......................................................... 125
5.50 TRST# (Test Reset)................................................................. 125
5.51 VCC2DET (VCC2 Detect) ...................................................... 126
5.52 VCC2H/L# (VCC2 High/Low)................................................. 127
5.53 W/R# (Write/Read) ................................................................. 128
5.54 WB/WT# (Writeback or Writethrough)................................. 129
5.55 Pin Tables by Type ................................................................. 130
5.56 Bus Cycle Definitions ............................................................. 132
6 Bus Cycles ................................................................................. 133
6.1 Timing Diagrams..................................................................... 133
6.2 Bus States................................................................................. 135
6.3 Memory Reads and Writes..................................................... 138
6.4 I/O Read and Write................................................................. 146
6.5 Inquire and Bus Arbitration Cycles ...................................... 148
6.6 Special Bus Cycles .................................................................. 170
7 Power-On Configuration and Initialization ........................... 179
7.1 Signals Sampled During the Falling Transition of RESET .. 179
7.2 RESET Requirements ............................................................ 180
7.3 State of Processor After RESET............................................ 180
Contents vii
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
7.4 State of Processor After INIT ................................................ 183
8 Cache Organization .................................................................. 185
8.1 MESI States in the Data Cache ............................................. 186
8.2 Predecode Bits......................................................................... 187
8.3 Cache Operation ..................................................................... 187
8.4 Cache Disabling and Flushing............................................... 190
8.5 Cache-Line Fills ...................................................................... 191
8.6 Cache-Line Replacements...................................................... 192
8.7 Write Allocate ......................................................................... 192
8.8 Prefetching .............................................................................. 197
8.9 Cache States ............................................................................ 198
8.10 Cache Coherency .................................................................... 199
8.11 Writethrough and Writeback Coherency States.................. 204
8.12 A20M# Masking of Cache Accesses ...................................... 204
9 Write Merge Buffer ................................................................. 205
9.1 EWBE# Control ....................................................................... 205
9.2 Memory Type Range Registers ............................................. 207
9.3 Memory-Range Restrictions .................................................. 209
9.4 Examples.................................................................................. 211
10 Floating-Point and Multimedia Execution Units .................. 213
10.1 Floating-Point Execution Unit............................................... 213
10.2 Multimedia and 3DNow!™ Execution Units........................ 215
10.3 Floating-Point and MMX™/3DNow!™
Instruction Compatibility....................................................... 216
11 System Management Mode (SMM) ........................................ 217
11.1 SMM Operating Mode and Default Register Values........... 217
11.2 SMM State-Save Area............................................................. 219
11.3 SMM Revision Identifier........................................................ 221
11.4 SMM Base Address ................................................................. 222
11.5 Halt Restart Slot ..................................................................... 222
11.6 I/O Trap Doubleword.............................................................. 223
11.7 I/O Trap Restart Slot .............................................................. 224
11.8 Exceptions, Interrupts, and Debug in SMM......................... 226
12 Test and Debug ......................................................................... 227
12.1 Built-In Self-Test (BIST) ......................................................... 227
12.2 Three-State Test Mode ........................................................... 228
12.3 Boundary-Scan Test Access Port (TAP)................................ 229
12.4 L1 Cache Inhibit...................................................................... 239
12.5 Debug ....................................................................................... 240
13 Clock Control ............................................................................ 247
13.1 Clock Control States ............................................................... 247
13.2 Halt State................................................................................. 249
13.3 Stop Grant State...................................................................... 250
13.4 Stop Grant Inquire State........................................................ 251
13.5 Stop Clock State...................................................................... 252
14 Electrical Data .......................................................................... 253
14.1 Operating Ranges ................................................................... 254
viii Contents
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.2 Absolute Ratings..................................................................... 255
14.3 DC Characteristics .................................................................. 256
14.4 Power Dissipation ................................................................... 258
14.5 Power Derating Based on Lower CPU Frequencies............ 260
14.6 Power and Grounding............................................................. 262
14.7 I/O Buffer Characteristics ...................................................... 264
15 Signal Switching Characteristics ............................................ 267
15.1 CLK Switching Characteristics.............................................. 267
15.2 Clock Switching Characteristics
for 100-MHz Bus Operation.................................................... 268
15.3 Clock Switching Characteristics
for 66-MHz Bus Operation...................................................... 268
15.4 Valid Delay, Float, Setup, and Hold Timings ...................... 269
15.5 Output Delay Timings for 100-MHz Bus Operation............. 270
15.6 Input Setup and Hold Timings
for 100-MHz Bus Operation.................................................... 272
15.7 Output Delay Timings for 66-MHz Bus Operation............... 274
15.8 Input Setup and Hold Timings
for 66-MHz Bus Operation...................................................... 276
15.9 RESET and Test Signal Timing............................................. 278
15.10 Timing Diagrams..................................................................... 281
16 Thermal Design ........................................................................ 285
16.1 Package Thermal Specifications ........................................... 285
16.2 Measuring Case Temperature ............................................... 288
16.3 Sample Heatsink Measured Data.......................................... 289
16.4 Layout and Airflow Considerations ...................................... 295
17 Pin Designation Diagrams ....................................................... 299
17.1 Pin Designations by Functional Grouping ........................... 301
18 Package Specifications ............................................................ 303
18.1 321-Pin Staggered CPGA Package Specification................. 303
19 Ordering Information .............................................................. 305
Index........................................................................................... 307
List of Figures ix
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
List of Figures
Figure 1. AMD-K6™-2E Processor Block Diagram.........................9
Figure 2. Cache Sector Organization.............................................12
Figure 3. The Instruction Buffer ....................................................14
Figure 4. AMD-K6™-2E Processor Decode Logic.........................15
Figure 5. AMD-K6™-2E Processor Scheduler...............................18
Figure 6. Register X and Y Functional Units ...............................20
Figure 7. EAX Register with 16-Bit and 8-Bit
Name Components ..........................................................24
Figure 8. Integer Data Registers....................................................25
Figure 9. Segment Register ............................................................26
Figure 10. Segment Usage ................................................................27
Figure 11. Floating-Point Register...................................................28
Figure 12. FPU Status Word Register .............................................28
Figure 13. FPU Control Word Register...........................................29
Figure 14. FPU Tag Word Register..................................................29
Figure 15. Packed Decimal Data Register ......................................30
Figure 16. Precision Real Data Registers .......................................30
Figure 17. MMX™/3DNow!™ Registers ..........................................31
Figure 18. MMX™ Data Types .........................................................32
Figure 19. 3DNow!™ Data Types .....................................................33
Figure 20. EFLAGS Register............................................................34
Figure 21. Control Register 4 (CR4)................................................35
Figure 22. Control Register 3 (CR3)................................................35
Figure 23. Control Register 2 (CR2)................................................35
Figure 24. Control Register 1 (CR1)................................................36
Figure 25. Control Register 0 (CR0)................................................36
Figure 26. Debug Register DR7 .......................................................37
Figure 27. Debug Register DR6 .......................................................38
Figure 28. Debug Registers DR5 and DR4......................................38
Figure 29. Debug Registers DR3, DR2, DR1, and DR0..................39
Figure 30. Machine-Check Address Register (MCAR) ..................41
Figure 31. Machine-Check Type Register (MCTR)........................41
Figure 32. Test Register 12 (TR12)..................................................42
Figure 33. Time Stamp Counter (TSC)............................................42
Figure 34. Extended Feature Enable Register (EFER).................43
Figure 35. SYSCALL/SYSRET Target Address
Register (STAR) ..............................................................44
Figure 36. Write Handling Control Register (WHCR)...................44
Figure 37. UC/WC Cacheability Control Register (UWCCR) .......45
Figure 38. Processor State Observability Register (PSOR) ..........45
Figure 39. Page Flush/Invalidate Register (PFIR).........................46
Figure 40. Memory Management Registers....................................47
Figure 41. Task State Segment (TSS)..............................................48
Figure 42. 4-Kbyte Paging Mechanism............................................49
Figure 43. 4-Mbyte Paging Mechanism ...........................................50
xList of Figures
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 44. Page Directory Entry 4-Kbyte Page Table (PDE)........51
Figure 45. Page Directory Entry 4-Mbyte Page Table (PDE) .......51
Figure 46. Page Table Entry (PTE)..................................................52
Figure 47. Application Segment Descriptor ...................................53
Figure 48. System Segment Descriptor ...........................................54
Figure 49. Gate Descriptor ...............................................................55
Figure 50. Waveform Definitions...................................................134
Figure 51. Bus State Machine Diagram.........................................135
Figure 52. Non-Pipelined Single-Transfer Memory
Read/Write and Write Delayed by EWBE# ................139
Figure 53. Misaligned Single-Transfer Memory
Read and Write..............................................................141
Figure 54. Burst Reads and Pipelined Burst Reads .....................143
Figure 55. Burst Writeback due to Cache-Line Replacement.....145
Figure 56. Basic I/O Read and Write .............................................146
Figure 57. Misaligned I/O Transfer................................................147
Figure 58. Basic HOLD/HLDA Operation .....................................149
Figure 59. HOLD-Initiated Inquire Hit to Shared
or Exclusive Line...........................................................151
Figure 60. HOLD-Initiated Inquire Hit to Modified Line............153
Figure 61. AHOLD-Initiated Inquire Miss ....................................155
Figure 62. AHOLD-Initiated Inquire Hit to Shared
or Exclusive Line...........................................................157
Figure 63. AHOLD-Initiated Inquire Hit to Modified Line.........159
Figure 64. AHOLD Restriction.......................................................161
Figure 65. BOFF# Timing................................................................163
Figure 66. Basic Locked Operation................................................165
Figure 67. Locked Operation with BOFF# Intervention..............167
Figure 68. Interrupt Acknowledge Operation ..............................169
Figure 69. Basic Special Bus Cycle (Halt Cycle) ..........................171
Figure 70. Shutdown Cycle.............................................................172
Figure 71. Stop Grant and Stop Clock Modes, Part 1 ..................174
Figure 72. Stop Grant and Stop Clock Modes, Part 2 ..................175
Figure 73. INIT-Initiated Transition from Protected Mode
to Real Mode..................................................................177
Figure 74. Cache Organization.......................................................185
Figure 75. Cache Sector Organization...........................................186
Figure 76. Write Handling Control Register (WHCR).................194
Figure 77. Write Allocate Logic Mechanisms and Conditions....195
Figure 78. Page Flush/Invalidate Register (PFIR).......................200
Figure 79. UC/WC Cacheability Control Register (UWCCR) .....208
Figure 80. External Logic for Supporting Floating-Point
Exceptions......................................................................215
Figure 81. SMM Memory.................................................................218
Figure 82. TAP State Diagram .......................................................237
Figure 83. Debug Register DR7 .....................................................241
Figure 84. Debug Register DR6 .....................................................242
List of Figures xi
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 85. Debug Registers DR5 and DR4....................................242
Figure 86. Debug Registers DR3, DR2, DR1, and DR0................243
Figure 87. Clock Control State Transitions...................................248
Figure 88. Suggested Component Placement ...............................263
Figure 89. CLK Waveform ..............................................................269
Figure 90. Key to Timing Diagrams...............................................281
Figure 91. Output Valid Delay Timing..........................................281
Figure 92. Maximum Float Delay Timing .....................................282
Figure 93. Input Setup and Hold Timing ......................................282
Figure 94. Reset and Configuration Timing .................................283
Figure 95. TCK Timing....................................................................284
Figure 96. TRST# Timing................................................................284
Figure 97. Test Signal Timing ........................................................284
Figure 98. Thermal Model ..............................................................286
Figure 99. Power Consumption v. Thermal Resistance...............287
Figure 100. Processor Heat Dissipation Path .................................288
Figure 101. Measuring Case Temperature......................................289
Figure 102. Heatsink A (15 mm height) ..........................................290
Figure 103. Heatsink B (20 mm height)...........................................290
Figure 104. Heatsink C (30 mm height) ..........................................290
Figure 105. Measured Thermal Resistance v. Airflow
(Socketed 321-Pin CPGA Package) .............................291
Figure 106. Measured Maximum Ambient Temperature
(Socketed 321-Pin CPGA Package) .............................292
Figure 107. Measured Thermal Resistance v. Airflow
(Soldered 321-Pin CPGA Package)..............................293
Figure 108. Measured Maximum Ambient Temperature
(Soldered, 321-Pin CPGA Package).............................294
Figure 109. Voltage Regulator Placement......................................295
Figure 110. Airflow for a Heatsink with Fan..................................296
Figure 111. Airflow Path in a Dual-Fan System .............................296
Figure 112. Airflow Path in an ATX Form-Factor System ............297
Figure 113. AMD-K6™-2E Processor Connection Diagram
(Top-Side View CPGA) .................................................299
Figure 114. AMD-K6™-2E Processor Connection Diagram
(Bottom-Side View CPGA)............................................300
Figure 115. 321-Pin Staggered CPGA Package Specification.......303
xii List of Figures
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
List of Tables xiii
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
List of Tables
Table 1. Execution Latency and Throughput of Execution Units .19
Table 2. General-Purpose Registers .................................................24
Table 3. General-Purpose Register Doubleword, Word,
and Byte Names....................................................................25
Table 4. Segment Registers ...............................................................26
Table 5. AMD-K6™-2E Processor Model 8/[F:8]
Model-Specific Registers.....................................................40
Table 6. Extended Feature Enable Register (EFER)Definition ...43
Table 7. SYSCALL/SYSRET Target Address Register
(STAR) Definition................................................................44
Table 8. Memory Management Registers.........................................47
Table 9. Application Segment Types ................................................53
Table 10. System Segment and Gate Types .......................................54
Table 11. Summary of Exceptions and Interrupts.............................55
Table 12. Integer Instructions .............................................................57
Table 13. Floating-Point Instructions .................................................74
Table 14. MMX™ Instructions.............................................................78
Table 15. 3DNow!™ Instructions.........................................................81
Table 16. Processor-to-Bus Clock Ratios ............................................93
Table 17. Output Pin Float Conditions.............................................127
Table 18. Input Pin Types ..................................................................130
Table 19. Output Pin Float Conditions.............................................131
Table 20. Input/Output Pin Float Conditions ..................................131
Table 21. Test Pins..............................................................................131
Table 22. Bus Cycle Definition ..........................................................132
Table 23. Special Cycles.....................................................................132
Table 24. Bus-Cycle Order During Misaligned Memory Transfers . 140
Table 25. A[4:3] Address-Generation Sequence During Bursts .....142
Table 26. Bus-Cycle Order During Misaligned I/O Transfers.........147
Table 27. Interrupt Acknowledge Operation Definition ................168
Table 28. Encodings for Special Bus Cycles.....................................170
Table 29. Output Signal State After RESET....................................180
Table 30. Register State After RESET .............................................181
Table 31. PWT Signal Generation.....................................................188
Table 32. PCD Signal Generation .....................................................189
Table 33. CACHE# Signal Generation..............................................189
Table 34. Data Cache States for Read and Write Accesses............198
Table 35. Cache States for Inquire Cycles, Snoops, Flushes,
and Invalidation .................................................................202
Table 36. Snoop Action ......................................................................203
Table 37. EWBEC Settings.................................................................207
Table 38. WC/UC Memory Type........................................................209
Table 39. Valid Masks and Range Sizes ...........................................210
Table 40. Initial State of Registers in System Management Mode . 219
Table 41. SMM State-Save Area Map ...............................................219
xiv List of Tables
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 42. SMM Revision Identifier...................................................222
Table 43. I/O Trap Doubleword Configuration ................................224
Table 44. I/O Trap Restart Slot .........................................................225
Table 45. Boundary Scan Bit Definitions .........................................233
Table 46. Device Identification Register .........................................234
Table 47. Supported Test Access Port (TAP) Instructions .............235
Table 48. DR7 LEN and RW Definitions ..........................................245
Table 49. Operating Ranges ..............................................................254
Table 50. Absolute Ratings................................................................255
Table 51. DC Characteristics .............................................................256
Table 52. Typical and Maximum Power Dissipation
for OPN Suffix AMZ (Low-Power Devices) .....................258
Table 53. Typical and Maximum Power Dissipation
for OPN Suffix AFR (Standard-Power Devices) .............259
Table 54. Power Derating Specification for Standard-Power
Devices (AMD-K6-2E/233AFR and 266AFR) ..................260
Table 55. Power Derating Specification for Low-Power
Devices (AMD-K6-2E/233AMZ and 266AMZ) .................261
Table 56. CLK Switching Characteristics
for 100-MHz Bus Operation...............................................268
Table 57. CLK Switching Characteristics
for 66-MHz Bus Operation.................................................268
Table 58. Output Delay Timings for 100-MHz Bus Operation........270
Table 59. Input Setup and Hold Timings
for 100-MHz Bus Operation...............................................272
Table 60. Output Delay Timings for 66-MHz Bus Operation..........274
Table 61. Input Setup and Hold Timings for 66-MHz Bus Operation 276
Table 62. RESET and Configuration Signals
for 100-MHz Bus Operation..............................................278
Table 63. RESET and Configuration Signals
for 66-MHz Bus Operation.................................................279
Table 64. TCK Waveform and TRST# Timing at 25 MHz ...............280
Table 65. Test Signal Timing at 25 MHz...........................................280
Table 66. Package Thermal Specification
for OPN Suffix AMZ (Low-Power Devices) .....................285
Table 67. Package Thermal Specification
for OPN Suffix AFR (Standard-Power Devices) .............285
Table 68. Passive Heatsink Samples.................................................289
Table 69. Socketed CPGA Package: Measured Thermal
Resistance (°C/W) qJC and qCA .........................................291
Table 70. Socketed CPGA Package: Measured Maximum
Ambient Temperature (°C) ...............................................292
Table 71. Soldered CPGA Package: Measured Thermal
Resistance (°C/W) qJC and qCA .........................................293
Table 72. Soldered CPGA Package: Measured Maximum
Ambient Temperature (°C) ...............................................294
Table 73. Valid Ordering Part Number Combinations ...................306
Revision History xv
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Revision History
Date Rev Description
June 1999 A Initial published release.
Jan 2000 B Replaced Figure 4, “AMD-K6™-2E Processor Decode Logic,” on page 15 with updated figure.
Jan 2000 B Replaced Table 45 on page 233 with revised boundary scan bit definitions.
Jan 2000 B Changed the Vcc2 maximum specification from 2.6 V to 2.4 V in Table 50, “Absolute Ratings,” on
page 255 for all OPNs with the exception of the 233AFR, 233AMZ, 266AFR, 266AMZ, and 300 AFR,
provided that the processor is not marked with a “7” following the date code.
Jan 2000 B
For the 300AMZ, 333AMZ, and 350AMZ ordering part numbers, added DC characteristics to
Table 51 on page 256, added power dissipation specifications to Table 52 on page 258, added
package thermal specifications to Table 66 on page 285, and added ordering information
beginning on page 305.
Jan 2000 B
For the 333AFR, 350AFR, and 400AFR ordering part numbers, added DC characteristics to Table 51
on page 256, added power dissipation specifications to Table 53 on page 259, added package
thermal specifications to Table 67 on page 285, and added ordering information beginning on
page 305.
Jan 2000 B Added power derating specifications beginning on page 260.
Jan 2000 B Added sample measured heat sink data beginning on page 289.
xvi Revision History
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
About this Data Sheet xvii
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
About this Data Sheet
The AMD-K6™-2E Processor Data Sheet is the complete specification of the
AMD-K6-2E embedded processor.
Overview
This data sheet is organized into the following sections:
Chapter 1, “AMD-K6™-2E Processor” on page 1, provides a list of the AMD-K6-2E
processor’s distinguishing characteristics, a description of the key features, and a
discussion about the Super7™ platform initiative.
Chapter 2, “Internal Architecture” on page 7, describes the functional elements of the
advanced design techniques, known as the RISC86® microarchitecture, implemented
by the AMD-K6-2E processor.
Chapter 3, “Software Environment” on page 23, provides a general overview of the
AMD-K6-2E processor’s x86 software environment and briefly describes the data
types, registers, operating modes, interrupts, and instructions supported by the
AMD-K6-2E processor’s architecture and design implementation.
Chapter 4, “Logic Symbol Diagram” on page 83, contains the AMD-K6-2E processor
logic symbol diagram.
Chapter 5, “Signal Descriptions” on page 85, lists the signals and their descriptions
alphabetically and by function.
Chapter 6, “Bus Cycles” on page 133, describes and illustrates the timing and
relationship of bus signals during various types of bus cycles.
Chapter 7, “Power-On Configuration and Initialization” on page 179, describes how
the system logic resets the AMD-K6-2E processor using the RESET signal.
Chapter 8, “Cache Organization” on page 185, describes the basic architecture and
resources of the AMD-K6-2E processor’s internal caches.
Chapter 9, “Write Merge Buffer” on page 205, describes the 8-byte write merge buffer
and how merging multiple write cycles into a single write cycle ultimately increases
overall system performance.
Chapter 10, “Floating-Point and Multimedia Execution Units” on page 213, describes
the AMD-K6-2E processor’s IEEE 754-compatible and 854-compatible floating point
xviii
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
execution unit, the multimedia and 3DNow!™ technology execution units, and the
floating-point and MMX/3DNow! technology instruction compatibility.
Chapter 11, “System Management Mode (SMM)” on page 217, describes SMM, the
state-save area, entry into and exit from SMM, exceptions and interrupts in SMM,
memory allocation and addressing in SMM, and the SMI# and SMIACT# signals.
Chapter 12, “Test and Debug” on page 227, describes the various test and debug modes
that enable the functional and manufacturing testing of systems and boards that use
the AMD-K6-2E processor and that allow designers to debug the instruction execution
of software components.
Chapter 13, “Clock Control” on page 247, describes the five modes of clock control
supported by the AMD-K6-2E processor.
Chapter 14, “Electrical Data” on page 253, includes operating ranges, absolute
ratings, DC characteristics, power dissipation data, power and grounding information,
decoupling recommendations, and I/O buffer characteristics.
Chapter 15, “Signal Switching Characteristics” on page 267, provides tables listing
valid delay, float, setup, and hold timing specifications for the AMD-K6-2E processor
signals.
Chapter 16, “Thermal Design” on page 285, lists the package thermal specifications,
discusses how to measure case temperature, and provides sample heat sink
measurement data, along with layout and airflow considerations.
Chapter 17, “Pin Designation Diagrams” on page 299, lists the AMD-K6-2E processor’s
pin designations by functional grouping.
Chapter 18, “Package Specifications” on page 303, provides a table and diagram
containing the 321-pin CPGA package specifications.
Chapter 19, “Ordering Information” on page 305, provides the ordering part number
(OPN) and valid OPN combinations.
Chapter 1 AMD-K6™-2E Processor 1
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
1AMD-K6™-2E Processor
The following are key features of the AMD-K6™-2E processor:
■Advanced 6-Issue RISC86® Superscalar Microarchitecture
•Ten parallel specialized execution units
•Multiple sophisticated x86-to-RISC86 instruction decoders
•Advanced two-level branch prediction
•Speculative execution
•Out-of-order execution
•Register renaming and data forwarding
•Up to six RISC86 instructions per clock
■Large on-chip split 64-Kbyte level-one (L1) cache
•32-Kbyte instruction cache with additional 20 Kbytes of predecode cache
•32-Kbyte writeback dual-ported data cache
•Two-way set associative
•MESI protocol support
■3DNow!™ technology
•Additional instructions to improve 3D graphics and multimedia performance
•Separate multiplier and ALU for superscalar instruction execution
■321-pin ceramic pin grid array (CPGA) package
■Socket 7 platform compatible, 66-MHz frontside bus
■Super7™ platform compatible, 100-MHz frontside bus supported on the 300-MHz,
350-MHz, and 400-MHz versions of the AMD-K6-2E processor
■High-performance industry-standard MMX™ instructions
•Dual integer ALU for superscalar execution
■High-performance IEEE 754-compatible and 854-compatible floating-point unit
■Industry-standard system management mode (SMM)
■IEEE 1149.1 boundary scan
■x86 binary software compatibility
■Low-power 0.25-micron process technology
•Split-plane power with support for full 3.3 V I/O
•Available with a low-power 1.9-V core voltage and extended temperature rating
or with a standard-power 2.2-V core voltage and standard temperature rating
2AMD-K6™-2E Processor Chapter 1
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
1.1 AMD-K6™-2E Embedded Processor Features
The AMD-K6-2E processor with 3DNow!™ technology is a functionally compatible
embedded version of the sixth generation, Microsoft® Windows® compatible
AMD-K6-2 processor. The AMD-K6-2E embedded processor delivers the same high
performance and incorporates the same leading-edge features, including the
innovative and efficient RISC86® microarchitecture, a large 64-Kbyte level-one cache
(32-Kbyte dual-ported data cache, 32-Kbyte instruction cache with predecode data),
and a powerful IEEE 754-compatible and 854-compatible floating-point execution
unit.
The AMD-K6-2E embedded processor also supports the new features incorporated
into the AMD-K6-2 processor. These features include superscalar MMX™ instruction
execution support, support for the Super7™ 100-MHz frontside bus, and AMD’s
innovative 3DNow!™ technology for high-performance multimedia and 3D graphics
operation based on high-performance single instruction multiple data (SIMD)
execution resources.
The AMD-K6-2E embedded processor includes several key features that are very
beneficial to the embedded market. The AMD-K6-2E processor offers leading-edge
performance for embedded systems requiring compatibility with the extensive
installed base of x86 software. The AMD-K6-2E processor’s Socket 7 and Super7
platform-compatible, 321-pin ceramic pin grid array (CPGA) package allows the
product designer to reduce time-to-market by leveraging today’s cost-effective
industry-standard infrastructure to deliver a superior-performing embedded solution.
The AMD-K6-2E embedded processor is available in two versions.
■The low-power version has a 1.9-V core voltage and extended temperature rating.
■The standard-power version has a 2.2-V core voltage and is the embedded
equivalent of the industry-standard desktop version of the AMD-K6-2 processor.
System Management Mode and Power Management Features
The AMD-K6-2E processor includes the complete industry-standard system
management mode (SMM), which is critical to system resource and power
management. (See “System Management Mode (SMM)” on page 217 for more
detailed information about this feature.)
The AMD-K6-2E processor also features the industry-standard Stop-Clock (STPCLK#)
control circuitry and the Halt instruction, both required for implementing the ACPI
power management specification. (“Clock Control” on page 247 provides more
information on these power management features.)
Chapter 1 AMD-K6™-2E Processor 3
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Microarchitecture
The AMD-K6-2E processor’s RISC86 microarchitecture is a decoupled
decode/execution superscalar design that implements state-of-the-art design
techniques to achieve leading-edge performance.
Advanced design techniques implemented in the AMD-K6-2E processor include
multiple x86 instruction decode, single-clock internal RISC operations, ten execution
units that support superscalar operation, out-of-order execution, data forwarding,
speculative execution, and register renaming.
In addition, the processor supports advanced branch prediction logic by
implementing an 8192-entry branch history table, a branch target cache, and a return
address stack, which combine to deliver better than a 95% prediction rate. These
design techniques enable the AMD-K6-2E to issue, execute, and retire multiple x86
instructions per clock, resulting in excellent scalable performance. The
microarchitecture of the AMD-K6-2E processor is more completely described in
“Internal Architecture” on page 7.
3DNow!™ Technology
AMD’s 3DNow! technology is an instruction-set extension to x86, which includes 21
new instructions to accelerate 3D graphics and other single-precision floating-point
compute intensive operations.
Improvements include fast frame rates on high-resolution graphics applications,
superior modeling of real-world environments and physics, life-like images, graphics,
and audio.
AMD has already shipped millions of processors with 3DNow! technology for desktop
and notebook PCs, revolutionizing the 3D experience with up to four times the peak
floating-point performance of previous sixth generation solutions. AMD is now
bringing this advanced capability to embedded systems.
AMD has taken a leadership role in developing these new instructions that enable
exciting new levels of performance and realism. 3DNow! technology was defined and
implemented in collaboration with Microsoft, application developers, and graphics
vendors, and has received an enthusiastic reception. It is compatible with today’s
existing x86 software, is supported by industry-standard APIs, and requires no
operating system support, thereby enabling a broad class of applications to benefit
from 3DNow! technology.
Industry-Standard x86 Architecture
The AMD-K6-2E processor is x86 binary code compatible. AMD’s extensive
experience through six generations of x86 processors has been carefully integrated
into the processor to enable compatibility with Windows®-based operating systems,
including Windows 95, Windows 98, Windows CE, Windows NT®, and Windows NTE.
4AMD-K6™-2E Processor Chapter 1
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The AMD-K6-2E processor is also compatible with DOS, OS/2, UNIX, and other
leading operating systems, including real-time operating systems (RTOS) commonly
used in embedded applications such as pSOS, QNX, RTXC, and VxWorks.
The AMD-K6-2E processor is compatible with more than 60,000 software
applications, including the latest software optimized for 3DNow! and MMX
technologies. AMD has shipped more than 120 million x86 microprocessors, including
more than 60 million Windows-compatible processors.
The AMD-K6-2E processor is among a long line of Microsoft Windows compatible
processors from AMD. The combination of state-of-the-art features, leading-edge
performance, high-performance multimedia engine, x86 compatibility, and low-cost
infrastructure enable decreased development costs and improved time-to-market,
making the AMD-K6-2E processor the superior choice for embedded systems.
1.2 Process Technology
The AMD-K6-2E processor is implemented using an advanced CMOS 0.25-micron
process technology that utilizes a split core and I/O voltage supply, which allows the
core of the processor to operate at a low voltage while the I/O portion operates at the
industry-standard 3.3 V.
This technology enables high performance while reducing power consumption by
operating the core at a low voltage and limiting power requirements to the acceptable
levels for today’s embedded systems.
1.3 Super7™ Platform Initiative
All AMD-K6-2E processors remain pin compatible with existing Socket 7 solutions;
however, for maximum system performance, the 300-MHz, 350-MHz, and 400-MHz
versions of the processor work optimally in Super7 designs that incorporate advanced
features such as support for the 100-MHz frontside bus and AGP graphics.
AMD and its industry partners are investing in the future of Socket 7 with the new
Super7 platform initiative. The goal of the initiative is to maintain the competitive
vitality of the Socket 7 infrastructure through a series of enhancements, including
the development of an industry-standard 100-MHz processor bus protocol. In addition
to the 100-MHz processor bus protocol, the Super7 initiative includes the
introduction of chipsets that support the AGP specification and support for a
backside L2 cache and frontside L3 cache.
Chapter 1 AMD-K6™-2E Processor 5
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Super7™ Platform Enhancements
■100-MHz processor bus—The AMD-K6-2E processor supports a 100-MHz, 800
Mbyte/second frontside bus to provide a high-speed interface to Super7
platform-based chipsets. The 100-MHz interface to the frontside Level 2 (L2)
cache and main system memory speeds up access to the frontside cache and main
memory by 50 percent over the 66-MHz Socket 7 interface, resulting in a
significant increase of 10% in overall system performance.
■Accelerated graphics port support—AGP improves the performance of video
graphics systems that have small amounts of video memory on the graphics card.
The industry-standard AGP specification enables a 133-MHz graphics interface
and will scale to even higher levels of performance.
■Support for backside L2 and frontside L3 cache—The Super7 platform has the
‘headroom’ to support higher-performance AMD-K6 processors with clock speeds
scaling to 500 MHz and beyond. The Super7 platform also supports the
AMD-K6-III processor, which features a full-speed, internal backside 256-Kbyte L2
cache designed to deliver new levels of system performance to desktop and
notebook PC systems. The AMD-K6-III processor also supports an optional
100-MHz frontside L3 cache for even higher-performance system configurations.
Super7™ Platform Advantages
The Super7 platform has the following advantages:
■Delivers performance and features competitive with alternate platforms at the
same clock speed, and at a significantly lower cost
■Takes advantage of existing system designs for superior value
■Enables OEMs and resellers to take advantage of mature, high-volume
infrastructure supported by multiple BIOS, chipset, graphics, and motherboard
suppliers
■Reduces inventory and design costs with one motherboard for a wide range of
products
■Builds on a huge installed base of more than 100 million motherboards
■Provides an easy upgrade path for future embedded applications, as well as a
bridge to legacy applications
By taking advantage of the low-cost, mature Socket 7 infrastructure, the Super7
platform will continue to provide superior value and leading-edge performance for
embedded systems.
6AMD-K6™-2E Processor Chapter 1
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 2 Internal Architecture 7
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
2 Internal Architecture
The AMD-K6-2E processor implements advanced design
techniques known as the RISC86 microarchitecture. The
RISC86 microarchitecture is a decoupled decode/execution
design approach that yields superior sixth-generation
performance for x86-based software. This chapter describes the
techniques used and the functional elements of the RISC86
microarchitecture.
2.1 AMD-K6™-2E Processor Microarchitecture Overview
When discussing processor design, it is important to understand
the terms architecture, microarchitecture, and design
implementation.
■Architecture refers to the instruction set and features of a
processor that are visible to software programs running on
the processor. The architecture determines which software
the processor can run. The architecture of the AMD-K6-2E
processor is the industry-standard x86 instruction set.
■Microarchitecture refers to the design techniques used in the
processor to reach the target cost, performance, and
functionality goals. The AMD-K6-2E processor is based on a
sophisticated RISC core known as the Enhanced RISC86
microarchitecture. The Enhanced RISC86 microarchitecture
is an advanced, second-order decoupled decode/execution
design approach that enables industry-leading performance
for x86-based software.
■Design implementation refers to the actual logic and circuit
designs from which the processor is created according to the
microarchitecture specifications.
8Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Enhanced RISC86®
Microarchitecture The Enhanced RISC86 microarchitecture defines the
characteristics of the AMD-K6-2E processor. The innovative
RISC86 microarchitecture approach implements the x86
instruction set by internally translating x86 instructions into
RISC86 operations. These RISC86 operations were specially
designed to include direct support for the x86 instruction set
while observing the RISC performance principles of fixed
length encoding, regularized instruction fields, and a large
register set.
The Enhanced RISC86 microarchitecture used in the
AMD-K6-2E processor enables higher processor core
performance and promotes straightforward extensibility in
future designs. Instead of directly executing complex x86
instructions, which have lengths of 1 to 15 bytes, the
AMD-K6-2E processor executes the simpler and easier
fixed-length RISC86 opcodes, while maintaining the instruction
coding efficiencies found in x86 programs.
The AMD-K6-2E processor contains parallel decoders, a
centralized RISC86 operation scheduler, and ten execution
units that support superscalar operation—multiple decode,
execution, and retirement—of x86 instructions. These elements
are packed into an aggressive and highly efficient six-stage
pipeline.
AMD-K6™-2E
Processor Block
Diagram
As shown in Figure 1 on page 9, the high-performance,
out-of-order execution engine of the AMD-K6-2E processor is
mated to a split level-one 64-Kbyte writeback cache with 32
Kbytes of instruction cache and 32 Kbytes of data cache. The
instruction cache feeds the decoders and, in turn, the decoders
feed the scheduler. The Instruction Control Unit (ICU) issues
and retires RISC86 operations contained in the scheduler. The
system bus interface is an industry-standard 64-bit Super7 and
Socket 7 demultiplexed bus.
The AMD-K6-2E processor combines the latest in processor
microarchitecture to provide the highest x86 performance for
today’s computational systems. The AMD-K6-2E offers true
sixth-generation performance and x86 binary software
compatibility.
Chapter 2 Internal Architecture 9
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 1. AMD-K6™-2E Processor Block Diagram
Decoders Decoding of the x86 instructions begins when the on-chip
instruction cache is filled. Predecode logic determines the
length of an x86 instruction on a byte-by-byte basis. This
predecode information is stored, along with the x86
instructions, in the instruction cache, to be used later by the
decoders. The decoders translate on-the-fly, with no additional
latency, up to two x86 instructions per clock into RISC86
operations.
Note: In this chapter, “clock” refers to a processor clock.
The AMD-K6-2E processor categorizes x86 instructions into
three types of decodes—short, long, and vector. The decoders
process either two short, one long, or one vector decode at a
time.
Store
Unit Branch
Unit
Store
Queue
Instruction
Control Unit
Scheduler
Buffer
(24 RISC86)
Six RISC86 ®
Operation Issue
Four RISC86
Decode
Out-of-Order
Execution Engine
32-KByte Level-One Dual-Port Data Cache 128-Entry DTLB
20-KByte Predecode Cache 64-Entry ITLB
Multiple Instruction Decoders
x86 to RISC86
Branch Logic
(8192-Entry BHT)
(16-Entry BTC)
(16-Entry RAS)
16-Byte Fetch
Load
Unit
Predecode
Logic
Level-One Cache
Controller
FPU
32-KByte Level-One Instruction Cache
Register Y Functional Units
Integer/
Multimedia /3DNow!
100 MHz
Super7
Bus
Interface
Register X Functional Units
Integer/
Multimedia/3DNow!
É
10 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The three types of decodes have the following characteristics:
■Short decodes—x86 instructions that are less than or equal
to seven bytes long
■Long decodes—x86 instructions less than or equal to 11
bytes long
■Vector decodes—complex x86 instructions
Short and long decodes are processed completely within the
decoders. Vector decodes are started by the decoders and then
completed by fetched sequences from an on-chip ROM. After
decoding, the RISC86 operations are delivered to the scheduler
for dispatching to the execution units.
Scheduler/Instruction
Control Unit The centralized scheduler or buffer is managed by the ICU. The
ICU buffers and manages up to 24 RISC86 operations at a time.
This equals from 6 to 12 x86 instructions. This buffer size (24) is
perfectly matched to the processor’s six-stage RISC86 pipeline,
four RISC86-operations decode rate, and ten parallel execution
units.
The scheduler accepts as many as four RISC86 operations at a
time from the decoders and retires up to four RISC86
operations per clock cycle. The ICU is capable of
simultaneously issuing up to six RISC86 operations at a time to
the execution units. This consists of the following types of
operations:
■Memory load operation
■Memory store operation
■Complex integer, MMX, or 3DNOW! register operation
■Simple integer register operation
■Floating-point register operation
■Branch condition evaluation
Registers When managing the RISC86 operations, the ICU uses 69
physical registers contained within the RISC86
microarchitecture.
■Forty-eight of the physical registers are located in a general
register file.
•Twenty-four of these are rename registers.
•The other twenty-four are committed or architectural
registers, consisting of 16 scratch registers and 8 registers
Chapter 2 Internal Architecture 11
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
that correspond to the x86 general-purpose registers—
EAX, EBX, ECX, EDX, EBP, ESP, ESI, and EDI.
■An analogous set of 21 registers is available specifically for
MMX and 3DNow! operations.
•Twelve of these are MMX/3DNow! rename registers.
•Nine are MMX/3DNow! committed or architectural
registers, consisting of one scratch register and eight
registers that correspond to the MMX registers (mm0–
mm7). For more detailed information, see the 3DNow!™
Technology Manual, order #21928.
Branch Logic The AMD-K6-2E processor is designed with highly
sophisticated dynamic branch logic consisting of the following:
■Branch history/Prediction table
■Branch target cache
■Return address stack
The AMD-K6-2E processor implements a two-level branch
prediction scheme based on an 8192-entry branch history table.
The branch history table stores prediction information that is
used for predicting conditional branches. Because the branch
history table does not store predicted target addresses, special
address ALUs calculate target addresses on-the-fly during
instruction decode.
The branch target cache augments predicted branch
performance by avoiding a one clock cache-fetch penalty. This
specialized target cache does this by supplying the first 16 bytes
of target instructions to the decoders when branches are
predicted. The return address stack is a unique device
specifically designed for optimizing CALL and RETURN pairs.
In summary, the AMD-K6-2E processor uses dynamic branch
logic to minimize delays due to the branch instructions that are
common in x86 software.
3DNow!™ Technology AMD has taken a lead role in improving the multimedia and 3D
capabilities of the x86 processor family with the introduction of
3DNow! technology, which uses a packed, single-precision,
floating-point data format and Single Instruction Multiple Data
(SIMD) operations also found in the MMX technology model.
12 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
2.2 Cache, Instruction Prefetch, and Predecode Bits
The writeback level-one cache on the AMD-K6-2E processor is
organized as a separate 32-Kbyte instruction cache and a
32-Kbyte data cache with two-way set associativity. The cache
line size is 32 bytes and lines are prefetched from main memory
using an efficient pipelined burst transaction.
As the instruction cache is filled, each instruction byte is
analyzed for instruction boundaries using predecoding logic.
Predecoding annotates each instruction byte with information
(5 bits per byte) that later enables the decoders to efficiently
decode multiple instructions simultaneously.
Cache The processor cache design takes advantage of a sectored
organization (see Figure 2). Each sector consists of 64 bytes
configured as two 32-byte cache lines. The two cache lines of a
sector share a common tag but have separate pairs of MESI
(Modified, Exclusive, Shared, Invalid) bits that track the state
of each cache line.
Figure 2. Cache Sector Organization
Two forms of cache misses and associated cache fills can take
place—a tag-miss cache fill and a tag-hit cache fill.
■Tag-miss cache fill—The miss is due to a tag mismatch, in
which case the required cache line is filled from external
memory, and the cache line within the sector that was not
required is marked as invalid.
■Tag-hit cache fill—The address matches the tag, but the
requested cache line is marked as invalid. The required
cache line is filled from external memory, and the cache line
within the sector that is not required remains in the same
cache state.
Prefetching The AMD-K6-2E processor conditionally performs cache
prefetching which results in the filling of the required cache
line first, and a prefetch of the second cache line making up the
other half of the sector. From the perspective of the external
bus, the two cache-line fills typically appear as two 32-byte
Tag Address Cache Line 0 Byte 31 Predecode Bits Byte 30 Predecode Bits ........ ........ Byte 0 Predecode Bits MESI Bits
Cache Line 1 Byte 31 Predecode Bits Byte 30 Predecode Bits ........ ........ Byte 0 Predecode Bits MESI Bits
Chapter 2 Internal Architecture 13
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
burst read cycles occurring back-to-back or, if allowed, as
pipelined cycles.
The 3DNow! technology includes a new instruction named
PREFETCH that allows a cache line to be prefetched into the
data cache. The PREFETCH instruction format is defined in
Table 15, “3DNow!™ Instructions,” on page 81. For more
detailed information, see the 3DNow!™ Technology Manual,
order #21928.
Predecode Bits Decoding x86 instructions is particularly difficult because the
instructions are variable in length (1 to 15 bytes). Predecode
logic supplies the five predecode bits associated with each
instruction byte. The predecode bits indicate the number of
bytes to the start of the next x86 instruction. The predecode
bits are stored in an extended instruction cache alongside each
x86 instruction byte, as shown in Figure 2 on page 12. The
predecode bits are passed with the instruction bytes to the
decoders where they assist with parallel x86 instruction
decoding.
2.3 Instruction Fetch and Decode
Instruction Fetch The processor can fetch up to 16 bytes per clock out of the
instruction cache or branch target cache. The fetched
information is placed into a 16-byte instruction buffer that
feeds directly into the decoders (see Figure 3 on page 14).
Fetching can occur along a single execution stream with up to
seven outstanding branches taken.
The instruction fetch logic is capable of retrieving any 16
contiguous bytes of information within a 32-byte boundary.
There is no additional penalty when the 16 bytes of instructions
lie across a cache line boundary. The instruction bytes are
loaded into the instruction buffer as they are consumed by the
decoders.
Although instructions can be consumed with byte granularity,
the instruction buffer is managed on a memory-aligned word
(two bytes) organization. Therefore, instructions are loaded and
replaced with word granularity. When a control transfer
occurs—such as a JMP instruction—the entire instruction
buffer is flushed and reloaded with a new set of 16 instruction
bytes.
14 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 3. The Instruction Buffer
Instruction Decode The AMD-K6-2E processor decode logic is designed to decode
multiple x86 instructions per clock (see Figure 4 on page 15).
The decode logic accepts x86 instruction bytes and their
predecode bits from the instruction buffer, locates the actual
instruction boundaries, and generates RISC86 operations from
these x86 instructions.
RISC86 operations are fixed-format internal instructions. Most
RISC86 operations execute in a single clock. RISC86 operations
are combined to perform every function of the x86 instruction
set. Some x86 instructions are decoded into as few as zero
RISC86 opcodes, for instance a NOP, or one RISC86 operation,
a register-to-register add. More complex x86 instructions are
decoded into several RISC86 operations.
16 Instruction Bytes
plus
16 Sets of Predecode Bits
Branch-Target Cache
16 x 16 Bytes
2:1
Instruction Buffer
16 Bytes
16 Bytes
Branch Target
Address Adders
Return Address Stack
16 x 16 Bytes
32-Kbyte Level-One
Instruction Cache
Fetch Unit
Chapter 2 Internal Architecture 15
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 4. AMD-K6™-2E Processor Decode Logic
The AMD-K6-2E processor uses a combination of decoders to
convert x86 instructions into RISC86 operations. The hardware
consists of three sets of decoders—two parallel short decoders,
one long decoder, and one vector decoder.
Parallel Short Decoders. The two parallel short decoders translate
the most commonly-used x86 instructions (moves, shifts,
branches, ALU, FPU) and the extensions to the x86 instruction
set (MMX and 3DNow! technology) into zero, one, or two
RISC86 operations each. The short decoders only operate on
x86 instructions that are up to seven bytes long. In addition, the
decoders are designed to decode up to two x86 instructions per
clock.
Instruction Buffer
4 RISC86 Operations
Long Decoder
Short Decoder #1
Short Decoder #2
Vector Address
Vector Decoder
RISC86® Sequencer
On-Chip ROM
16 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Long Decoder. The commonly-used x86 instructions that are
greater than seven bytes but not more than 11 bytes long, and
the x86 instructions that are slightly less common and are up to
seven bytes long are handled by the long decoder. The long
decoder only performs one decode per clock and generates up
to four RISC86 operations.
Vector Decoder. All other translations (complex instructions,
serializing conditions, interrupts and exceptions, etc.) are
handled by a combination of the vector decoder and RISC86
operation sequences fetched from an on-chip ROM. For
complex operations, the vector decoder logic provides the first
set of RISC86 operations and a vector (initial ROM address) to a
sequence of further RISC86 operations. The same types of
RISC86 operations are fetched from the ROM as those that are
generated by the hardware decoders.
Note: Although all three sets of decoders are simultaneously fed a
copy of the instruction buffer contents, only one of the three
types of decoders is used during any one decode clock.
Grouped Operations. The decoders or the RISC86 sequencer
always generate a group of four RISC86 operations. For decodes
that cannot fill the entire group with four RISC86 operations,
RISC86 NOP operations are placed in the empty locations of
the grouping. For example, a long-decoded x86 instruction that
converts to only three RISC86 operations is padded with a
single RISC86 NOP operation and then passed to the scheduler.
Up to six groups, or 24 RISC86 operations, can be placed in the
scheduler at a time.
Floating Point Instructions. All of the common, and a few of the
uncommon, floating-point instructions (also known as ESC
instructions) are hardware decoded as short decodes. This
decode generates a RISC86 floating-point operation and,
optionally, an associated floating-point load or store operation.
Floating-point or ESC instruction decode is only allowed in the
first short decoder, but non-ESC instructions, excluding MMX
instructions, can be decoded simultaneously by the second
short decoder along with an ESC instruction decode in the first
short decoder.
MMX and 3DNow!™ Instructions. All of the MMX and 3DNow!
instructions, with the exception of the EMMS, FEMMS, and
PREFETCH instructions, are hardware decoded as short
Chapter 2 Internal Architecture 17
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
decodes. The MMX instruction decode generates a RISC86
MMX operation and, optionally, an associated MMX load or
store operation. A 3DNow! instruction decode generates a
RISC86 3DNow! operation and, optionally, an associated load or
store operation. MMX and 3DNow! instructions can be decoded
in either or both of the short decoders.
2.4 Centralized Scheduler
The scheduler is the heart of the AMD-K6-2E processor (see
Figure 5 on page 18). The scheduler contains the logic necessary
to manage out-of-order execution, data forwarding, register
renaming, simultaneous issue and retirement of multiple
RISC86 operations, and speculative execution.
The scheduler’s buffer can hold up to 24 RISC86 operations.
This equates to a maximum of 12 x86 instructions. When
possible, the scheduler can simultaneously issue a RISC86
operation to any available execution unit (store, load, branch,
integer, integer/multimedia, or floating-point). In total, the
scheduler can issue up to six and retire up to four RISC86
operations per clock.
The main advantage of the scheduler and its operation buffer is
the ability to examine an x86 instruction window equal to 12
x86 instructions at one time. This advantage is due to the fact
that the scheduler operates on the RISC86 operations in
parallel and allows the AMD-K6-2E processor to perform
dynamic on-the-fly instruction code scheduling for optimized
execution. Although the scheduler can issue RISC86 operations
for out-of-order execution, it always retires x86 instructions in
order.
18 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 5. AMD-K6™-2E Processor Scheduler
2.5 Execution Units
The AMD-K6-2E processor contains ten parallel execution
units—store, load, integer X ALU, integer Y ALU, MMX ALU
(X), MMX ALU (Y), MMX/3DNow! multiplier, 3DNow! ALU,
floating-point, and branch condition. Each unit is independent
and capable of handling the RISC86 operations. Table 1 on
page 19 details the execution units, functions performed within
these units, operation latency, and operation throughput.
The store and load execution units are two-stage pipelined
designs.
■The store unit performs data writes and register calculation
for LEA/PUSH. Data memory and register writes from stores
are available after one clock. Store operations are held in a
store queue prior to execution. From there, they execute in
order.
■The load unit performs data memory reads. Data is available
from the load unit after two clocks.
RISC86 Operation Buffer
RISC86 Issue Buses
RISC86 #0 RISC86 #1 RISC86 #2 RISC86 #3
Centralized RISC86®
Operation Scheduler
From Decode Logic
Chapter 2 Internal Architecture 19
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The Integer X execution unit can operate on all ALU
operations, multiplies, divides (signed and unsigned), shifts,
and rotates.
The Integer Y execution unit can operate on the basic word and
doubleword ALU operations—ADD, AND, CMP, OR, SUB,
XOR, zero-extend, and sign-extend operands.
The functional units that execute MMX and 3DNow!
instructions share pipeline control with the Integer X and
Integer Y units.
Register X and Y
Pipelines The register X and Y functional units are attached to the issue
bus for the register X execution pipeline or the issue bus for the
register Y execution pipeline or both. Each register pipeline
has dedicated resources that consist of an integer execution
unit and an MMX ALU execution unit, therefore allowing
superscalar operation on integer and MMX instructions. In
addition, both the X and Y issue buses are connected to the
3DNow! ALU, the MMX/3DNow! multiplier, and MMX shifter,
which allows the appropriate RISC86 operation to be issued
through either bus. Figure 6 on page 20 shows the details of the
X and Y register pipelines.
Table 1. Execution Latency and Throughput of Execution Units
Functional Unit Function Latency Throughput
Store LEA/PUSH, Address (Pipelined) 1 1
Memory Store (Pipelined) 1 1
Load Memory Loads (Pipelined) 2 1
Integer X
Integer ALU 1 1
Integer Multiply 2–3 2–3
Integer Shift 1 1
Multimedia
(processes
MMX instructions)
MMX ALU 1 1
MMX Shifts, Packs, Unpack 1 1
MMX Multiply 2 1
Integer Y Basic ALU (16-bit and 32-bit operands) 1 1
Branch Resolves Branch Conditions 1 1
FPU FADD, FSUB, FMUL 2 2
3DNow!
3DNow! ALU 2 1
3DNow! Multiply 2 1
3DNow! Convert 2 1
20 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 6. Register X and Y Functional Units
The branch condition unit is separate from the branch
prediction logic in that it resolves conditional branches such as
JCC and LOOP after the branch condition has been evaluated.
2.6 Branch-Prediction Logic
Sophisticated branch logic that can minimize or hide the impact
of changes in program flow is designed into the AMD-K6-2E
processor. Branches in x86 code fit into two categories:
■Unconditional branches always change program flow (that is,
the branches are always taken)
■Conditional branches may or may not divert program flow
(that is, the branches are taken or not-taken). When a
conditional branch is not taken, the processor simply
continues decoding and executing the next instructions in
memory.
MMX/
3DNow!É
Multiplier
Integer X
ALU MMXÉ
ALU MMX
Shifter 3DNow!
ALU MMX
ALU Integer Y
ALU
Scheduler
Buffer
(24 RISC86® Operations)
Issue Bus
for the
Register X
Execution
Pipeline
Issue Bus
for the
Register Y
Execution
Pipeline
Chapter 2 Internal Architecture 21
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Typical applications have up to 10% of unconditional branches
and another 10% to 20% conditional branches. The AMD-K6-2E
processor branch logic has been designed to handle this type of
program behavior and its negative effects on instruction
execution, such as stalls due to delayed instruction fetching and
the draining of the processor pipeline. The branch logic
contains an 8192-entry branch history table, a 16-entry by
16-byte branch target cache, a 16-entry return address stack,
and a branch execution unit.
Branch History Table The AMD-K6-2E processor handles unconditional branches
without any penalty by redirecting instruction fetching to the
target address of the unconditional branch. However,
conditional branches require the use of the dynamic
branch-prediction mechanism built into the AMD-K6-2E
processor.
A two-level adaptive history algorithm is implemented in an
8192-entry branch history table. This table stores executed
branch information, predicts individual branches, and predicts
the behavior of groups of branches.
To accommodate the large branch history table, the AMD-K6-2E
processor does not store predicted target addresses. Instead,
the branch target addresses are calculated on-the-fly using
ALUs during the decode stage. The adders calculate all
possible target addresses before the instructions are fully
decoded, and the processor chooses which addresses are valid.
Branch Target Cache To avoid a one clock cache-fetch penalty when a branch is
predicted taken, a built-in branch target cache supplies the first
16 bytes of instructions directly to the instruction buffer
(assuming the target address hits this cache). (See Figure 3 on
page 14.)
The branch target cache is organized as 16 entries of 16 bytes.
In total, the branch prediction logic achieves branch prediction
rates greater than 95%.
Return Address Stack The return address stack is a special device designed to
optimize CALL and RET pairs. Software is typically compiled
with subroutines that are frequently called from various places
in a program. This is usually done to save space.
22 Internal Architecture Chapter 2
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Entry into the subroutine occurs with the execution of a CALL
instruction. At that time, the processor pushes the address of
the next instruction in memory following the CALL instruction
onto the stack (allocated space in memory). When the processor
encounters a RET instruction (within or at the end of the
subroutine), the branch logic pops the address from the stack
and begins fetching from that location. To avoid the latency of
main memory accesses during CALL and RET operations, the
return address stack caches the pushed addresses.
Branch Execution
Unit The branch execution unit enables efficient speculative
execution. This unit gives the processor the ability to execute
instructions beyond conditional branches before knowing
whether the branch prediction was correct.
The AMD-K6-2E processor does not permanently update the
x86 registers or memory locations until all speculatively
executed conditional branch instructions are resolved. When a
prediction is incorrect, the processor backs out to the point of
the mispredicted branch instruction and restores all registers.
The AMD-K6-2E processor can support up to seven outstanding
branches.
Chapter 3 Software Environment 23
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
3 Software Environment
This chapter provides a general overview of the AMD-K6-2E
processor’s x86 software environment and briefly describes the
data types, registers, operating modes, interrupts, and
instructions supported by the AMD-K6-2E architecture and
design implementation.
The stepping of the Model 8 version of the processor
determines the implementation and format of ten Model-
Specific Registers (MSRs).
The AMD-K6-2E processor supports Model 8 steppings [F:8] in
any of eight possible model/steppings—Models 8/8, 8/9, 8/A, 8/B,
8/C, 8/D, 8/E, or 8/F.
Note that the name AMD-K6-2E processor by itself refers to all
steppings of the Model 8/[F:8] version.
3.1 Registers
The AMD-K6-2E processor contains all the registers defined by
the x86 architecture, including general-purpose, segment,
floating-point, MMX/3DNow!, EFLAGS, control, task, debug,
test, and descriptor/memory-management registers.
In addition to information about these registers, this chapter
provides information on the AMD-K6-2E processor MSRs.
Note: Areas of the register designated as Reserved should not be
modified by software.
24 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
General-Purpose
Registers The eight 32-bit x86 general-purpose registers are used to hold
integer data or memory pointers used by instructions. Table 2
contains a list of the general-purpose registers and the
functions for which they are used.
To support byte and word operations, EAX, EBX, ECX, and
EDX can also be used as 8-bit and 16-bit registers. The shorter
registers are overlaid on the longer ones. For example, the
name of the 16-bit version of EAX is AX (low 16 bits of EAX)
and the 8-bit names for AX are AH (high order bits) and AL (low
order bits). The same naming convention applies to EBX, ECX,
and EDX.
EDI, ESI, ESP, and EBP can be used as smaller 16-bit registers
called DI, SI, SP, and BP respectively, but these registers do not
have 8-bit versions. Figure 7 shows the EAX register with its
name components, and Table 3 on page 25 lists the doubleword
(32-bit) general-purpose registers and their corresponding word
(16-bit) and byte (8-bit) versions.
Figure 7. EAX Register with 16-Bit and 8-Bit Name Components
Table 2. General-Purpose Registers
Register Function
EAX Commonly used as an accumulator
EBX Commonly used as a pointer
ECX Commonly used for counting in loop operations
EDX Commonly used to hold I/O information and to pass parameters
EDI Commonly used as a destination pointer by the ES segment
ESI Commonly used as a source pointer by the DS segment
ESP Used to point to the stack segment
EBP Used to point to data within the stack segment
87 0151631
EAX
AX
AH AL
Chapter 3 Software Environment 25
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Integer Data Types Four types of data are used in general-purpose registers—byte,
word, doubleword, and quadword integers. Figure 8 shows the
format of the integer data registers.
Figure 8. Integer Data Registers
Table 3. General-Purpose Register Doubleword, Word, and Byte Names
32-Bit Name
(Doubleword) 16-Bit Name
(Word) 8-Bit Name
(High-order Bits) 8-Bit Name
(Low-order Bits)
EAX AX AH AL
EBX BX BH BL
ECX CX CH CL
EDX DX DH DL
EDI DI ––
ESI SI – –
ESP SP – –
EBP BP – –
15 0
31 0
Precision — 32 Bits
Precision — 16 Bits
Word Integer
Doubleword Integer
70
Precision —
8 Bits
Byte Integer
63 0
Precision — 64 Bits
Quadword Integer
26 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Segment Registers The six 16-bit segment registers are used as pointers to areas
(segments) of memory. Table 4 lists the segment registers and
their functions. Figure 9 shows the format for all six segment
registers.
Figure 9. Segment Register
Segment Usage The operating system determines the type of memory model
that is implemented. The segment register usage is determined
by the operating system’s memory model. In a real mode
memory model, the segment register points to the base address
in memory.
In a protected mode memory model, the segment register is
called a selector and it selects a segment descriptor in a
descriptor table. This descriptor contains a pointer to the base
of the segment, the limit of the segment, and various protection
attributes. For more information on descriptor formats, see
“Descriptors and Gates” on page 52. Figure 10 on page 27 shows
segment usage for real mode and protected mode memory
models.
Table 4. Segment Registers
Segment
Register Segment Register Function
CS Code segment, where instructions are located
DS Data segment, where data is located
ES Data segment, where data is located
FS Data segment, where data is located
GS Data segment, where data is located
SS Stack segment
015
Chapter 3 Software Environment 27
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 10. Segment Usage
Instruction Pointer The instruction pointer (EIP or IP) is used in conjunction with
the code segment register (CS). The instruction pointer is
either a 32-bit register (EIP) or a 16-bit register (IP) that keeps
track of where the next instruction resides within memory. This
register cannot be directly manipulated, but can be altered by
modifying return pointers when a JMP or CALL instruction is
used.
Floating-Point
Registers The floating-point execution unit in the AMD-K6-2E processor
is designed to perform mathematical operations on non-integer
numbers. This floating-point unit conforms to the IEEE 754 and
854 standards and uses several registers to meet these
standards—eight numeric floating-point registers, a status
word register, a control word register, and a tag word register.
Segment Register
Real Mode Memory Model
Segment Selector
Physical Memory
Protected Mode Memory Model
Base
Descriptor Table
Physical Memory
Segment Base
Base
Limit
BaseLimit
Segment Base
28 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The eight floating-point registers are physically 80 bits wide
and labeled FPR0–FPR7. Figure 11 shows the format of these
floating-point registers. See “Floating-Point Register Data
Types” on page 30 for information on allowable floating-point
data types.
Figure 11. Floating-Point Register
The 16-bit FPU status word register contains information about
the state of the floating-point unit. Figure 12 shows the format
of the FPU status word register.
Figure 12. FPU Status Word Register
64 63 07879
Sign Exponent Significand
9876543210101112131415
P
EO
E
E
S
C
0
C
1I
E
Z
E
U
E
S
F
TOSP
C
3
BC
2D
E
Symbol Description Bits
BFPU Busy15
C3 Condition Code 14
TOSP Top of Stack Pointer 13–11
C2 Condition Code 10
C1 Condition Code 9
C0 Condition Code 8
ES Error Summary Status 7
SF Stack Fault 6
Exception Flags
PE Precision Error 5
UE Underflow Error 4
OE Overflow Error 3
ZE Zero Divide Error 2
DE Denormalized Operation Error 1
IE Invalid Operation Error 0
TOSP Information
000 = FPR0
111 = FPR7
Chapter 3 Software Environment 29
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The FPU control word register allows a programmer to manage
the FPU processing options. Figure 13 shows the format of the
FPU control word register.
Figure 13. FPU Control Word Register
The FPU tag word register contains information about the
registers in the register stack. Figure 14 shows the format of the
FPU tag word register.
Figure 14. FPU Tag Word Register
9876543210101112131415
P
MO
M
P
C
R
CI
M
Z
M
U
MD
M
Rounding Control Information
00b = Round to the nearest or even number
01b = Round down toward negative infinity
10b = Round up toward positive infinity
11b = Truncate toward zero
Y
Precision Control Information
00b = 24 bits Single Precision Real
01b = Reserved
10b = 53 bits Double Precision Real
11b = 64 bits Extended Precision Real
Reserved
Symbol Description Bits
Y Infinity Bit (80287 compatibility) 12
RC Rounding Control 11–10
PC Precision Control 9–8
Exception Masks
PM Precision 5
UM Underflow 4
OM Overflow 3
ZM Zero Divide 2
DM Denormalized Operation 1
IM Invalid Operation 0
Tag Values
00 = Valid
01 = Zero
10 = Special
11 = Empty
87 65 43 2 1 0101112131415
TAG
(FPR7) TAG
(FPR6) TAG
(FPR4)
TAG
(FPR5) TAG
(FPR2)
TAG
(FPR3) TAG
(FPR0)
TAG
(FPR1)
9
30 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Floating-Point
Register Data Types Floating-point registers use four different types of data—
packed decimal, single-precision real, double-precision real,
and extended-precision real. Figures 15 and 16 show the
formats for these registers.
Figure 15. Packed Decimal Data Register
Figure 16. Precision Real Data Registers
079
Precision — 18 Digits, 72 Bits Used, 4-Bits/Digit
71
SIgnore
or
Zero
78 72
Description Bits
Ignored on Load, Zeros on Store 78-72
Sign Bit 79
063
Double-Precision Real
31 0
Single-Precision Real
079
22
SBiased
Exponent
78
23
SBiased
Exponent
6364
5152
Biased
Exponent
S
Significand
Significand
Significand
30
62
Extended-Precision Real
S = Sign Bit
S = Sign Bit
S = Sign Bit
I
62
I = Integer Bit
Chapter 3 Software Environment 31
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
MMX™/3DNow!™
Registers The AMD-K6-2E processor implements eight 64-bit
MMX/3DNow! registers for use by multimedia software. These
registers are mapped on the floating-point register stack. The
MMX and 3DNow! instructions refer to these registers as mm0
to mm7. Figure 17 shows the format of these registers. For more
information, see the AMD-K6® Processor Multimedia
Technology Manual, order #20726 and the 3DNow! Technology
Manual, order #21928.
Figure 17. MMX™/3DNow!™ Registers
63 0
mm0
mm7
mm1
mm6
mm5
mm2
mm3
mm4
32 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
MMX™ Data Types For the MMX instructions, the MMX registers use three types of
data—packed 8-byte integer, packed quadword integer, and
packed dual doubleword integer. Figure 18 shows the format of
the MMX data types.
Figure 18. MMX™ Data Types
63 0
Packed Bytes Integer
63 0
Packed Words Integer
63 0
Packed Doubleword Integer
32 31
48 47 32 31 16 15
56 55 48 47 40 39 32 31 24 23 16 15 8 7
Byte 7 Byte 6 Byte 5 Byte 4 Byte 3 Byte 2 Byte 1 Byte 0
Word 0Word 1Word 2Word 3
Doubleword 1 Doubleword 0
Chapter 3 Software Environment 33
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
3DNow!™ Data Types For 3DNow! instructions, the MMX/3DNow! registers use
packed single-precision real data. Figure 19 shows the format of
the 3DNow! data type.
Figure 19. 3DNow!™ Data Types
63 62 0
32 31 30
Packed Single Precision Floating Point
55 54 23 22
Biased
Exponent
SSignificand
S = Sign Bit
Biased
Exponent
SSignificand
S = Sign Bit
34 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
EFLAGS Register The EFLAGS register provides for three different types of
flags—system, control, and status. The system flags provide
operating system controls, the control flag provides directional
information for string operations, and the status flags provide
information resulting from logical and arithmetic operations.
Figure 20 shows the format of the EFLAGS register.
Figure 20. EFLAGS Register
9876543210101112131415161718192021
I
O
P
L
31 30 29 28 27 26 25 24 23 22
A
FP
F
Z
F
S
F
I
F
D
FT
F
O
F
N
T
R
F
V
M
A
C
V
I
F
V
I
P
I
DC
F
Reserved
Symbol Description Bits
ID ID Flag 21
VIP Virtual Interrupt Pending 20
VIF Virtual Interrupt Flag 19
AC Alignment Check 18
VM Virtual-8086 Mode 17
RF Resume Flag 16
NT Nested Task 14
IOPL I/O Privilege Level 13–12
OF Overflow Flag 11
DF Direction Flag 10
IF Interrupt Flag 9
TF Trap Flag 8
SF Sign Flag 7
ZF Zero Flag 6
AF Auxiliary Flag 4
PF Parity Flag 2
CF Carry Flag 0
Chapter 3 Software Environment 35
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Control Registers The five control registers contain system control bits and
pointers. Figures 21 through 25 show the formats of the control
registers.
Figure 21. Control Register 4 (CR4)
Figure 22. Control Register 3 (CR3)
Figure 23. Control Register 2 (CR2)
7654321031
P
S
E
T
S
D
M
C
E
V
M
E
D
E
P
V
I
Reserved
Symbol Description Bit
MCE Machine Check Enable 6
PSE Page Size Extensions 4
DE Debugging Extensions 3
TSD Time Stamp Disable 2
PVI Protected Virtual Interrupts 1
VME Virtual-8086 Mode Extensions 0
P
C
D
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Page Directory Base P
W
T
Reserved
Symbol Description Bit
PCD Page Cache Disable 4
PWT Page Writethrough 3
031
Page Fault Linear Address
36 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 24. Control Register 1 (CR1)
Figure 25. Control Register 0 (CR0)
031
Reserved
E
TT
S
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
A
ME
M
W
PM
PP
E
N
E
P
GC
DN
W
Reserved
Symbol Description Bit
AM Alignment Mask 18
WP Write Protect 16
NE Numeric Error 5
ET Extension Type 4
TS Task Switched 3
EM Emulation 2
MP Monitor Co-processor 1
PE Protection Enabled 0
Symbol Description Bit
PG Paging 31
CD Cache Disable 30
NW Not Writethrough 29
Chapter 3 Software Environment 37
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Debug Registers Figures 26 through 29 show the 32-bit debug registers
supported by the processor. These registers are further
described in “Debug” on page 240.
Figure 26. Debug Register DR7
9876543210101112131415
L
2L
1
L
3
G
3
G
EL
EL
0
G
0
G
1
L
2
G
D
25 24 23 22 21 20 19 18 17 16262728293031
R/W
3
LEN
3R/W
2
LEN
2R/W
1
LEN
1R/W
0
LEN
0
Reserved
Symbol Description Bit
GD General Detect Enabled 13
GE Global Exact Breakpoint Enabled 9
LE Local Exact Breakpoint Enabled 8
G3 Global Exact Breakpoint # 3 Enabled 7
L3 Local Exact Breakpoint # 3 Enabled 6
G2 Global Exact Breakpoint # 2 Enabled 5
L2 Local Exact Breakpoint # 2 Enabled 4
G1 Global Exact Breakpoint # 1 Enabled 3
L1 Local Exact Breakpoint # 1 Enabled 2
G0 Global Exact Breakpoint # 0 Enabled 1
L0 Local Exact Breakpoint # 0 Enabled 0
Symbol Description Bits
LEN 3 Length of Breakpoint #3 31–30
R/W 3 Type of Transaction(s) to Trap 29–28
LEN 2 Length of Breakpoint #2 27–26
R/W 2 Type of Transaction(s) to Trap 25–24
LEN 1 Length of Breakpoint #1 23–22
R/W 1 Type of Transaction(s) to Trap 21–20
LEN 0 Length of Breakpoint #0 19–18
R/W 0 Type of Transaction(s) to Trap 17–16
38 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 27. Debug Register DR6
Figure 28. Debug Registers DR5 and DR4
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
B
1
B
2
B
SB
0
B
TB
DB
3
Reserved
Symbol Description Bit
BT Breakpoint Task Switch 15
BS Breakpoint Single Step 14
BD Breakpoint Debug Access Detected 13
B3 Breakpoint #3 Condition Detected 3
B2 Breakpoint #2 Condition Detected 2
B1 Breakpoint #1 Condition Detected 1
B0 Breakpoint #0 Condition Detected 0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Reserved
DR5
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Reserved
DR4
Chapter 3 Software Environment 39
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 29. Debug Registers DR3, DR2, DR1, and DR0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 3 32-bit Linear Address
DR3
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 0 32-bit Linear Address
DR0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 2 32-bit Linear Address
DR2
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 1 32-bit Linear Address
DR1
40 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
3.2 Model-Specific Registers (MSR)
The AMD-K6-2E processor is based on and functionally
identical to the AMD-K6-2 processor Model 8/[F:8], which
provides ten model-specific registers (MSRs).
■The value in the ECX register selects the MSR to be
addressed by the RDMSR and WRMSR instructions.
■The values in EAX and EDX are used as inputs and outputs
by the RDMSR and WRMSR instructions.
Table 5 lists the MSRs and the corresponding value of the ECX
register. Figures 30 through 39 show the MSR formats.
For more information about the MSRs, see the AMD-K6®
Processor BIOS Design Application Note, order #21329.
For more information about the RDMSR and WRMSR
instructions, see the AMD K86™ Family BIOS and Software Tools
Development Guide, order #21062.
Table 5. AMD-K6™-2E Processor Model 8/[F:8] Model-Specific Registers
Model-Specific Register Value of ECX
Machine-Check Address Register (MCAR) 00h
Machine-Check Type Register (MCTR) 01h
Test Register 12 (TR12) 0Eh
Time Stamp Counter (TSC) 10h
Extended Feature Enable Register (EFER) C000_0080h
SYSCALL/SYSRET Target Address Register (STAR) C000_0081h
Write Handling Control Register (WHCR) C000_0082h
UC/WC Cacheability Control Register (UWCCR) C000_0085h
Processor State Observability Register (PSOR) C000_0087h
Page Flush/Invalidate Register (PFIR) C000_0088h
Chapter 3 Software Environment 41
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
MCAR and MCTR The AMD-K6-2E processor does not support the generation of a
machine-check exception. However, the processor does provide
a 64-bit machine-check address register (MCAR), a 64-bit
machine-check type register (MCTR), and a machine check
enable (MCE) bit in CR4.
Because the processor does not support machine check
exceptions, the contents of the MCAR and MCTR registers are
only affected by the WRMSR instruction and by RESET being
sampled asserted (where all bits in each register are reset to 0).
The formats for the machine-check address register and the
machine-check type register are shown in Figure 30 and Figure
31, respectively.
Figure 30. Machine-Check Address Register (MCAR)
Figure 31. Machine-Check Type Register (MCTR)
0
63
MCAR
54 063
MCTR
Reserved
42 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Test Register 12
(TR12) Test register 12 provides a method for disabling the L1 caches.
Figure 32 shows the format of the TR12 register.
Figure 32. Test Register 12 (TR12)
Time Stamp Counter With each processor clock cycle, the processor increments the
64-bit time stamp counter (TSC) MSR. Figure 33 shows the
format of the TSC register.
Figure 33. Time Stamp Counter (TSC)
421063
C
I
3
Reserved Symbol Description Bit
CI Cache Inhibit Bit 3
063
TSC
Chapter 3 Software Environment 43
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Extended Feature
Enable Register
(EFER)
The extended feature enable register (EFER) contains the
control bits that enable the extended features of the
AMD-K6-2E processor. Figure 34 shows the format of the EFER
register, and Table 6 defines the function of each bit in the
EFER register.
Figure 34. Extended Feature Enable Register (EFER)
For more information about the EWBEC bits, see “EWBE#
Control” on page 205.
Table 6. Extended Feature Enable Register (EFER)Definition
Bit Description R/W Function
63–4 Reserved R Writing a 1 to any reserved bit causes a general protection fault to occur. All
reserved bits are always read as 0.
3-2 EWBE# Control
(EWBEC)
R/W This 2-bit field controls the behavior of the processor with respect to the ordering of
write cycles and the EWBE# signal. EFER[3] and EFER[2] are Global EWBE# Disable
(GEWBED) and Speculative EWBE# Disable (SEWBED), respectively.
1 Data Prefetch Enable
(DPE)
R/W DPE must be set to 1 to enable data prefetching (this is the default setting following
reset). If enabled, cache misses initiated by a memory read within a 32-byte cache
line are conditionally followed by cache-line fetches of the other line in the 64-byte
sector.
0 System Call Extension
(SCE)
R/W SCE must be set to 1 to enable the usage of the SYSCALL and SYSRET instructions.
1063
S
C
E
234
D
P
E
EWBEC
Symbol Description Bit
EWBEC EWBE# Control 3-2
DPE Data Prefetch Enable 1
SCE System Call Extension 0
Reserved
44 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
SYSCALL/SYSRET
Target Address
Register (STAR)
The SYSCALL/SYSRET target address register (STAR)
contains the target EIP address used by the SYSCALL
instruction and the 16-bit code and stack segment selector
bases used by the SYSCALL and SYSRET instructions. Figure
35 shows the format of the STAR register, and Table 7 defines
the function of each bit of the STAR register. For more
information, see the SYSCALL and SYSRET Instruction
Specification Application Note, order #21086.
Figure 35. SYSCALL/SYSRET Target Address Register (STAR)
Write Handling
Control Register
(WHCR)
The write handling control register (WHCR) is an MSR that
contains two fields—the write allocate enable limit (WAELIM)
field, and the write allocate enable 15-to-16-Mbyte (WAE15M)
bit. Figure 36 shows the format of the WHCR register. See
“Write Allocate” on page 192 for more information.
Figure 36. Write Handling Control Register (WHCR)
31 063
Target EIP Address
32
47
48
SYSCALL CS Selector and SS
Selector Base
SYSRET CS Selector and SS
Selector Base
Table 7. SYSCALL/SYSRET Target Address Register (STAR) Definition
Bit Description R/W
63–48 SYSRET CS and SS Selector Base R/W
47–32 SYSCALL CS and SS Selector Base R/W
31–0 Target EIP Address R/W
1522 063
WAELIM
16
Note: Hardware RESET initializes this MSR to all zeros.
W
A
E
1
5
M
Symbol Description Bits
WAELIM Write Allocate Enable Limit 31-22
WAE15M Write Allocate Enable 15-to-16-Mbyte 16
17
213132
Reserved
Chapter 3 Software Environment 45
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
UC/WC Cacheability
Control Register
(UWCCR)
The AMD-K6-2E processor provides two variable-range Memory
Type Range Registers (MTRRs)—MTRR0 and MTRR1—that
each specify a range of memory. Each range can be defined as
uncacheable (UC) or write-combining (WC) memory. Figure 37
shows the format of the UWCCR register. For more detailed
information about the MTRR0, MTRR1, and UWCCR registers,
see “Memory Type Range Registers” on page 207.
.
Figure 37. UC/WC Cacheability Control Register (UWCCR)
Processor State
Observability
Register (PSOR)
The AMD-K6-2E processor provides the processor state
observability register (PSOR) (see Figure 38).
Figure 38. Processor State Observability Register (PSOR)
16 063
Physical Address Mask 0
1731
Physical Base Address 0
12
Physical Address Mask 1Physical Base Address 1
3233344849
U
C
0
W
C
0
U
C
1
W
C
1
MTRR1 MTRR0
Symbol Description Bits
UC0 Uncacheable Memory Type 0
WC0 Write-Combining Memory Type 1
Symbol Description Bits
UC1 Uncacheable Memory Type 32
WC1 Write-Combining Memory Type 33
20
63
BF
Symbol Description Bit
NOL2 No L2 Functionality 8
STEP Processor Stepping 7-4
BF Bus Frequency Divisor 2-0
34
STEP
789
N
O
L
2
Reserved
46 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Page Flush/Invalidate
Register (PFIR) The AMD-K6-2E processor contains the Page Flush/Invalidate
Register (PFIR) (see Figure 39) that allows cache invalidation
and optional flushing of a specific 4-Kbyte page from the linear
address space. Using this register can result in a much lower
cycle count for flushing particular pages versus flushing the
entire cache. When the PFIR is written to (using the WRMSR
instruction), the invalidation and, optionally, the flushing
begins.
Figure 39. Page Flush/Invalidate Register (PFIR)
LINPAGE
10
63
F
/
I
Symbol Description Bit
LINPAGE 20-bit Linear Page Address 31-12
PF Page Fault Occurred 8
F/I Flush/Invalidate Command 0
1131 1232
P
F
987
Reserved
Chapter 3 Software Environment 47
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
3.3 Memory Management Registers
The AMD-K6-2E processor controls segmented memory
management with the registers listed in Table 8. Figure 40
shows the formats of the memory management registers.
Figure 40. Memory Management Registers
Table 8. Memory Management Registers
Register Name Function
Global Descriptor Table Register Contains a pointer to the base of the global descriptor table
Interrupt Descriptor Table Register Contains a pointer to the base of the interrupt descriptor table
Local Descriptor Table Register Contains a pointer to the local descriptor table of the current task
Task Register Contains a pointer to the task state segment of the current task
15 0
16-Bit Limit
16
47
32-Bit Linear Base Address
Global and Interrupt Descriptor Table Registers
31 063
32-Bit Limit
32
32-Bit Linear Base Address
15 0
Local Descriptor Table Register and Task Register
Attributes
15 0
Selector
48 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Task State Segment Figure 41 shows the format of the task state segment (TSS).
Figure 41. Task State Segment (TSS)
31
Interrupt Redirection Bitmap (IRB)
(eight 32-bit locations)
0
I/O Permission Bitmap (IOPB)
(up to 8 Kbytes)
Operating System
Data Structure
Base Address of IOPB
LDT Selector
0000h
0000h
0000h
0000h
0000h
0000h
0000h
GS
FS
DS
SS
CS
ES
EDI
ESI
EBP
ESP
EBX
EDX
ECX
EAX
CR3
EFLAGS
EIP
0000h
0000h
0000h
0000h
SS2
SS1
SS0
Link (Prior TSS Selector)
ESP0
ESP1
ESP2
TSS Limit
from TR
64h
0
T
0000h
Chapter 3 Software Environment 49
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
3.4 Paging
The AMD-K6-2E processor can physically address up to four
Gbytes of memory. This memory can be segmented into pages.
The size of these pages is determined by the operating system
design and the values set up in the page directory entries (PDE)
and page table entries (PTE).
The processor can access both 4-Kbyte pages and 4-Mbyte
pages, and the page sizes can be intermixed within a page
directory. When the page size extension (PSE) bit in CR4 is set,
the processor translates linear addresses using either the
4-Kbyte translation lookaside buffer (TLB) or the 4-Mbyte TLB,
depending on the state of the page size (PS) bit in the page
directory entry. Figures 42 and 43 show how 4-Kbyte and
4-Mbyte page translations work.
Figure 42. 4-Kbyte Paging Mechanism
Linear Address
Page
Directory
Page
Table
4-Kbyte
Page
Frame
CR3
011122131 22
Page Directory
Offset
Page Table
Offset
Page
Offset
PDE
PTE
Physical
Address
50 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 43. 4-Mbyte Paging Mechanism
Figures 44 through 46 show the formats of the PDE and PTE.
These entries contain information regarding the location of
pages and their status.
Linear Address
Page
Directory
4-Mbyte
Page
Frame
CR3
02131 22
Page Directory
Offset
Page
Offset
PDE
Physical
Address
Chapter 3 Software Environment 51
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 44. Page Directory Entry 4-Kbyte Page Table (PDE)
Figure 45. Page Directory Entry 4-Mbyte Page Table (PDE)
876543210
31
P
C
D
U
/
S
W
/
R
9101112
A
V
L0 A P
W
TP
Page Table Base Address
Symbol Description Bits
AVL Available to Software 11–9
Reserved 8
PS Page Size 7
Reserved 6
A Accessed 5
PCD Page Cache Disable 4
PWT Page Writethrough 3
U/S User/Supervisor 2
W/R Write/Read 1
P Present (valid) 0
876543210
31
P
C
D
U
/
S
W
/
R
9101112
A
V
L1 A P
W
TP
Physical Page Base Address Reserved
2122
Symbol Description Bits
AVL Available to Software 11–9
Reserved 8
PS Page Size 7
Reserved 6
A Accessed 5
PCD Page Cache Disable 4
PWT Page Writethrough 3
U/S User/Supervisor 2
W/R Write/Read 1
P Present (valid) 0
52 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 46. Page Table Entry (PTE)
3.5 Descriptors and Gates
There are various types of structures and registers in the x86
architecture that define, protect, and isolate code segments,
data segments, task state segments, and gates. These structures
are called descriptors.
■The application segment descriptor is used to point to either a
data or code segment. Figure 47 on page 53 shows the
application segment descriptor format. Table 9 contains
information describing the memory segment type to which
the descriptor points.
■The system segment descriptor is used to point to a task state
segment, a call gate, or a local descriptor table. Figure 48 on
page 54 shows the system segment descriptor format. Table
10 contains information describing the type of segment or
gate to which the descriptor points.
■The AMD-K6-2E processor uses gates to transfer control
between executable segments with different privilege
levels. Figure 49 on page 55 shows the format of the gate
descriptor types. Table 10 contains information describing
the type of segment or gate to which the descriptor points.
876543210
31
P
C
D
U
/
S
W
/
R
9101112
A
V
LAP
W
TP
Physical Page Base Address D
Symbol Description Bits
AVL Available to Software 11–9
Reserved 8–7
D Dirty 6
A Accessed 5
PCD Page Cache Disable 4
PWT Page Writethrough 3
U/S User/Supervisor 2
W/R Write/Read 1
P Present (valid) 0
Chapter 3 Software Environment 53
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 47. Application Segment Descriptor
Base Address 15–0 Segment Limit 15–0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Segment
Limit PDPL 1Type
A
V
L
G DBase Address 31–24 Base Address 23–16
Reserved
Symbol Description Bits
G Granularity 23
D 32-Bit/16-Bit 22
AVL Available to Software 20
P Present/Valid Bit 15
DPL Descriptor Privilege Level 14-13
DT Descriptor Type 12
Type See Table 9 11-8
Table 9. Application Segment Types
Type Data/Code Description
0
Data
Read-Only
1Read-Only—Accessed
2 Read/Write
3 Read/Write—Accessed
4 Read-Only—Expand-down
5 Read-Only—Expand-down, Accessed
6 Read/Write—Expand-down
7 Read/Write—Expand-down, Accessed
8
Code
Execute-Only
9 Execute-Only—Accessed
A Execute/Read
B Execute/Read—Accessed
C Execute-Only—Conforming
D Execute-Only—Conforming, Accessed
E Execute/Read-Only—Conforming
F Execute/Read-Only—Conforming, Accessed
54 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 48. System Segment Descriptor
Base Address 15–0 Segment Limit 15–0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Segment
Limit PDPL 0Type
A
V
L
G XBase Address 31–24 Base Address 23–16
Reserved
Symbol Description Bits
G Granularity 23
X Not Needed 22
AVL Availability to Software 20
P Present/Valid Bit 15
DPL Descriptor Privilege Level 14-13
DT Descriptor Type 12
Type See Table 10 11-8
Table 10. System Segment and Gate Types
Type Description
0 Reserved
1 Available 16-bit TSS
2LDT
3Busy 16-bit TSS
4 16-bit Call Gate
5Task Gate
6 16-bit Interrupt Gate
7 16-bit Trap Gate
8 Reserved
9 Available 32-bit TSS
A Reserved
B Busy 32-bit TSS
C 32-bit Call Gate
D Reserved
E 32-bit Interrupt Gate
F 32-bit Trap Gate
Chapter 3 Software Environment 55
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 49. Gate Descriptor
3.6 Exceptions and Interrupts
Table 11 summarizes the exceptions and interrupts.
DPL 0TypeOffset 31–16 P
Segment Selector Offset 15–0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Reserved
Symbol Description Bits
P Present/Valid Bit 15
DPL Descriptor Privilege Level 14-13
DT Descriptor Type 12
Type See Table 10 11-8
Table 11. Summary of Exceptions and Interrupts
Interrupt
Number Interrupt Type Cause
0 Divide by Zero Error DIV, IDIV
1 Debug Debug trap or fault
2 Non-Maskable Interrupt NMI signal sampled asserted
3 Breakpoint Int 3
4 Overflow INTO
5 Bounds Check BOUND
6 Invalid Opcode Invalid instruction
7 Device Not Available ESC and WAIT
8 Double Fault Fault occurs while handling a fault
9 Reserved - Interrupt 13 —
10 Invalid TSS Task switch to an invalid segment
11 Segment Not Present Instruction loads a segment and present bit is 0 (invalid segment)
12 Stack Segment Stack operation causes limit violation or present bit is 0
13 General Protection Segment related or miscellaneous invalid actions
14 Page Fault Page protection violation or a reference to missing page
16 Floating-Point Error Arithmetic error generated by floating-point instruction
17 Alignment Check Data reference to an unaligned operand. (The AC flag and the AM bit of CR0 are set to 1.)
0–255 Software Interrupt INT n
56 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
3.7 Instructions Supported by the AMD-K6™-2E Processor
This section documents all of the x86 instructions supported by
the AMD-K6™-2E processor. Tables 12 through 15 define the
integer, floating-point, MMX, and 3DNow! instructions for the
AMD-K6-2E processor, respectively.
Each table shows the instruction mnemonic, opcode, modR/M
byte, decode type, and RISC86 operation(s) for each
instruction.
Instruction
Mnemonic and
Operand Types
The first column in each table indicates the instruction
mnemonic and operand types, with the following notations:
■disp16/32—16-bit or 32-bit displacement value
■disp32/48—doubleword or 48-bit displacement value
■disp8—8-bit displacement value
■eXX—register width depending on the operand size
■imm16/32—16-bit or 32-bit immediate value
■imm8—8-bit immediate value
■mem16/32—word or doubleword integer value in memory
■mem32/48—doubleword or 48-bit integer value in memory
■mem32real—32-bit floating-point value in memory
■mem48—48-bit integer value in memory
■mem64—64-bit value in memory
■mem64real—64-bit floating-point value in memory
■mem8—byte integer value in memory
■mem80real—80-bit floating-point value in memory
■mmreg—MMX/3DNow! register
■mmreg1—MMX/3DNow! register defined by bits 5, 4, and 3
of the modR/M byte
■mmreg2—MMX/3DNow! register defined by bits 2, 1, and 0
of the modR/M byte
■mreg16/32—word or doubleword integer register, or word or
doubleword integer value in memory defined by the
modR/M byte
■mreg8—byte integer register or byte integer value in
memory defined by the modR/M byte
■reg8—byte integer register defined by instruction byte(s) or
bits 5, 4, and 3 of the modR/M byte
Chapter 3 Software Environment 57
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
■reg16/32—word or doubleword integer register defined by
instruction byte(s) or bits 5, 4, and 3 of the modR/M byte
Opcode Bytes The second and third columns list all applicable opcode bytes.
ModR/M Byte The fourth column lists the modR/M byte when used by the
instruction. The modR/M byte defines the instruction as a
register or memory form. If modR/M bits 7 and 6 are documented
as mm (memory form), mm can only be 10b, 01b or 00b.
Decode Type The fifth column lists the type of instruction decode—short,
long, and vector. The AMD-K6-2E processor decode logic can
process two short, one long, or one vector decode per clock.
RISC86 Operation The sixth column lists the type of RISC86 operation(s) required
for the instruction. The operation types and corresponding
execution units are as follows:
■alu—either of the integer execution units
■alux—integer X execution unit only
■branch—branch condition unit
■float—floating-point execution unit
■limm—load immediate, instruction control unit
■load, fload, mload—load unit
■meu—Multimedia execution units for MMX and 3DNow!
instructions
■store, fstore, mstore—store unit
Table 12. Integer Instructions
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
AAA 37h vector
AAD D5h 0Ah vector
AAM D4h 0Ah vector
AAS 3Fh vector
ADC mreg8, reg8 10h 11-xxx-xxx vector
ADC mem8, reg8 10h mm-xxx-xxx vector
ADC mreg16/32, reg16/32 11h 11-xxx-xxx vector
ADC mem16/32, reg16/32 11h mm-xxx-xxx vector
ADC reg8, mreg8 12h 11-xxx-xxx vector
ADC reg8, mem8 12h mm-xxx-xxx vector
58 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
ADC reg16/32, mreg16/32 13h 11-xxx-xxx vector
ADC reg16/32, mem16/32 13h mm-xxx-xxx vector
ADC AL, imm8 14h vector
ADC EAX, imm16/32 15h vector
ADC mreg8, imm8 80h 11-010-xxx vector
ADC mem8, imm8 80h mm-010-xxx vector
ADC mreg16/32, imm16/32 81h 11-010-xxx vector
ADC mem16/32, imm16/32 81h mm-010-xxx vector
ADC mreg16/32, imm8 (signed ext.) 83h 11-010-xxx vector
ADC mem16/32, imm8 (signed ext.) 83h mm-010-xxx vector
ADD mreg8, reg8 00h 11-xxx-xxx short alux
ADD mem8, reg8 00h mm-xxx-xxx long load, alux, store
ADD mreg16/32, reg16/32 01h 11-xxx-xxx short alu
ADD mem16/32, reg16/32 01h mm-xxx-xxx long load, alu, store
ADD reg8, mreg8 02h 11-xxx-xxx short alux
ADD reg8, mem8 02h mm-xxx-xxx short load, alux
ADD reg16/32, mreg16/32 03h 11-xxx-xxx short alu
ADD reg16/32, mem16/32 03h mm-xxx-xxx short load, alu
ADD AL, imm8 04h short alux
ADD EAX, imm16/32 05h short alu
ADD mreg8, imm8 80h 11-000-xxx short alux
ADD mem8, imm8 80h mm-000-xxx long load, alux, store
ADD mreg16/32, imm16/32 81h 11-000-xxx short alu
ADD mem16/32, imm16/32 81h mm-000-xxx long load, alu, store
ADD mreg16/32, imm8 (signed ext.) 83h 11-000-xxx short alux
ADD mem16/32, imm8 (signed ext.) 83h mm-000-xxx long load, alux, store
AND mreg8, reg8 20h 11-xxx-xxx short alux
AND mem8, reg8 20h mm-xxx-xxx long load, alux, store
AND mreg16/32, reg16/32 21h 11-xxx-xxx short alu
AND mem16/32, reg16/32 21h mm-xxx-xxx long load, alu, store
AND reg8, mreg8 22h 11-xxx-xxx short alux
AND reg8, mem8 22h mm-xxx-xxx short load, alux
AND reg16/32, mreg16/32 23h 11-xxx-xxx short alu
AND reg16/32, mem16/32 23h mm-xxx-xxx short load, alu
AND AL, imm8 24h short alux
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 59
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
AND EAX, imm16/32 25h short alu
AND mreg8, imm8 80h 11-100-xxx short alux
AND mem8, imm8 80h mm-100-xxx long load, alux, store
AND mreg16/32, imm16/32 81h 11-100-xxx short alu
AND mem16/32, imm16/32 81h mm-100-xxx long load, alu, store
AND mreg16/32, imm8 (signed ext.) 83h 11-100-xxx short alux
AND mem16/32, imm8 (signed ext.) 83h mm-100-xxx long load, alux, store
ARPL mreg16, reg16 63h 11-xxx-xxx vector
ARPL mem16, reg16 63h mm-xxx-xxx vector
BOUND 62h vector
BSF reg16/32, mreg16/32 0Fh BCh 11-xxx-xxx vector
BSF reg16/32, mem16/32 0Fh BCh mm-xxx-xxx vector
BSR reg16/32, mreg16/32 0Fh BDh 11-xxx-xxx vector
BSR reg16/32, mem16/32 0Fh BDh mm-xxx-xxx vector
BSWAP EAX 0Fh C8h long alu
BSWAP ECX 0Fh C9h long alu
BSWAP EDX 0Fh CAh long alu
BSWAP EBX 0Fh CBh long alu
BSWAP ESP 0Fh CCh long alu
BSWAP EBP 0Fh CDh long alu
BSWAP ESI 0Fh CEh long alu
BSWAP EDI 0Fh CFh long alu
BT mreg16/32, reg16/32 0Fh A3h 11-xxx-xxx vector
BT mem16/32, reg16/32 0Fh A3h mm-xxx-xxx vector
BT mreg16/32, imm8 0Fh BAh 11-100-xxx vector
BT mem16/32, imm8 0Fh BAh mm-100-xxx vector
BTC mreg16/32, reg16/32 0Fh BBh 11-xxx-xxx vector
BTC mem16/32, reg16/32 0Fh BBh mm-xxx-xxx vector
BTC mreg16/32, imm8 0Fh BAh 11-111-xxx vector
BTC mem16/32, imm8 0Fh BAh mm-111-xxx vector
BTR mreg16/32, reg16/32 0Fh B3h 11-xxx-xxx vector
BTR mem16/32, reg16/32 0Fh B3h mm-xxx-xxx vector
BTR mreg16/32, imm8 0Fh BAh 11-110-xxx vector
BTR mem16/32, imm8 0Fh BAh mm-110-xxx vector
BTS mreg16/32, reg16/32 0Fh ABh 11-xxx-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
60 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
BTS mem16/32, reg16/32 0Fh ABh mm-xxx-xxx vector
BTS mreg16/32, imm8 0Fh BAh 11-101-xxx vector
BTS mem16/32, imm8 0Fh BAh mm-101-xxx vector
CALL full pointer 9Ah vector
CALL near imm16/32 E8h short store
CALL mem16:16/32 FFh 11-011-xxx vector
CALL near mreg32 (indirect) FFh 11-010-xxx vector
CALL near mem32 (indirect) FFh mm-010-xxx vector
CBW/CWDE EAX 98h vector
CLC F8h vector
CLD FCh vector
CLI FAh vector
CLTS 0Fh 06h vector
CMC F5h vector
CMP mreg8, reg8 38h 11-xxx-xxx short alux
CMP mem8, reg8 38h mm-xxx-xxx short load, alux
CMP mreg16/32, reg16/32 39h 11-xxx-xxx short alu
CMP mem16/32, reg16/32 39h mm-xxx-xxx short load, alu
CMP reg8, mreg8 3Ah 11-xxx-xxx short alux
CMP reg8, mem8 3Ah mm-xxx-xxx short load, alux
CMP reg16/32, mreg16/32 3Bh 11-xxx-xxx short alu
CMP reg16/32, mem16/32 3Bh mm-xxx-xxx short load, alu
CMP AL, imm8 3Ch short alux
CMP EAX, imm16/32 3Dh short alu
CMP mreg8, imm8 80h 11-111-xxx short alux
CMP mem8, imm8 80h mm-111-xxx short load, alux
CMP mreg16/32, imm16/32 81h 11-111-xxx short alu
CMP mem16/32, imm16/32 81h mm-111-xxx short load, alu
CMP mreg16/32, imm8 (signed ext.) 83h 11-111-xxx long load, alu
CMP mem16/32, imm8 (signed ext.) 83h mm-111-xxx long load, alu
CMPSB mem8, mem8 A6h vector
CMPSW mem16, mem32 A7h vector
CMPSD mem32, mem32 A7h vector
CMPXCHG mreg8, reg8 0Fh B0h 11-xxx-xxx vector
CMPXCHG mem8, reg8 0Fh B0h mm-xxx-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 61
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
CMPXCHG mreg16/32, reg16/32 0Fh B1h 11-xxx-xxx vector
CMPXCHG mem16/32, reg16/32 0Fh B1h mm-xxx-xxx vector
CMPXCHG8B EDX:EAX 0Fh C7h 11-xxx-xxx vector
CMPXCHG8B mem64 0Fh C7h mm-xxx-xxx vector
CPUID 0Fh A2h vector
CWD/CDQ EDX, EAX 99h vector
DAA 27h vector
DAS 2Fh vector
DEC EAX 48h short alu
DEC ECX 49h short alu
DEC EDX 4Ah short alu
DEC EBX 4Bh short alu
DEC ESP 4Ch short alu
DEC EBP 4Dh short alu
DEC ESI 4Eh short alu
DEC EDI 4Fh short alu
DEC mreg8 FEh 11-001-xxx vector
DEC mem8 FEh mm-001-xxx long load, alux, store
DEC mreg16/32 FFh 11-001-xxx vector
DEC mem16/32 FFh mm-001-xxx long load, alu, store
DIV AL, mreg8 F6h 11-110-xxx vector
DIV AL, mem8 F6h mm-110-xxx vector
DIV EAX, mreg16/32 F7h 11-110-xxx vector
DIV EAX, mem16/32 F7h mm-110-xxx vector
IDIV mreg8 F6h 11-111-xxx vector
IDIV mem8 F6h mm-111-xxx vector
IDIV EAX, mreg16/32 F7h 11-111-xxx vector
IDIV EAX, mem16/32 F7h mm-111-xxx vector
IMUL reg16/32, imm16/32 69h 11-xxx-xxx vector
IMUL reg16/32, mreg16/32, imm16/32 69h 11-xxx-xxx vector
IMUL reg16/32, mem16/32, imm16/32 69h mm-xxx-xxx vector
IMUL reg16/32, imm8 (sign extended) 6Bh 11-xxx-xxx vector
IMUL reg16/32, mreg16/32, imm8 (signed) 6Bh 11-xxx-xxx vector
IMUL reg16/32, mem16/32, imm8 (signed) 6Bh mm-xxx-xxx vector
IMUL AX, AL, mreg8 F6h 11-101-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
62 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
IMUL AX, AL, mem8 F6h mm-101-xxx vector
IMUL EDX:EAX, EAX, mreg16/32 F7h 11-101-xxx vector
IMUL EDX:EAX, EAX, mem16/32 F7h mm-101-xxx vector
IMUL reg16/32, mreg16/32 0Fh AFh 11-xxx-xxx vector
IMUL reg16/32, mem16/32 0Fh AFh mm-xxx-xxx vector
IN AL, imm8 E4h vector
IN AX, imm8 E5h vector
IN EAX, imm8 E5h vector
IN AL, DX ECh vector
IN AX, DX EDh vector
IN EAX, DX EDh vector
INC EAX 40h short alu
INC ECX 41h short alu
INC EDX 42h short alu
INC EBX 43h short alu
INC ESP 44h short alu
INC EBP 45h short alu
INC ESI 46h short alu
INC EDI 47h short alu
INC mreg8 FEh 11-000-xxx vector
INC mem8 FEh mm-000-xxx long load, alux, store
INC mreg16/32 FFh 11-000-xxx vector
INC mem16/32 FFh mm-000-xxx long load, alu, store
INVD 0Fh 08h vector
INVLPG 0Fh 01h mm-111-xxx vector
JO short disp8 70h short branch
JB/JNAE short disp8 71h short branch
JNO short disp8 71h short branch
JNB/JAE short disp8 73h short branch
JZ/JE short disp8 74h short branch
JNZ/JNE short disp8 75h short branch
JBE/JNA short disp8 76h short branch
JNBE/JA short disp8 77h short branch
JS short disp8 78h short branch
JNS short disp8 79h short branch
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 63
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
JP/JPE short disp8 7Ah short branch
JNP/JPO short disp8 7Bh short branch
JL/JNGE short disp8 7Ch short branch
JNL/JGE short disp8 7Dh short branch
JLE/JNG short disp8 7Eh short branch
JNLE/JG short disp8 7Fh short branch
JCXZ/JEC short disp8 E3h vector
JO near disp16/32 0Fh 80h short branch
JNO near disp16/32 0Fh 81h short branch
JB/JNAE near disp16/32 0Fh 82h short branch
JNB/JAE near disp16/32 0Fh 83h short branch
JZ/JE near disp16/32 0Fh 84h short branch
JNZ/JNE near disp16/32 0Fh 85h short branch
JBE/JNA near disp16/32 0Fh 86h short branch
JNBE/JA near disp16/32 0Fh 87h short branch
JS near disp16/32 0Fh 88h short branch
JNS near disp16/32 0Fh 89h short branch
JP/JPE near disp16/32 0Fh 8Ah short branch
JNP/JPO near disp16/32 0Fh 8Bh short branch
JL/JNGE near disp16/32 0Fh 8Ch short branch
JNL/JGE near disp16/32 0Fh 8Dh short branch
JLE/JNG near disp16/32 0Fh 8Eh short branch
JNLE/JG near disp16/32 0Fh 8Fh short branch
JMP near disp16/32 (direct) E9h short branch
JMP far disp32/48 (direct) EAh vector
JMP disp8 (short) EBh short branch
JMP far mreg32 (indirect) EFh 11-101-xxx vector
JMP far mem32 (indirect) EFh mm-101-xxx vector
JMP near mreg16/32 (indirect) FFh 11-100-xxx vector
JMP near mem16/32 (indirect) FFh mm-100-xxx vector
LAHF 9Fh vector
LAR reg16/32, mreg16/32 0Fh 02h 11-xxx-xxx vector
LAR reg16/32, mem16/32 0Fh 02h mm-xxx-xxx vector
LDS reg16/32, mem32/48 C5h mm-xxx-xxx vector
LEA reg16/32, mem16/32 8Dh mm-xxx-xxx short load, alu
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
64 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
LEAVE C9h long load, alu, alu
LES reg16/32, mem32/48 C4h mm-xxx-xxx vector
LFS reg16/32, mem32/48 0Fh B4h vector
LGDT mem48 0Fh 01h mm-010-xxx vector
LGS reg16/32, mem32/48 0Fh B5h vector
LIDT mem48 0Fh 01h mm-011-xxx vector
LLDT mreg16 0Fh 00h 11-010-xxx vector
LLDT mem16 0Fh 00h mm-010-xxx vector
LMSW mreg16 0Fh 01h 11-100-xxx vector
LMSW mem16 0Fh 01h mm-100-xxx vector
LODSB AL, mem8 ACh long load, alu
LODSW AX, mem16 ADh long load, alu
LODSD EAX, mem32 ADh long load, alu
LOOP disp8 E2h short alu, branch
LOOPE/LOOPZ disp8 E1h vector
LOOPNE/LOOPNZ disp8 E0h vector
LSL reg16/32, mreg16/32 0Fh 03h 11-xxx-xxx vector
LSL reg16/32, mem16/32 0Fh 03h mm-xxx-xxx vector
LSS reg16/32, mem32/48 0Fh B2h mm-xxx-xxx vector
LTR mreg16 0Fh 00h 11-011-xxx vector
LTR mem16 0Fh 00h mm-011-xxx vector
MOV mreg8, reg8 88h 11-xxx-xxx short alux
MOV mem8, reg8 88h mm-xxx-xxx short store
MOV mreg16/32, reg16/32 89h 11-xxx-xxx short alu
MOV mem16/32, reg16/32 89h mm-xxx-xxx short store
MOV reg8, mreg8 8Ah 11-xxx-xxx short alux
MOV reg8, mem8 8Ah mm-xxx-xxx short load
MOV reg16/32, mreg16/32 8Bh 11-xxx-xxx short alu
MOV reg16/32, mem16/32 8Bh mm-xxx-xxx short load
MOV mreg16, segment reg 8Ch 11-xxx-xxx long load
MOV mem16, segment reg 8Ch mm-xxx-xxx vector
MOV segment reg, mreg16 8Eh 11-xxx-xxx vector
MOV segment reg, mem16 8Eh mm-xxx-xxx vector
MOV AL, mem8 A0h short load
MOV EAX, mem16/32 A1h short load
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 65
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
MOV mem8, AL A2h short store
MOV mem16/32, EAX A3h short store
MOV AL, imm8 B0h short limm
MOV CL, imm8 B1h short limm
MOV DL, imm8 B2h short limm
MOV BL, imm8 B3h short limm
MOV AH, imm8 B4h short limm
MOV CH, imm8 B5h short limm
MOV DH, imm8 B6h short limm
MOV BH, imm8 B7h short limm
MOV EAX, imm16/32 B8h short limm
MOV ECX, imm16/32 B9h short limm
MOV EDX, imm16/32 BAh short limm
MOV EBX, imm16/32 BBh short limm
MOV ESP, imm16/32 BCh short limm
MOV EBP, imm16/32 BDh short limm
MOV ESI, imm16/32 BEh short limm
MOV EDI, imm16/32 BFh short limm
MOV mreg8, imm8 C6h 11-000-xxx short limm
MOV mem8, imm8 C6h mm-000-xxx long store
MOV mreg16/32, imm16/32 C7h 11-000-xxx short limm
MOV mem16/32, imm16/32 C7h mm-000-xxx long store
MOVSB mem8,mem8 A4h long load, store, alux, alux
MOVSD mem16, mem16 A5h long load, store, alu, alu
MOVSW mem32, mem32 A5h long load, store, alu, alu
MOVSX reg16/32, mreg8 0Fh BEh 11-xxx-xxx short alu
MOVSX reg16/32, mem8 0Fh BEh mm-xxx-xxx short load, alu
MOVSX reg32, mreg16 0Fh BFh 11-xxx-xxx short alu
MOVSX reg32, mem16 0Fh BFh mm-xxx-xxx short load, alu
MOVZX reg16/32, mreg8 0Fh B6h 11-xxx-xxx short alu
MOVZX reg16/32, mem8 0Fh B6h mm-xxx-xxx short load, alu
MOVZX reg32, mreg16 0Fh B7h 11-xxx-xxx short alu
MOVZX reg32, mem16 0Fh B7h mm-xxx-xxx short load, alu
MUL AL, mreg8 F6h 11-100-xxx vector
MUL AL, mem8 F6h mm-100-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
66 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
MUL EAX, mreg16/32 F7h 11-100-xxx vector
MUL EAX, mem16/32 F7h mm-100-xxx vector
NEG mreg8 F6h 11-011-xxx short alux
NEG mem8 F6h mm-011-xxx vector
NEG mreg16/32 F7h 11-011-xxx short alu
NEG mem16/32 F7h mm-011-xxx vector
NOP (XCHG EAX, EAX) 90h short limm
NOT mreg8 F6h 11-010-xxx short alux
NOT mem8 F6h mm-010-xxx vector
NOT mreg16/32 F7h 11-010-xxx short alu
NOT mem16/32 F7h mm-010-xxx vector
OR mreg8, reg8 08h 11-xxx-xxx short alux
OR mem8, reg8 08h mm-xxx-xxx long load, alux, store
OR mreg16/32, reg16/32 09h 11-xxx-xxx short alu
OR mem16/32, reg16/32 09h mm-xxx-xxx long load, alu, store
OR reg8, mreg8 0Ah 11-xxx-xxx short alux
OR reg8, mem8 0Ah mm-xxx-xxx short load, alux
OR reg16/32, mreg16/32 0Bh 11-xxx-xxx short alu
OR reg16/32, mem16/32 0Bh mm-xxx-xxx short load, alu
OR AL, imm8 0Ch short alux
OR EAX, imm16/32 0Dh short alu
OR mreg8, imm8 80h 11-001-xxx short alux
OR mem8, imm8 80h mm-001-xxx long load, alux, store
OR mreg16/32, imm16/32 81h 11-001-xxx short alu
OR mem16/32, imm16/32 81h mm-001-xxx long load, alu, store
OR mreg16/32, imm8 (signed ext.) 83h 11-001-xxx short alux
OR mem16/32, imm8 (signed ext.) 83h mm-001-xxx long load, alux, store
OUT imm8, AL E6h vector
OUT imm8, AX E7h vector
OUT imm8, EAX E7h vector
OUT DX, AL EEh vector
OUT DX, AX EFh vector
OUT DX, EAX EFh vector
POP ES 07h vector
POP SS 17h vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 67
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
POP DS 1Fh vector
POP FS 0Fh A1h vector
POP GS 0Fh A9h vector
POP EAX 58h short load, alu
POP ECX 59h short load, alu
POP EDX 5Ah short load, alu
POP EBX 5Bh short load, alu
POP ESP 5Ch short load, alu
POP EBP 5Dh short load, alu
POP ESI 5Eh short load, alu
POP EDI 5Fh short load, alu
POP mreg 16/32 8Fh 11-000-xxx short load, alu
POP mem 16/32 8Fh mm-000-xxx long load, store, alu
POPA/POPAD 61h vector
POPF/POPFD 9Dh vector
PUSH ES 06h long load, store
PUSH CS 0Eh vector
PUSH FS 0Fh A0h vector
PUSH GS 0Fh A8h vector
PUSH SS 16h vector
PUSH DS 1Eh long load, store
PUSH EAX 50h short store
PUSH ECX 51h short store
PUSH EDX 52h short store
PUSH EBX 53h short store
PUSH ESP 54h short store
PUSH EBP 55h short store
PUSH ESI 56h short store
PUSH EDI 57h short store
PUSH imm8 6Ah long store
PUSH imm16/32 68h long store
PUSH mreg16/32 FFh 11-110-xxx vector
PUSH mem16/32 FFh mm-110-xxx long load, store
PUSHA/PUSHAD 60h vector
PUSHF/PUSHFD 9Ch vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
68 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
RCL mreg8, imm8 C0h 11-010-xxx vector
RCL mem8, imm8 C0h mm-010-xxx vector
RCL mreg16/32, imm8 C1h 11-010-xxx vector
RCL mem16/32, imm8 C1h mm-010-xxx vector
RCL mreg8, 1 D0h 11-010-xxx vector
RCL mem8, 1 D0h mm-010-xxx vector
RCL mreg16/32, 1 D1h 11-010-xxx vector
RCL mem16/32, 1 D1h mm-010-xxx vector
RCL mreg8, CL D2h 11-010-xxx vector
RCL mem8, CL D2h mm-010-xxx vector
RCL mreg16/32, CL D3h 11-010-xxx vector
RCL mem16/32, CL D3h mm-010-xxx vector
RCR mreg8, imm8 C0h 11-011-xxx vector
RCR mem8, imm8 C0h mm-011-xxx vector
RCR mreg16/32, imm8 C1h 11-011-xxx vector
RCR mem16/32, imm8 C1h mm-011-xxx vector
RCR mreg8, 1 D0h 11-011-xxx vector
RCR mem8, 1 D0h mm-011-xxx vector
RCR mreg16/32, 1 D1h 11-011-xxx vector
RCR mem16/32, 1 D1h mm-011-xxx vector
RCR mreg8, CL D2h 11-011-xxx vector
RCR mem8, CL D2h mm-011-xxx vector
RCR mreg16/32, CL D3h 11-011-xxx vector
RCR mem16/32, CL D3h mm-011-xxx vector
RET near imm16 C2h vector
RET near C3h vector
RET far imm16 CAh vector
RET far CBh vector
ROL mreg8, imm8 C0h 11-000-xxx vector
ROL mem8, imm8 C0h mm-000-xxx vector
ROL mreg16/32, imm8 C1h 11-000-xxx vector
ROL mem16/32, imm8 C1h mm-000-xxx vector
ROL mreg8, 1 D0h 11-000-xxx vector
ROL mem8, 1 D0h mm-000-xxx vector
ROL mreg16/32, 1 D1h 11-000-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 69
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
ROL mem16/32, 1 D1h mm-000-xxx vector
ROL mreg8, CL D2h 11-000-xxx vector
ROL mem8, CL D2h mm-000-xxx vector
ROL mreg16/32, CL D3h 11-000-xxx vector
ROL mem16/32, CL D3h mm-000-xxx vector
ROR mreg8, imm8 C0h 11-001-xxx vector
ROR mem8, imm8 C0h mm-001-xxx vector
ROR mreg16/32, imm8 C1h 11-001-xxx vector
ROR mem16/32, imm8 C1h mm-001-xxx vector
ROR mreg8, 1 D0h 11-001-xxx vector
ROR mem8, 1 D0h mm-001-xxx vector
ROR mreg16/32, 1 D1h 11-001-xxx vector
ROR mem16/32, 1 D1h mm-001-xxx vector
ROR mreg8, CL D2h 11-001-xxx vector
ROR mem8, CL D2h mm-001-xxx vector
ROR mreg16/32, CL D3h 11-001-xxx vector
ROR mem16/32, CL D3h mm-001-xxx vector
SAHF 9Eh vector
SAR mreg8, imm8 C0h 11-111-xxx short alux
SAR mem8, imm8 C0h mm-111-xxx vector
SAR mreg16/32, imm8 C1h 11-111-xxx short alu
SAR mem16/32, imm8 C1h mm-111-xxx vector
SAR mreg8, 1 D0h 11-111-xxx short alux
SAR mem8, 1 D0h mm-111-xxx vector
SAR mreg16/32, 1 D1h 11-111-xxx short alu
SAR mem16/32, 1 D1h mm-111-xxx vector
SAR mreg8, CL D2h 11-111-xxx short alux
SAR mem8, CL D2h mm-111-xxx vector
SAR mreg16/32, CL D3h 11-111-xxx short alu
SAR mem16/32, CL D3h mm-111-xxx vector
SBB mreg8, reg8 18h 11-xxx-xxx vector
SBB mem8, reg8 18h mm-xxx-xxx vector
SBB mreg16/32, reg16/32 19h 11-xxx-xxx vector
SBB mem16/32, reg16/32 19h mm-xxx-xxx vector
SBB reg8, mreg8 1Ah 11-xxx-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
70 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
SBB reg8, mem8 1Ah mm-xxx-xxx vector
SBB reg16/32, mreg16/32 1Bh 11-xxx-xxx vector
SBB reg16/32, mem16/32 1Bh mm-xxx-xxx vector
SBB AL, imm8 1Ch vector
SBB EAX, imm16/32 1Dh vector
SBB mreg8, imm8 80h 11-011-xxx vector
SBB mem8, imm8 80h mm-011-xxx vector
SBB mreg16/32, imm16/32 81h 11-011-xxx vector
SBB mem16/32, imm16/32 81h mm-011-xxx vector
SBB mreg16/32, imm8 (signed ext.) 83h 11-011-xxx vector
SBB mem16/32, imm8 (signed ext.) 83h mm-011-xxx vector
SCASB AL, mem8 AEh vector
SCASW AX, mem16 AFh vector
SCASD EAX, mem32 AFh vector
SETO mreg8 0Fh 90h 11-xxx-xxx vector
SETO mem8 0Fh 90h mm-xxx-xxx vector
SETNO mreg8 0Fh 91h 11-xxx-xxx vector
SETNO mem8 0Fh 91h mm-xxx-xxx vector
SETB/SETNAE mreg8 0Fh 92h 11-xxx-xxx vector
SETB/SETNAE mem8 0Fh 92h mm-xxx-xxx vector
SETNB/SETAE mreg8 0Fh 93h 11-xxx-xxx vector
SETNB/SETAE mem8 0Fh 93h mm-xxx-xxx vector
SETZ/SETE mreg8 0Fh 94h 11-xxx-xxx vector
SETZ/SETE mem8 0Fh 94h mm-xxx-xxx vector
SETNZ/SETNE mreg8 0Fh 95h 11-xxx-xxx vector
SETNZ/SETNE mem8 0Fh 95h mm-xxx-xxx vector
SETBE/SETNA mreg8 0Fh 96h 11-xxx-xxx vector
SETBE/SETNA mem8 0Fh 96h mm-xxx-xxx vector
SETNBE/SETA mreg8 0Fh 97h 11-xxx-xxx vector
SETNBE/SETA mem8 0Fh 97h mm-xxx-xxx vector
SETS mreg8 0Fh 98h 11-xxx-xxx vector
SETS mem8 0Fh 98h mm-xxx-xxx vector
SETNS mreg8 0Fh 99h 11-xxx-xxx vector
SETNS mem8 0Fh 99h mm-xxx-xxx vector
SETP/SETPE mreg8 0Fh 9Ah 11-xxx-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 71
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
SETP/SETPE mem8 0Fh 9Ah mm-xxx-xxx vector
SETNP/SETPO mreg8 0Fh 9Bh 11-xxx-xxx vector
SETNP/SETPO mem8 0Fh 9Bh mm-xxx-xxx vector
SETL/SETNGE mreg8 0Fh 9Ch 11-xxx-xxx vector
SETL/SETNGE mem8 0Fh 9Ch mm-xxx-xxx vector
SETNL/SETGE mreg8 0Fh 9Dh 11-xxx-xxx vector
SETNL/SETGE mem8 0Fh 9Dh mm-xxx-xxx vector
SETLE/SETNG mreg8 0Fh 9Eh 11-xxx-xxx vector
SETLE/SETNG mem8 0Fh 9Eh mm-xxx-xxx vector
SETNLE/SETG mreg8 0Fh 9Fh 11-xxx-xxx vector
SETNLE/SETG mem8 0Fh 9Fh mm-xxx-xxx vector
SGDT mem48 0Fh 01h mm-000-xxx vector
SIDT mem48 0Fh 01h mm-001-xxx vector
SHL/SAL mreg8, imm8 C0h 11-100-xxx short alux
SHL/SAL mem8, imm8 C0h mm-100-xxx vector
SHL/SAL mreg16/32, imm8 C1h 11-100-xxx short alu
SHL/SAL mem16/32, imm8 C1h mm-100-xxx vector
SHL/SAL mreg8, 1 D0h 11-100-xxx short alux
SHL/SAL mem8, 1 D0h mm-100-xxx vector
SHL/SAL mreg16/32, 1 D1h 11-100-xxx short alu
SHL/SAL mem16/32, 1 D1h mm-100-xxx vector
SHL/SAL mreg8, CL D2h 11-100-xxx short alux
SHL/SAL mem8, CL D2h mm-100-xxx vector
SHL/SAL mreg16/32, CL D3h 11-100-xxx short alu
SHL/SAL mem16/32, CL D3h mm-100-xxx vector
SHR mreg8, imm8 C0h 11-101-xxx short alux
SHR mem8, imm8 C0h mm-101-xxx vector
SHR mreg16/32, imm8 C1h 11-101-xxx short alu
SHR mem16/32, imm8 C1h mm-101-xxx vector
SHR mreg8, 1 D0h 11-101-xxx short alux
SHR mem8, 1 D0h mm-101-xxx vector
SHR mreg16/32, 1 D1h 11-101-xxx short alu
SHR mem16/32, 1 D1h mm-101-xxx vector
SHR mreg8, CL D2h 11-101-xxx short alux
SHR mem8, CL D2h mm-101-xxx vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
72 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
SHR mreg16/32, CL D3h 11-101-xxx short alu
SHR mem16/32, CL D3h mm-101-xxx vector
SHLD mreg16/32, reg16/32, imm8 0Fh A4h 11-xxx-xxx vector
SHLD mem16/32, reg16/32, imm8 0Fh A4h mm-xxx-xxx vector
SHLD mreg16/32, reg16/32, CL 0Fh A5h 11-xxx-xxx vector
SHLD mem16/32, reg16/32, CL 0Fh A5h mm-xxx-xxx vector
SHRD mreg16/32, reg16/32, imm8 0Fh ACh 11-xxx-xxx vector
SHRD mem16/32, reg16/32, imm8 0Fh ACh mm-xxx-xxx vector
SHRD mreg16/32, reg16/32, CL 0Fh ADh 11-xxx-xxx vector
SHRD mem16/32, reg16/32, CL 0Fh ADh mm-xxx-xxx vector
SLDT mreg16 0Fh 00h 11-000-xxx vector
SLDT mem16 0Fh 00h mm-000-xxx vector
SMSW mreg16 0Fh 01h 11-100-xxx vector
SMSW mem16 0Fh 01h mm-100-xxx vector
STC F9h vector
STD FDh vector
STI FBh vector
STOSB mem8, AL AAh long store, alux
STOSW mem16, AX ABh long store, alux
STOSD mem32, EAX ABh long store, alux
STR mreg16 0Fh 00h 11-001-xxx vector
STR mem16 0Fh 00h mm-001-xxx vector
SUB mreg8, reg8 28h 11-xxx-xxx short alux
SUB mem8, reg8 28h mm-xxx-xxx long load, alux, store
SUB mreg16/32, reg16/32 29h 11-xxx-xxx short alu
SUB mem16/32, reg16/32 29h mm-xxx-xxx long load, alu, store
SUB reg8, mreg8 2Ah 11-xxx-xxx short alux
SUB reg8, mem8 2Ah mm-xxx-xxx short load, alux
SUB reg16/32, mreg16/32 2Bh 11-xxx-xxx short alu
SUB reg16/32, mem16/32 2Bh mm-xxx-xxx short load, alu
SUB AL, imm8 2Ch short alux
SUB EAX, imm16/32 2Dh short alu
SUB mreg8, imm8 80h 11-101-xxx short alux
SUB mem8, imm8 80h mm-101-xxx long load, alux, store
SUB mreg16/32, imm16/32 81h 11-101-xxx short alu
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 73
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
SUB mem16/32, imm16/32 81h mm-101-xxx long load, alu, store
SUB mreg16/32, imm8 (signed ext.) 83h 11-101-xxx short alux
SUB mem16/32, imm8 (signed ext.) 83h mm-101-xxx long load, alux, store
SYSCALL 0Fh 05h vector
SYSRET 0Fh 07h vector
TEST mreg8, reg8 84h 11-xxx-xxx short alux
TEST mem8, reg8 84h mm-xxx-xxx vector
TEST mreg16/32, reg16/32 85h 11-xxx-xxx short alu
TEST mem16/32, reg16/32 85h mm-xxx-xxx vector
TEST AL, imm8 A8h long alux
TEST EAX, imm16/32 A9h long alu
TEST mreg8, imm8 F6h 11-000-xxx long alux
TEST mem8, imm8 F6h mm-000-xxx long load, alux
TEST mreg16/32, imm16/32 F7h 11-000-xxx long alu
TEST mem16/32, imm16/32 F7h mm-000-xxx long load, alu
VERR mreg16 0Fh 00h 11-100-xxx vector
VERR mem16 0Fh 00h mm-100-xxx vector
VERW mreg16 0Fh 00h 11-101-xxx vector
VERW mem16 0Fh 00h mm-101-xxx vector
WAIT 9Bh vector
WBINVD 0Fh 09h vector
XADD mreg8, reg8 0Fh C0h 11-100-xxx vector
XADD mem8, reg8 0Fh C0h mm-100-xxx vector
XADD mreg16/32, reg16/32 0Fh C1h 11-101-xxx vector
XADD mem16/32, reg16/32 0Fh C1h mm-101-xxx vector
XCHG reg8, mreg8 86h 11-xxx-xxx vector
XCHG reg8, mem8 86h mm-xxx-xxx vector
XCHG reg16/32, mreg16/32 87h 11-xxx-xxx vector
XCHG reg16/32, mem16/32 87h mm-xxx-xxx vector
XCHG EAX, EAX 90h short limm
XCHG EAX, ECX 91h long alu, alu, alu
XCHG EAX, EDX 92h long alu, alu, alu
XCHG EAX, EBX 93h long alu, alu, alu
XCHG EAX, ESP 94h long alu, alu, alu
XCHG EAX, EBP 95h long alu, alu, alu
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
74 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
XCHG EAX, ESI 96h long alu, alu, alu
XCHG EAX, EDI 97h long alu, alu, alu
XLAT D7h vector
XOR mreg8, reg8 30h 11-xxx-xxx short alux
XOR mem8, reg8 30h mm-xxx-xxx long load, alux, store
XOR mreg16/32, reg16/32 31h 11-xxx-xxx short alu
XOR mem16/32, reg16/32 31h mm-xxx-xxx long load, alu, store
XOR reg8, mreg8 32h 11-xxx-xxx short alux
XOR reg8, mem8 32h mm-xxx-xxx short load, alux
XOR reg16/32, mreg16/32 33h 11-xxx-xxx short alu
XOR reg16/32, mem16/32 33h mm-xxx-xxx short load, alu
XOR AL, imm8 34h short alux
XOR EAX, imm16/32 35h short alu
XOR mreg8, imm8 80h 11-110-xxx short alux
XOR mem8, imm8 80h mm-110-xxx long load, alux, store
XOR mreg16/32, imm16/32 81h 11-110-xxx short alu
XOR mem16/32, imm16/32 81h mm-110-xxx long load, alu, store
XOR mreg16/32, imm8 (signed ext.) 83h 11-110-xxx short alux
XOR mem16/32, imm8 (signed ext.) 83h mm-110-xxx long load, alux, store
Table 13. Floating-Point Instructions
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
F2XM1 D9h F0h short float
FABS D9h F1h short float
FADD ST(0), ST(i)1D8h 11-000-xxx short float
FADD ST(0), mem32real D8h mm-000-xxx short fload, float
FADD ST(i), ST(0)1DCh 11-000-xxx short float
FADD ST(0), mem64real DCh mm-000-xxx short fload, float
FADDP ST(i), ST(0)1DEh 11-000-xxx short float
FBLD DFh mm-100-xxx vector
FBSTP DFh mm-110-xxx vector
FCHS D9h E0h short float
FCLEX DBh E2h vector
Table 12. Integer Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 75
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
FCOM ST(0), ST(i)1D8h 11-010-xxx short float
FCOM ST(0), mem32real D8h mm-010-xxx short fload, float
FCOM ST(0), mem64real DCh mm-010-xxx short fload, float
FCOMP ST(0), ST(i)1D8h 11-011-xxx short float
FCOMP ST(0), mem32real D8h mm-011-xxx short fload, float
FCOMP ST(0), mem64real DCh mm-011-xxx short fload, float
FCOMPP DEh D9h 11-011-001 short float
FCOS D9h FFh short float
FDECSTP D9h F6h short float
FDIV ST(0), ST(i) (single precision)1D8h 11-110-xxx short float
FDIV ST(0), ST(i) (double precision)1D8h 11-110-xxx short float
FDIV ST(0), ST(i) (extended precision)1D8h 11-110-xxx short float
FDIV ST(i), ST(0) (single precision)1DCh 11-111-xxx short float
FDIV ST(i), ST(0) (double precision)1DCh 11-111-xxx short float
FDIV ST(i), ST(0) (extended precision)1DCh 11-111-xxx short float
FDIV ST(0), mem32real D8h mm-110-xxx short fload, float
FDIV ST(0), mem64real DCh mm-110-xxx short fload, float
FDIVP ST(0), ST(i)1DEh 11-111-xxx short float
FDIVR ST(0), ST(i)1D8h 11-110-xxx short float
FDIVR ST(i), ST(0)1DCh 11-111-xxx short float
FDIVR ST(0), mem32real D8h mm-111-xxx short fload, float
FDIVR ST(0), mem64real DCh mm-111-xxx short fload, float
FDIVRP ST(i), ST(0)1DEh 11-110-xxx short float
FFREE ST(i)1DDh 11-000-xxx short float
FIADD ST(0), mem32int DAh mm-000-xxx short fload, float
FIADD ST(0), mem16int DEh mm-000-xxx short fload, float
FICOM ST(0), mem32int DAh mm-010-xxx short fload, float
FICOM ST(0), mem16int DEh mm-010-xxx short fload, float
FICOMP ST(0), mem32int DAh mm-011-xxx short fload, float
FICOMP ST(0), mem16int DEh mm-011-xxx short fload, float
FIDIV ST(0), mem32int DAh mm-110-xxx short fload, float
FIDIV ST(0), mem16int DEh mm-110-xxx short fload, float
FIDIVR ST(0), mem32int DAh mm-111-xxx short fload, float
Table 13. Floating-Point Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
76 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
FIDIVR ST(0), mem16int DEh mm-111-xxx short fload, float
FILD mem16int DFh mm-000-xxx short fload, float
FILD mem32int DBh mm-000-xxx short fload, float
FILD mem64int DFh mm-101-xxx short fload, float
FIMUL ST(0), mem32int DAh mm-001-xxx short fload, float
FIMUL ST(0), mem16int DEh mm-001-xxx short fload, float
FINCSTP D9h F7h short
FINIT DBh E3h vector
FIST mem16int DFh mm-010-xxx short fload, float
FIST mem32int DBh mm-010-xxx short fload, float
FISTP mem16int DFh mm-011-xxx short fload, float
FISTP mem32int DBh mm-011-xxx short fload, float
FISTP mem64int DFh mm-111-xxx short fload, float
FISUB ST(0), mem32int DAh mm-100-xxx short fload, float
FISUB ST(0), mem16int DEh mm-100-xxx short fload, float
FISUBR ST(0), mem32int DAh mm-101-xxx short fload, float
FISUBR ST(0), mem16int DEh mm-101-xxx short fload, float
FLD ST(i)1D9h 11-000-xxx short fload, float
FLD mem32real D9h mm-000-xxx short fload, float
FLD mem64real DDh mm-000-xxx short fload, float
FLD mem80real DBh mm-101-xxx vector
FLD1 D9h E8h short fload, float
FLDCW D9h mm-101-xxx vector
FLDENV D9h mm-100-xxx short fload, float
FLDL2E D9h EAh short float
FLDL2T D9h E9h short float
FLDLG2 D9h ECh short float
FLDLN2 D9h EDh short float
FLDPI D9h EBh short float
FLDZ D9h EEh short float
FMUL ST(0), ST(i)1D8h 11-001-xxx short float
FMUL ST(i), ST(0)1DCh 11-001-xxx short float
FMUL ST(0), mem32real D8h mm-001-xxx short fload, float
FMUL ST(0), mem64real DCh mm-001-xxx short fload, float
FMULP ST(0), ST(i)1DEh 11-001-xxx short float
Table 13. Floating-Point Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 77
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
FNOP D9h D0h short float
FPATAN D9h F3h short float
FPREM D9h F8h short float
FPREM1 D9h F5h short float
FPTAN D9h F2h vector
FRNDINT D9h FCh short float
FRSTOR DDh mm-100-xxx vector
FSAVE DDh mm-110-xxx vector
FSCALE D9h FDh short float
FSIN D9h FEh short float
FSINCOS D9h FBh vector
FSQRT (single precision) D9h FAh short float
FSQRT (double precision) D9h FAh short float
FSQRT (extended precision) D9h FAh short float
FST mem32real D9h mm-010-xxx short fstore
FST mem64real DDh mm-010-xxx short fstore
FST ST(i)1DDh 11-010-xxx short fstore
FSTCW D9h mm-111-xxx vector
FSTENV D9h mm-110-xxx vector
FSTP mem32real D9h mm-011-xxx short fstore
FSTP mem64real DDh mm-011-xxx short fstore
FSTP mem80real D9h mm-111-xxx vector
FSTP ST(i)1DDh 11-011-xxx short float
FSTSW AX DFh E0h vector
FSTSW mem16 DDh mm-111-xxx vector
FSUB ST(0), mem32real D8h mm-100-xxx short fload, float
FSUB ST(0), mem64real DCh mm-100-xxx short fload, float
FSUB ST(0), ST(i)1D8h 11-100-xxx short float
FSUB ST(i), ST(0)1DCh 11-101-xxx short float
FSUBP ST(0), ST(i)1DEh 11-101-xxx short float
FSUBR ST(0), mem32real D8h mm-101-xxx short fload, float
FSUBR ST(0), mem64real DCh mm-101-xxx short fload, float
FSUBR ST(0), ST(i)1D8h 11-100-xxx short float
FSUBR ST(i), ST(0)1DCh 11-101-xxx short float
Table 13. Floating-Point Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
78 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
FSUBRP ST(i), ST(0)1DEh 11-100-xxx short float
FTST D9h E4h short float
FUCOM DDh 11-100-xxx short float
FUCOMP DDh 11-101-xxx short float
FUCOMPP DAh E9h short float
FXAM D9h E5h short float
FXCH D9h 11-001-xxx short float
FXTRACT D9h F4h vector
FYL2X D9h F1h short float
FYL2XP1 D9h F9h short float
FWAIT 9Bh vector
Notes:
1. The last three bits of the modR/M byte select the stack entry ST(i).
Table 14. MMX™ Instructions
Instruction Mnemonic Prefix
Byte(s) First
Byte ModR/M
Byte Decode
Type RISC86
Operations
EMMS 0Fh 77h vector
MOVD mmreg, mreg3210Fh 6Eh 11-xxx-xxx short meu
MOVD mmreg, mem32 0Fh 6Eh mm-xxx-xxx short mload
MOVD mreg32, mmreg10Fh 7Eh 11-xxx-xxx short mstore, load
MOVD mem32, mmreg 0Fh 7Eh mm-xxx-xxx short mstore
MOVQ mmreg1, mmreg2 0Fh 6Fh 11-xxx-xxx short meu
MOVQ mmreg, mem64 0Fh 6Fh mm-xxx-xxx short mload
MOVQ mmreg2, mmreg1 0Fh 7Fh 11-xxx-xxx short meu
MOVQ mem64, mmreg 0Fh 7Fh mm-xxx-xxx short mstore
PACKSSDW mmreg1, mmreg2 0Fh 6Bh 11-xxx-xxx short meu
PACKSSDW mmreg, mem64 0Fh 6Bh mm-xxx-xxx short mload, meu
PACKSSWB mmreg1, mmreg2 0Fh 63h 11-xxx-xxx short meu
PACKSSWB mmreg, mem64 0Fh 63h mm-xxx-xxx short mload, meu
PACKUSWB mmreg1, mmreg2 0Fh 67h 11-xxx-xxx short meu
PACKUSWB mmreg, mem64 0Fh 67h mm-xxx-xxx short mload, meu
PADDB mmreg1, mmreg2 0Fh FCh 11-xxx-xxx short meu
PADDB mmreg, mem64 0Fh FCh mm-xxx-xxx short mload, meu
Table 13. Floating-Point Instructions (continued)
Instruction Mnemonic First
Byte Second
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 79
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
PADDD mmreg1, mmreg2 0Fh FEh 11-xxx-xxx short meu
PADDD mmreg, mem64 0Fh FEh mm-xxx-xxx short mload, meu
PADDSB mmreg1, mmreg2 0Fh ECh 11-xxx-xxx short meu
PADDSB mmreg, mem64 0Fh ECh mm-xxx-xxx short mload, meu
PADDSW mmreg1, mmreg2 0Fh EDh 11-xxx-xxx short meu
PADDSW mmreg, mem64 0Fh EDh mm-xxx-xxx short mload, meu
PADDUSB mmreg1, mmreg2 0Fh DCh 11-xxx-xxx short meu
PADDUSB mmreg, mem64 0Fh DCh mm-xxx-xxx short mload, meu
PADDUSW mmreg1, mmreg2 0Fh DDh 11-xxx-xxx short meu
PADDUSW mmreg, mem64 0Fh DDh mm-xxx-xxx short mload, meu
PADDW mmreg1, mmreg2 0Fh FDh 11-xxx-xxx short meu
PADDW mmreg, mem64 0Fh FDh mm-xxx-xxx short mload, meu
PAND mmreg1, mmreg2 0Fh DBh 11-xxx-xxx short meu
PAND mmreg, mem64 0Fh DBh mm-xxx-xxx short mload, meu
PANDN mmreg1, mmreg2 0Fh DFh 11-xxx-xxx short meu
PANDN mmreg, mem64 0Fh DFh mm-xxx-xxx short mload, meu
PCMPEQB mmreg1, mmreg2 0Fh 74h 11-xxx-xxx short meu
PCMPEQB mmreg, mem64 0Fh 74h mm-xxx-xxx short mload, meu
PCMPEQD mmreg1, mmreg2 0Fh 76h 11-xxx-xxx short meu
PCMPEQD mmreg, mem64 0Fh 76h mm-xxx-xxx short mload, meu
PCMPEQW mmreg1, mmreg2 0Fh 75h 11-xxx-xxx short meu
PCMPEQW mmreg, mem64 0Fh 75h mm-xxx-xxx short mload, meu
PCMPGTB mmreg1, mmreg2 0Fh 64h 11-xxx-xxx short meu
PCMPGTB mmreg, mem64 0Fh 64h mm-xxx-xxx short mload, meu
PCMPGTD mmreg1, mmreg2 0Fh 66h 11-xxx-xxx short meu
PCMPGTD mmreg, mem64 0Fh 66h mm-xxx-xxx short mload, meu
PCMPGTW mmreg1, mmreg2 0Fh 65h 11-xxx-xxx short meu
PCMPGTW mmreg, mem64 0Fh 65h mm-xxx-xxx short mload, meu
PMADDWD mmreg1, mmreg2 0Fh F5h 11-xxx-xxx short meu
PMADDWD mmreg, mem64 0Fh F5h mm-xxx-xxx short mload, meu
PMULHW mmreg1, mmreg2 0Fh E5h 11-xxx-xxx short meu
PMULHW mmreg, mem64 0Fh E5h mm-xxx-xxx short mload, meu
PMULLW mmreg1, mmreg2 0Fh D5h 11-xxx-xxx short meu
PMULLW mmreg, mem64 0Fh D5h mm-xxx-xxx short mload, meu
POR mmreg1, mmreg2 0Fh EBh 11-xxx-xxx short meu
Table 14. MMX™ Instructions (continued)
Instruction Mnemonic Prefix
Byte(s) First
Byte ModR/M
Byte Decode
Type RISC86
Operations
80 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
POR mmreg, mem64 0Fh EBh mm-xxx-xxx short mload, meu
PSLLD mmreg1, mmreg2 0Fh F2h 11-xxx-xxx short meu
PSLLD mmreg, mem64 0Fh F2h mm-xxx-xxx short mload, meu
PSLLD mmreg, imm8 0Fh 72h 11-110-xxx short meu
PSLLQ mmreg1, mmreg2 0Fh F3h 11-xxx-xxx short meu
PSLLQ mmreg, mem64 0Fh F3h mm-xxx-xxx short mload, meu
PSLLQ mmreg, imm8 0Fh 73h 11-110-xxx short meu
PSLLW mmreg1, mmreg2 0Fh F1h 11-xxx-xxx short meu
PSLLW mmreg, mem64 0Fh F1h mm-xxx-xxx short mload, meu
PSLLW mmreg, imm8 0Fh 71h 11-110-xxx short meu
PSRAD mmreg1, mmreg2 0Fh E2h 11-xxx-xxx short meu
PSRAD mmreg, mem64 0Fh E2h mm-xxx-xxx short mload, meu
PSRAD mmreg, imm8 0Fh 72h 11-100-xxx short meu
PSRAW mmreg1, mmreg2 0Fh E1h 11-xxx-xxx short meu
PSRAW mmreg, mem64 0Fh E1h mm-xxx-xxx short mload, meu
PSRAW mmreg, imm8 0Fh 71h 11-100-xxx short meu
PSRLD mmreg1, mmreg2 0Fh D2h 11-xxx-xxx short meu
PSRLD mmreg, mem64 0Fh D2h mm-xxx-xxx short mload, meu
PSRLD mmreg, imm8 0Fh 72h 11-010-xxx short meu
PSRLQ mmreg1, mmreg2 0Fh D3h 11-xxx-xxx short meu
PSRLQ mmreg, mem64 0Fh D3h mm-xxx-xxx short mload, meu
PSRLQ mmreg, imm8 0Fh 73h 11-010-xxx short meu
PSRLW mmreg1, mmreg2 0Fh D1h 11-xxx-xxx short meu
PSRLW mmreg, mem64 0Fh D1h mm-xxx-xxx short mload, meu
PSRLW mmreg, imm8 0Fh 71h 11-010-xxx short meu
PSUBB mmreg1, mmreg2 0Fh F8h 11-xxx-xxx short meu
PSUBB mmreg, mem64 0Fh F8h mm-xxx-xxx short mload, meu
PSUBD mmreg1, mmreg2 0Fh FAh 11-xxx-xxx short meu
PSUBD mmreg, mem64 0Fh FAh mm-xxx-xxx short mload, meu
PSUBSB mmreg1, mmreg2 0Fh E8h 11-xxx-xxx short meu
PSUBSB mmreg, mem64 0Fh E8h mm-xxx-xxx short mload, meu
PSUBSW mmreg1, mmreg2 0Fh E9h 11-xxx-xxx short meu
PSUBSW mmreg, mem64 0Fh E9h mm-xxx-xxx short mload, meu
PSUBUSB mmreg1, mmreg2 0Fh D8h 11-xxx-xxx short meu
PSUBUSB mmreg, mem64 0Fh D8h mm-xxx-xxx short mload, meu
Table 14. MMX™ Instructions (continued)
Instruction Mnemonic Prefix
Byte(s) First
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 3 Software Environment 81
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
PSUBUSW mmreg1, mmreg2 0Fh D9h 11-xxx-xxx short meu
PSUBUSW mmreg, mem64 0Fh D9h mm-xxx-xxx short mload, meu
PSUBW mmreg1, mmreg2 0Fh F9h 11-xxx-xxx short meu
PSUBW mmreg, mem64 0Fh F9h mm-xxx-xxx short mload, meu
PUNPCKHBW mmreg1, mmreg2 0Fh 68h 11-xxx-xxx short meu
PUNPCKHBW mmreg, mem64 0Fh 68h mm-xxx-xxx short mload, meu
PUNPCKHDQ mmreg1, mmreg2 0Fh 6Ah 11-xxx-xxx short meu
PUNPCKHDQ mmreg, mem64 0Fh 6Ah mm-xxx-xxx short mload, meu
PUNPCKHWD mmreg1, mmreg2 0Fh 69h 11-xxx-xxx short meu
PUNPCKHWD mmreg, mem64 0Fh 69h mm-xxx-xxx short mload, meu
PUNPCKLBW mmreg1, mmreg2 0Fh 60h 11-xxx-xxx short meu
PUNPCKLBW mmreg, mem64 0Fh 60h mm-xxx-xxx short mload, meu
PUNPCKLDQ mmreg1, mmreg2 0Fh 62h 11-xxx-xxx short meu
PUNPCKLDQ mmreg, mem64 0Fh 62h mm-xxx-xxx short mload, meu
PUNPCKLWD mmreg1, mmreg2 0Fh 61h 11-xxx-xxx short meu
PUNPCKLWD mmreg, mem64 0Fh 61h mm-xxx-xxx short mload, meu
PXOR mmreg1, mmreg2 0Fh EFh 11-xxx-xxx short meu
PXOR mmreg, mem64 0Fh EFh mm-xxx-xxx short mload, meu
Notes:
1. Bits 2, 1, and 0 of the modR/M byte select the integer register.
Table 15. 3DNow!™ Instructions
Instruction Mnemonic Prefix
Byte(s) Opcode
Byte ModR/M
Byte Decode
Type RISC86
Operations
FEMMS 0Fh 0Eh vector
PAVGUSB mmreg1, mmreg2 0Fh, 0Fh BFh 11-xxx-xxx short meu
PAVGUSB mmreg, mem64 0Fh, 0Fh BFh mm-xxx-xxx short mload, meu
PF2ID mmreg1, mmreg2 0Fh, 0Fh 1Dh 11-xxx-xxx short meu
PF2ID mmreg, mem64 0Fh, 0Fh 1Dh mm-xxx-xxx short mload, meu
PFACC mmreg1, mmreg2 0Fh, 0Fh AEh 11-xxx-xxx short meu
PFACC mmreg, mem64 0Fh, 0Fh AEh mm-xxx-xxx short mload, meu
PFADD mmreg1, mmreg2 0Fh, 0Fh 9Eh 11-xxx-xxx short meu
PFADD mmreg, mem64 0Fh, 0Fh 9Eh mm-xxx-xxx short mload, meu
PFCMPEQ mmreg1, mmreg2 0Fh, 0Fh B0h 11-xxx-xxx short meu
PFCMPEQ mmreg, mem64 0Fh, 0Fh B0h mm-xxx-xxx short mload, meu
Table 14. MMX™ Instructions (continued)
Instruction Mnemonic Prefix
Byte(s) First
Byte ModR/M
Byte Decode
Type RISC86
Operations
82 Software Environment Chapter 3
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
PFCMPGE mmreg1, mmreg2 0Fh, 0Fh 90h 11-xxx-xxx short meu
PFCMPGE mmreg, mem64 0Fh, 0Fh 90h mm-xxx-xxx short mload, meu
PFCMPGT mmreg1, mmreg2 0Fh, 0Fh A0h 11-xxx-xxx short meu
PFCMPGT mmreg, mem64 0Fh, 0Fh A0h mm-xxx-xxx short mload, meu
PFMAX mmreg1, mmreg2 0Fh, 0Fh A4h 11-xxx-xxx short meu
PFMAX mmreg, mem64 0Fh, 0Fh A4h mm-xxx-xxx short mload, meu
PFMIN mmreg1, mmreg2 0Fh, 0Fh 94h 11-xxx-xxx short meu
PFMIN mmreg, mem64 0Fh, 0Fh 94h mm-xxx-xxx short mload, meu
PFMUL mmreg1, mmreg2 0Fh, 0Fh B4h 11-xxx-xxx short meu
PFMUL mmreg, mem64 0Fh, 0Fh B4h mm-xxx-xxx short mload, meu
PFRCP mmreg1, mmreg2 0Fh, 0Fh 96h 11-xxx-xxx short meu
PFRCP mmreg, mem64 0Fh, 0Fh 96h mm-xxx-xxx short mload, meu
PFRCPIT1 mmreg1, mmreg2 0Fh, 0Fh A6h 11-xxx-xxx short meu
PFRCPIT1 mmreg, mem64 0Fh, 0Fh A6h mm-xxx-xxx short mload, meu
PFRCPIT2 mmreg1, mmreg2 0Fh, 0Fh B6h 11-xxx-xxx short meu
PFRCPIT2 mmreg, mem64 0Fh, 0Fh B6h mm-xxx-xxx short mload, meu
PFRSQIT1 mmreg1, mmreg2 0Fh, 0Fh A7h 11-xxx-xxx short meu
PFRSQIT1 mmreg, mem64 0Fh, 0Fh A7h mm-xxx-xxx short mload, meu
PFRSQRT mmreg1, mmreg2 0Fh, 0Fh 97h 11-xxx-xxx short meu
PFRSQRT mmreg, mem64 0Fh, 0Fh 97h mm-xxx-xxx short mload, meu
PFSUB mmreg1, mmreg2 0Fh, 0Fh 9Ah 11-xxx-xxx short meu
PFSUB mmreg, mem64 0Fh, 0Fh 9Ah mm-xxx-xxx short mload, meu
PFSUBR mmreg1, mmreg2 0Fh, 0Fh AAh 11-xxx-xxx short meu
PFSUBR mmreg, mem64 0Fh, 0Fh AAh mm-xxx-xxx short mload, meu
PI2FD mmreg1, mmreg2 0Fh, 0Fh 0Dh 11-xxx-xxx short meu
PI2FD mmreg, mem64 0Fh, 0Fh 0Dh mm-xxx-xxx short mload, meu
PMULHRW mmreg1, mmreg2 0Fh, 0Fh B7h 11-xxx-xxx short meu
PMULHRW mmreg1, mem64 0Fh, 0Fh B7h mm-xxx-xxx short mload, meu
PREFETCH mem810Fh 0Dh mm-000-xxx vector load
PREFETCHW mem81,2 0Fh 0Dh mm-001-xxx vector load
Notes:
1. For PREFETCH and PREFETCHW, the mem8 value refers to a byte address within the 32-byte line that will be prefetched.
2. PREFETCHW will be implemented in a future K86 processor. On the AMD-K6-2E processor, this instruction performs in the same
manner as the PREFETCH instruction.
Table 15. 3DNow!™ Instructions (continued)
Instruction Mnemonic Prefix
Byte(s) Opcode
Byte ModR/M
Byte Decode
Type RISC86
Operations
Chapter 4 Logic Symbol Diagram 83
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
4 Logic Symbol Diagram
A20M#
A[31:3]
AP
ADS#
ADSC#
APCHK#
BE[7:0]#
AHOLD
BOFF#
BREQ
HLDA
HOLD
D/C#
EWBE#
LOCK#
M/IO#
NA#
SCYC
W/R#
CACHE#
KEN#
PCD
PWT
WB/WT#
Clock
Bus
Arbitration
CLK BF[2:0]
TCK TDI TDO TMS TRST#
BRDY#
BRDYC#
D[63:0]
DP[7:0]
PCHK#
EADS#
HIT#
HITM#
INV
FERR#
IGNNE#
FLUSH#
INIT
INTR
NMI
RESET
SMI#
SMIACT#
STPCLK#
JTAG Test
Data
and
Data
Parity
Inquire
Cycles
Floating-Point
Error Handling
External
Interrupts,
SMM, Reset and
Initialization
Address
and
Address
Parity
Cycle
Definition
and
Control
Cache
Control
AMD-K6™-2E
Processor
Voltage Detection
VCC2DET VCC2H/L#
Notes:
The signals are grouped by function. The arrows show the direction of the signal, either into or out of the processor. Signals with double-
headed arrows are bidirectional. Signals with pound signs (#) are active Low.
84 Logic Symbol Diagram Chapter 4
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 5 Signal Descriptions 85
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5 Signal Descriptions
This chapter includes a detailed description of each signal
supported on the AMD-K6-2E processor. This chapter also
provides tables listing the signals grouped by type, beginning
on page 130.
The logic symbol diagram on page 83 shows the signals grouped
by function.
Connection diagrams and pins listed by high-level function are
included in Chapter 17, “Pin Designation Diagrams” on page
299.
5.1 Signal Terminology
The following terminology is used in this chapter:
■Driven—The processor actively pulls the signal up to the
High-voltage state or pulls the signal down to the
Low-voltage state.
■Floated—The the signal is not being driven by the processor
(high-impedance state), which allows another device to
drive this signal.
■Asserted—For all active-High signals, the term asserted
means the signal is in the High-voltage state. For all
active-Low signals, the term asserted means the signal is in
the Low-voltage state.
■Negated—For all active-High signals, the term negated
means the signal is in the Low-voltage state. For all
active-Low signals, the term negated means the signal is in
the High-voltage state.
■Sampled—The processor has measured the state of a signal
at predefined points in time and will take the appropriate
action based on the state of the signal. If a signal is not
sampled by the processor, its assertion or negation has no
effect on the operation of the processor.
86 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.2 A20M# (Address Bit 20 Mask)
Pin Attribute Input
Pin Location AK-08
Summary A20M# is used to simulate the behavior of the 8086 when
running in real mode. The assertion of A20M# causes the
processor to force bit 20 of the physical address to 0 prior to
accessing the cache or driving out a memory bus cycle. The
clearing of address bit 20 maps addresses that wrap above
1 Mbyte to addresses below 1 Mbyte.
Sampled The processor samples A20M# as a level-sensitive input on
every clock edge. The system logic can drive the signal either
synchronously or asynchronously. If it is asserted
asynchronously, it must be asserted for a minimum pulse width
of two clocks.
The following list explains the effects of the processor sampling
A20M# asserted under various conditions:
■Inquire cycles and writeback cycles are not affected by the
state of A20M#.
■The assertion of A20M# in system management mode
(SMM) is ignored.
■When A20M# is sampled asserted in protected mode, it
causes unpredictable processor operation. A20M# is only
defined in real mode.
■To ensure that A20M# is recognized before the first ADS#
occurs following the negation of RESET, A20M# must be
sampled asserted on the same clock edge that RESET is
sampled negated or on one of the two subsequent clock
edges.
■To ensure A20M# is recognized before the execution of an
instruction, a serializing instruction must be executed
between the instruction that asserts A20M# and the
targeted instruction.
Chapter 5 Signal Descriptions 87
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.3 A[31:3] (Address Bus)
Pin Attribute A[31:5] Bidirectional, A[4:3] Output
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary A[31:3] contain the physical address for the current bus cycle.
The processor drives addresses on A[31:3] during memory and
I/O cycles, and cycle definition information during special bus
cycles. The processor samples addresses on A[31:5] during
inquire cycles.
Driven, Sampled, and
Floated As Outputs: A[31:3] are driven valid off the same clock edge as
ADS# and remain in the same state until the clock edge on
which NA# or the last expected BRDY# of the cycle is sampled
asserted. A[31:3] are driven during memory cycles, I/O cycles,
special bus cycles, and interrupt acknowledge cycles. The
processor continues to drive the address bus while the bus is
idle.
As Inputs: The processor samples A[31:5] during inquire cycles
on the clock edge on which EADS# is sampled asserted. Even
though A4 and A3 are not used during the inquire cycle, they
must be driven to a valid state and must meet the same timings
as A[31:5].
A[31:3] are floated off the clock edge that AHOLD or BOFF# is
sampled asserted and off the clock edge that the processor
asserts HLDA in recognition of HOLD.
The processor resumes driving A[31:3] off the clock edge on
which the processor samples AHOLD or BOFF# negated and off
the clock edge on which the processor negates HLDA.
88 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.4 ADS# (Address Strobe)
Pin Attribute Output
Pin Location AJ-05
Summary The assertion of ADS# indicates the beginning of a new bus
cycle. The address bus and all cycle definition signals
corresponding to this bus cycle are driven valid off the same
clock edge as ADS#.
Driven and Floated ADS# is asserted for one clock at the beginning of each bus
cycle. For non-pipelined cycles, ADS# can be asserted as early
as the clock edge after the clock edge on which the last
expected BRDY# of the cycle is sampled asserted, resulting in a
single idle state between cycles. For pipelined cycles if the
processor is prepared to start a new cycle, ADS# can be asserted
as early as one clock edge after NA# is sampled asserted.
If AHOLD is sampled asserted, ADS# is only driven in order to
perform a writeback cycle due to an inquire cycle that hits a
modified cache line.
The processor floats ADS# off the clock edge that BOFF# is
sampled asserted and off the clock edge that the processor
asserts HLDA in recognition of HOLD.
5.5 ADSC# (Address Strobe Copy)
Pin Attribute Output
Pin Location AM-02
Summary ADSC# has the identical function and timing as ADS#. In the
event ADS# becomes too heavily loaded due to a large fanout in
a system, ADSC# can be used to split the load across two
outputs, which can improve system timing.
Chapter 5 Signal Descriptions 89
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.6 AHOLD (Address Hold)
Pin Attribute Input
Pin Location V-04
Summary AHOLD can be asserted by the system to initiate one or more
inquire cycles. To allow the system to drive the address bus
during an inquire cycle, the processor floats A[31:3] and AP off
the clock edge on which AHOLD is sampled asserted. The data
bus and all other control and status signals remain under the
control of the processor and are not floated. This allows a bus
cycle that is in progress when AHOLD is sampled asserted to
continue to completion. The processor resumes driving the
address bus off the clock edge on which AHOLD is sampled
negated.
If AHOLD is sampled asserted, ADS# is only asserted in order
to perform a writeback cycle due to an inquire cycle that hits a
modified cache line.
Sampled The processor samples AHOLD on every clock edge. AHOLD is
recognized while INIT and RESET are sampled asserted.
90 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.7 AP (Address Parity)
Pin Attribute Bidirectional
Pin Location AK-02
Summary AP contains the even parity bit for cache line addresses driven
and sampled on A[31:5]. Even parity means that the total
number of 1 bits on AP and A[31:5] is even. (A4 and A3 are not
used for the generation or checking of address parity because
these bits are not required to address a cache line.) AP is driven
by the processor during processor-initiated cycles and is
sampled by the processor during inquire cycles. If AP does not
reflect even parity during an inquire cycle, the processor
asserts APCHK# to indicate an address bus parity check. The
processor does not take an internal exception as the result of
detecting an address bus parity check, and system logic must
respond appropriately to the assertion of this signal.
Driven, Sampled, and
Floated As an Output: The processor drives AP valid off the clock edge
on which ADS# is asserted until the clock edge on which NA# or
the last expected BRDY# of the cycle is sampled asserted. AP is
driven during memory cycles, I/O cycles, special bus cycles, and
interrupt acknowledge cycles. The processor continues to drive
AP while the bus is idle.
As an Input: The processor samples AP during inquire cycles on
the clock edge on which EADS# is sampled asserted.
The processor floats AP off the clock edge that AHOLD or
BOFF# is sampled asserted and off the clock edge that the
processor asserts HLDA in recognition of HOLD.
The processor resumes driving AP off the clock edge on which
the processor samples AHOLD or BOFF# negated and off the
clock edge on which the processor negates HLDA.
Chapter 5 Signal Descriptions 91
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.8 APCHK# (Address Parity Check)
Pin Attribute Output
Pin Location AE-05
Summary If the processor detects an address parity error during an
inquire cycle, APCHK# is asserted for one clock. The processor
does not take an internal exception as the result of detecting an
address bus parity check, and system logic must respond
appropriately to the assertion of this signal.
The processor is designed so that APCHK# does not glitch,
enabling the signal to be used as a clocking source for system
logic.
Driven APCHK# is driven valid off the clock edge after the clock edge
on which the processor samples EADS# asserted. It is negated
off the next clock edge.
APCHK# is always driven except in the three-state test mode.
92 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.9 BE[7:0]# (Byte Enables)
Pin Attribute Output
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary BE[7:0]# are used by the processor to indicate the valid data
bytes during a write cycle and the requested data bytes during
a read cycle. The byte enables can be used to derive address bits
A[2:0], which are not physically part of the processor’s address
bus. The processor checks and generates valid data parity for
the data bytes that are valid as defined by the byte enables. The
eight byte enables correspond to the eight bytes of the data bus
as follows:
The processor expects data to be driven by the system logic on
all eight bytes of the data bus during a burst cache-line read
cycle, independent of the byte enables that are asserted.
The byte enables are also used to distinguish between special
bus cycles as defined in Table 23 on page 132.
Driven and Floated BE[7:0]# are driven off the same clock edge as ADS# and
remain in the same state until the clock edge on which NA# or
the last expected BRDY# of the cycle is sampled asserted.
BE[7:0]# are driven during memory cycles, I/O cycles, special
bus cycles, and interrupt acknowledge cycles.
The processor floats BE[7:0]# off the clock edge that BOFF# is
sampled asserted and off the clock edge that the processor
asserts HLDA in recognition of HOLD. Unlike the address bus,
BE[7:0]# are not floated in response to AHOLD.
■BE7#: D[63:56] ■BE3#: D[31:24]
■BE6#: D[55:48] ■BE2#: D[23:16]
■BE5#: D[47:40] ■BE1#: D[15:8]
■BE4#: D[39:32] ■BE0#: D[7:0]
Chapter 5 Signal Descriptions 93
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.10 BF[2:0] (Bus Frequency)
Pin Attributes Inputs, Internal Pullups
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary BF[2:0] determine the internal operating frequency of the
processor. The frequency of the CLK input signal is multiplied
internally by a ratio determined by the state of these signals as
defined in Table 16. BF[2:0] have weak internal pullups and
default to the 3.5 multiplier if left unconnected.
Sampled BF[2:0] are sampled during the falling transition of RESET.
They must meet a minimum setup time of 1.0 ms and a
minimum hold time of two clocks relative to the negation of
RESET.
Table 16. Processor-to-Bus Clock Ratios
State of BF[2:0] Inputs Processor-Clock to Bus-Clock Ratio
100b 2.5x
101b 3.0x
111b 3.5x
010b 4.0x
000b 4.5x
001b 5.0x
011b 5.5x
110b 6.0x
94 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.11 BOFF# (Backoff)
Pin Attribute Input
Pin Location Z-04
Summary If BOFF# is sampled asserted, the processor unconditionally
aborts any cycles in progress and transitions to a bus hold state
by floating the following signals: A[31:3], ADS#, ADSC#, AP,
BE[7:0]#, CACHE#, D[63:0], D/C#, DP[7:0], LOCK#, M/IO#,
PCD, PWT, SCYC, and W/R#. These signals remain floated until
BOFF# is sampled negated. This allows an alternate bus master
or the system to control the bus.
When BOFF# is sampled negated, any processor cycle that was
aborted due to the assertion of BOFF# is restarted from the
beginning of the cycle, regardless of the number of transfers
that were completed. If BOFF# is sampled asserted on the same
clock edge as BRDY# of a bus cycle of any length, then BOFF#
takes precedence over the BRDY#. In this case, the cycle is
aborted and restarted after BOFF# is sampled negated.
Sampled BOFF# is sampled on every clock edge. The processor floats its
bus signals off the clock edge on which BOFF# is sampled
asserted. These signals remain floated until the clock edge on
which BOFF# is sampled negated.
BOFF# is recognized while INIT and RESET are sampled
asserted.
Chapter 5 Signal Descriptions 95
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.12 BRDY# (Burst Ready)
Pin Attribute Input, Internal Pullup
Pin Location X-04
Summary BRDY# is asserted to the processor by system logic to indicate
either that the data bus is being driven with valid data during a
read cycle or that the data bus has been latched during a write
cycle. If necessary, the system logic can insert bus cycle wait
states by negating BRDY# until it is ready to continue the data
transfer. BRDY# is also used to indicate the completion of
special bus cycles.
Sampled BRDY# is sampled every clock edge within a bus cycle starting
with the clock edge after the clock edge that negates ADS#.
BRDY# is ignored while the bus is idle. The processor samples
the following inputs on the clock edge on which BRDY# is
sampled asserted: D[63:0], DP[7:0], and KEN# during read
cycles, EWBE# during write cycles (if not masked off), and
WB/WT# during read and write cycles. If NA# is sampled
asserted prior to BRDY#, then KEN# and WB/WT# are sampled
on the clock edge on which NA# is sampled asserted.
The number of times the processor expects to sample BRDY#
asserted depends on the type of bus cycle, as follows:
■One time for a single-transfer cycle, a special bus cycle, or
each of two cycles in an interrupt acknowledge sequence
■Four times for a burst cycle (once for each data transfer)
BRDY# can be held asserted for four consecutive clocks
throughout the four transfers of the burst, or it can be negated
to insert wait states.
96 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.13 BRDYC# (Burst Ready Copy)
Pin Attribute Input, Internal Pullup
Pin Location Y-03
Summary BRDYC# has the identical function as BRDY#. In the event
BRDY# becomes too heavily loaded due to a large fanout or
loading in a system, BRDYC# can be used to reduce this
loading, which improves timing.
Sampled BRDYC# is sampled every clock edge within a bus cycle starting
with the clock edge after the clock edge that negates ADS#.
5.14 BREQ (Bus Request)
Pin Attribute Output
Pin Location AJ-01
Summary BREQ is asserted by the processor to request the bus in order to
complete an internally pending bus cycle. The system logic can
use BREQ to arbitrate among the bus participants. If the
processor does not own the bus, BREQ is asserted until the
processor gains access to the bus in order to begin the pending
cycle or until the processor no longer needs to run the pending
cycle. If the processor currently owns the bus, BREQ is asserted
with ADS#. The processor asserts BREQ for each assertion of
ADS# but does not necessarily assert ADS# for each assertion of
BREQ.
Driven BREQ is asserted off the same clock edge on which ADS# is
asserted. BREQ can also be asserted off any clock edge,
independent of the assertion of ADS#. BREQ can be negated
one clock edge after it is asserted.
The processor always drives BREQ except in the three-state test
mode.
Chapter 5 Signal Descriptions 97
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.15 CACHE# (Cacheable Access)
Pin Attribute Output
Pin Location U-03
Summary For reads, CACHE# is asserted to indicate the cacheability of
the current bus cycle. In addition, if the processor samples
KEN# asserted, which indicates the driven address is
cacheable, the cycle is a 32-byte burst read cycle. For write
cycles, CACHE# is asserted to indicate the current bus cycle is a
modified cache-line writeback. KEN# is ignored during
writebacks. If CACHE# is not asserted, or if KEN# is sampled
negated during a read cycle, the cycle is not cacheable and
defaults to a single-transfer cycle.
Driven and Floated CACHE# is driven off the same clock edge as ADS# and remains
in the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted.
CACHE# is floated off the clock edge that BOFF# is sampled
asserted and off the clock edge that the processor asserts HLDA
in recognition of HOLD.
5.16 CLK (Clock)
Pin Attribute Input
Pin Location AK-18
Summary The CLK signal is the bus clock for the processor and is the
reference for all signal timings under normal operation (except
for TDI, TDO, TMS, and TRST#). BF[2:0] determine the internal
frequency multiplier applied to CLK to obtain the processor’s
core operating frequency. See “BF[2:0] (Bus Frequency)” on
page 93 for a list of the processor-to-bus clock ratios.
Sampled The CLK signal must be stable a minimum of 1.0 ms prior to the
negation of RESET to ensure the proper operation of the
processor. See “CLK Switching Characteristics” on page 267 for
details regarding the CLK specifications.
98 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.17 D/C# (Data/Code)
Pin Attribute Output
Pin Location AK-04
Summary The processor drives D/C# during a memory bus cycle to
indicate whether it is addressing data or executable code. D/C#
is also used to define other bus cycles, including interrupt
acknowledge and special cycles. See Table 23 on page 132 for
more details.
Driven and Floated D/C# is driven off the same clock edge as ADS# and remains in
the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted. D/C# is
driven during memory cycles, I/O cycles, special bus cycles, and
interrupt acknowledge cycles.
D/C# is floated off the clock edge that BOFF# is sampled
asserted and off the clock edge that the processor asserts HLDA
in recognition of HOLD.
Chapter 5 Signal Descriptions 99
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.18 D[63:0] (Data Bus)
Pin Attribute Bidirectional
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary D[63:0] represent the processor’s 64-bit data bus. Each of the
eight bytes of data that comprise this bus is qualified as valid
by its corresponding byte enable. See “BE[7:0]# (Byte
Enables)” on page 92.
Driven, Sampled, and
Floated As Outputs: For single-transfer write cycles, the processor drives
D[63:0] with valid data one clock edge after the clock edge on
which ADS# is asserted and D[63:0] remain in the same state
until the clock edge on which BRDY# is sampled asserted. If the
cycle is a writeback—in which case four, 8-byte transfers
occur—D[63:0] are driven one clock edge after the clock edge
on which ADS# is asserted and are subsequently changed off
the clock edge on which each BRDY# assertion of the burst
cycle is sampled.
If the assertion of ADS# represents a pipelined write cycle that
follows a read cycle, the processor does not drive D[63:0] until it
is certain that contention on the data bus will not occur. In this
case, D[63:0] are driven the clock edge after the last expected
BRDY# of the previous cycle is sampled asserted.
As Inputs: During read cycles, the processor samples D[63:0] on
the clock edge on which BRDY# is sampled asserted.
The processor always floats D[63:0] except when they are being
driven during a write cycle as described above. In addition,
D[63:0] are floated off the clock edge that BOFF# is sampled
asserted and off the clock edge that the processor asserts
HLDA in recognition of HOLD.
100 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.19 DP[7:0] (Data Parity)
Pin Attribute Bidirectional
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary DP[7:0] are even parity bits for each valid byte of data—as
defined by BE[7:0]#—driven and sampled on the D[63:0] data
bus. Even parity means that the total number of odd (1) bits
within each byte of data and its respective data parity bit is an
even number. DP[7:0] are driven by the processor during write
cycles and sampled by the processor during read cycles.
If the processor detects bad parity on any valid byte of data
during a read cycle, PCHK# is asserted for one clock beginning
the clock edge after BRDY# is sampled asserted. The processor
does not take an internal exception as the result of detecting a
data parity check, and system logic must respond appropriately
to the assertion of this signal.
The eight data parity bits correspond to the eight bytes of the
data bus as follows:
For systems that do not support data parity, DP[7:0] should be
connected to VCC3 through pullup resistors.
Driven, Sampled, and
Floated As Outputs: For single-transfer write cycles, the processor drives
DP[7:0] with valid parity one clock edge after the clock edge on
which ADS# is asserted and DP[7:0] remain in the same state
until the clock edge on which BRDY# is sampled asserted. If the
cycle is a writeback, DP[7:0] are driven one clock edge after the
clock edge on which ADS# is asserted and are subsequently
changed off the clock edge on which each BRDY# assertion of
the burst cycle is sampled.
As Inputs: During read cycles, the processor samples DP[7:0] on
the clock edge on which BRDY# is sampled asserted.
■DP7: D[63:56] ■DP3: D[31:24]
■DP6: D[55:48] ■DP2: D[23:16]
■DP5: D[47:40] ■DP1: D[15:8]
■DP4: D[39:32] ■DP0: D[7:0]
Chapter 5 Signal Descriptions 101
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The processor always floats DP[7:0] except when they are being
driven during a write cycle as described above. In addition,
DP[7:0] are floated off the clock edge that BOFF# is sampled
asserted and off the clock edge that the processor asserts HLDA
in recognition of HOLD.
5.20 EADS# (External Address Strobe)
Pin Attribute Input
Pin Location AM-04
Summary System logic asserts EADS# during a cache inquire cycle to
indicate that the address bus contains a valid address. EADS#
can only be driven after the system logic has taken control of
the address bus by asserting AHOLD or BOFF# or by receiving
HLDA. The processor responds to the sampling of EADS# and
the address bus by driving HIT#, which indicates if the inquired
cache line exists in the processor’s cache, and HITM#, which
indicates if it is in the modified state.
Sampled If AHOLD or BOFF# is asserted by the system logic in order to
execute a cache inquire cycle, the processor begins sampling
EADS# two clock edges after AHOLD or BOFF# is sampled
asserted. If the system logic asserts HOLD in order to execute a
cache inquire cycle, the processor begins sampling EADS# two
clock edges after the clock edge HLDA is asserted by the
processor.
EADS# is ignored during the following conditions:
■One clock edge after the clock edge on which EADS# is
sampled asserted
■Two clock edges after the clock edge on which ADS# is
asserted
■When the processor is driving the address bus
■When the processor asserts HITM#
102 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.21 EWBE# (External Write Buffer Empty)
Pin Attribute Input
Pin Location W-03
Summary The system logic can negate EWBE# to the processor to indicate
that its external write buffers are full and that additional data
cannot be stored at this time. This causes the processor to delay
the following activities until EWBE# is sampled asserted:
■The commitment of write hit cycles to cache lines in the
modified state or exclusive state in the processor’s cache
■The decode and execution of an instruction that follows a
currently-executing serializing instruction
■The assertion or negation of SMIACT#
■The entering of the Halt state and the Stop Grant state
Negating EWBE# does not prevent the completion of any type
of cycle that is currently in progress.
Sampled The processor samples EWBE# on each clock edge that BRDY#
is sampled asserted during all memory write cycles (except
writeback cycles), I/O write cycles, and special bus cycles.
If EWBE# is sampled negated, it is sampled on every clock edge
until it is asserted, and then it is ignored until BRDY# is
sampled asserted in the next write cycle or special cycle.
If EFER[3] is 1, then EWBE# is ignored by the processor. For
more information on the EFER settings and EWBE#, see
“EWBE# Control” on page 205.
Chapter 5 Signal Descriptions 103
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.22 FERR# (Floating-Point Error)
Pin Attribute Output
Pin Location Q-05
Summary The assertion of FERR# indicates the occurrence of an
unmasked floating-point exception resulting from the
execution of a floating-point instruction. This signal is provided
to allow the system logic to handle this exception in a manner
consistent with IBM-compatible PC/AT systems. See “Handling
Floating-Point Exceptions” on page 213 for a system logic
implementation that supports floating-point exceptions.
The state of the numeric error (NE) bit in CR0 does not affect
the FERR# signal.
The processor is designed so that FERR# does not glitch,
enabling the signal to be used as a clocking source for system
logic.
Driven The processor asserts FERR# on the instruction boundary of
the next floating-point instruction, MMX instruction, 3DNow!
instruction, or WAIT instruction that occurs following the
floating-point instruction that caused the unmasked
floating-point exception—that is, FERR# is not asserted at the
time the exception occurs. The IGNNE# signal does not affect
the assertion of FERR#.
FERR# is negated during the following conditions:
■Following the successful execution of the floating-point
instructions FCLEX, FINIT, FSAVE, and FSTENV
■Under certain circumstances, following the successful
execution of the floating-point instructions FLDCW,
FLDENV, and FRSTOR, which load the floating-point status
word or the floating-point control word
■Following the falling transition of RESET
FERR# is always driven except in the three-state test mode.
See “IGNNE# (Ignore Numeric Exception)” on page 108 for
more details on floating-point exceptions.
104 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.23 FLUSH# (Cache Flush)
Pin Attribute Input
Pin Location AN-07
Summary In response to sampling FLUSH# asserted, the processor writes
back any data cache lines that are in the modified state,
invalidates all lines in the instruction and data caches, and then
executes a flush acknowledge special cycle. See Table 23 on
page 132 for the bus definition of special cycles.
In addition, FLUSH# is sampled when RESET is negated to
determine if the processor enters the three-state test mode. If
FLUSH# is 0 during the falling transition of RESET, the
processor enters the three-state test mode instead of
performing the normal RESET functions.
Sampled FLUSH# is sampled and latched as a falling edge-sensitive
signal. During normal operation (not RESET), FLUSH# is
sampled on every clock edge but is not recognized until the next
instruction boundary. If FLUSH# is asserted synchronously, it
can be asserted for a minimum of one clock. If FLUSH# is
asserted asynchronously, it must have been negated for a
minimum of two clocks, followed by an assertion of a minimum
of two clocks.
FLUSH# is also sampled during the falling transition of RESET.
If RESET and FLUSH# are driven synchronously, FLUSH# is
sampled on the clock edge prior to the clock edge on which
RESET is sampled negated. If RESET is driven asynchronously,
the minimum setup and hold time for FLUSH#, relative to the
negation of RESET, is two clocks.
Chapter 5 Signal Descriptions 105
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.24 HIT# (Inquire Cycle Hit)
Pin Attribute Output
Pin Location AK-06
Summary The processor asserts HIT# during an inquire cycle to indicate
that the cache line is valid within the processor’s instruction or
data cache (also known as a cache hit). The cache line can be in
the modified, exclusive, or shared state.
Driven HIT# is always driven—except in the three-state test mode—
and only changes state the clock edge after the clock edge on
which EADS# is sampled asserted. It is driven in the same state
until the next inquire cycle.
5.25 HITM# (Inquire Cycle Hit To Modified Line)
Pin Attribute Output
Pin Location AL-05
Summary The processor asserts HITM# during an inquire cycle to
indicate that the cache line exists in the processor’s data cache
in the modified state. The processor performs a writeback cycle
as a result of this cache hit. If an inquire cycle hits a cache line
that is currently being written back, the processor asserts
HITM# but does not execute another writeback cycle. The
system logic must not expect the processor to assert ADS# each
time HITM# is asserted.
Driven HITM# is always driven—except in the three-state test mode—
and, in particular, is driven to represent the result of an inquire
cycle the clock edge after the clock edge on which EADS# is
sampled asserted. If HITM# is negated in response to the
inquire address, it remains negated until the next inquire cycle.
If HITM# is asserted in response to the inquire address, it
remains asserted throughout the writeback cycle and is negated
one clock edge after the last BRDY# of the writeback is
sampled asserted.
106 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.26 HLDA (Hold Acknowledge)
Pin Attribute Output
Pin Location AJ-03
Summary When HOLD is sampled asserted, the processor completes the
current bus cycles, floats the processor bus, and asserts HLDA
in an acknowledgment that these events have been completed.
The processor does not assert HLDA until the completion of a
locked sequence of cycles. While HLDA is asserted, another bus
master can drive cycles on the bus, including inquire cycles to
the processor. The following signals are floated when HLDA is
asserted: A[31:3], ADS#, ADSC#, AP, BE[7:0]#, CACHE#,
D[63:0], D/C#, DP[7:0], LOCK#, M/IO#, PCD, PWT, SCYC, and
W/R#.
The processor is designed so that HLDA does not glitch.
Driven HLDA is always driven except in the three-state test mode. If a
processor cycle is in progress while HOLD is sampled asserted,
HLDA is asserted one clock edge after the last BRDY# of the
cycle is sampled asserted. If the bus is idle, HLDA is asserted
one clock edge after HOLD is sampled asserted. HLDA is
negated one clock edge after the clock edge on which HOLD is
sampled negated.
The assertion of HLDA is independent of the sampled state of
BOFF#.
The processor floats the bus every clock in which HLDA is
asserted.
Chapter 5 Signal Descriptions 107
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.27 HOLD (Bus Hold Request)
Pin Attribute Input
Pin Location AB-04
Summary The system logic can assert HOLD to gain control of the
processor’s bus. When HOLD is sampled asserted, the processor
completes the current bus cycles, floats the processor bus, and
asserts HLDA in an acknowledgment that these events have
been completed.
Sampled The processor samples HOLD on every clock edge. If a
processor cycle is in progress while HOLD is sampled asserted,
HLDA is asserted one clock edge after the last BRDY# of the
cycle is sampled asserted. If the bus is idle, HLDA is asserted
one clock edge after HOLD is sampled asserted. HOLD is
recognized while INIT and RESET are sampled asserted.
108 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.28 IGNNE# (Ignore Numeric Exception)
Pin Attribute Input
Pin Location AA-35
Summary IGNNE#, in conjunction with the numeric error (NE) bit in CR0,
is used by the system logic to control the effect of an unmasked
floating-point exception on a previous floating-point instruction
during the execution of a floating-point instruction, MMX
instruction, 3DNow! instruction, or the WAIT instruction—
hereafter referred to as the target instruction.
If an unmasked floating-point exception is pending and the
target instruction is considered error-sensitive, then the
relationship between NE and IGNNE# is as follows:
■If NE = 0, then:
•If IGNNE# is sampled asserted, the processor ignores the
floating-point exception and continues with the
execution of the target instruction.
•If IGNNE# is sampled negated, the processor waits until
it samples IGNNE#, INTR, SMI#, NMI, or INIT asserted.
If IGNNE# is sampled asserted while waiting, the
processor ignores the floating-point exception and
continues with the execution of the target instruction.
If INTR, SMI#, NMI, or INIT is sampled asserted while
waiting, the processor handles its assertion
appropriately.
■If NE = 1, the processor invokes the INT 10h exception
handler.
If an unmasked floating-point exception is pending and the
target instruction is considered error-insensitive, then the
processor ignores the floating-point exception and continues
with the execution of the target instruction.
FERR# is not affected by the state of the NE bit or IGNNE#.
FERR# is always asserted at the instruction boundary of the
target instruction that follows the floating-point instruction
that caused the unmasked floating-point exception.
Chapter 5 Signal Descriptions 109
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
This signal is provided to allow the system logic to handle
exceptions in a manner consistent with IBM-compatible PC/AT
systems.
Sampled The processor samples IGNNE# as a level-sensitive input on
every clock edge. The system logic can drive the signal either
synchronously or asynchronously. If it is asserted
asynchronously, it must be asserted for a minimum pulse width
of two clocks.
5.29 INIT (Initialization)
Pin Attribute Input
Pin Location AA-33
Summary The assertion of INIT causes the processor to empty its
pipelines, to initialize most of its internal state, and to branch
to address FFFF_FFF0h—the same instruction execution
starting point used after RESET. Unlike RESET, the processor
preserves the contents of its caches, the floating-point state, the
MMX state, model-specific registers, the CD and NW bits of the
CR0 register, and other specific internal resources.
INIT can be used as an accelerator for 80286 code that requires
a reset to exit from protected mode back to real mode.
Sampled INIT is sampled and latched as a rising edge-sensitive signal.
INIT is sampled on every clock edge but is not recognized until
the next instruction boundary. During an I/O write cycle, it must
be sampled asserted a minimum of three clock edges before
BRDY# is sampled asserted if it is to be recognized on the
boundary between the I/O write instruction and the following
instruction.
If INIT is asserted synchronously, it can be asserted for a
minimum of one clock. If it is asserted asynchronously, it must
have been negated for a minimum of two clocks, followed by an
assertion of a minimum of two clocks.
110 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.30 INTR (Maskable Interrupt)
Pin Attribute Input
Pin Location AD-34
Summary INTR is the system’s maskable interrupt input to the processor.
When the processor samples and recognizes INTR asserted, the
processor executes a pair of interrupt acknowledge bus cycles
and then jumps to the interrupt service routine specified by the
interrupt number that was returned during the interrupt
acknowledge sequence. The processor only recognizes INTR if
the interrupt flag (IF) in the EFLAGS register equals 1.
Sampled The processor samples INTR as a level-sensitive input on every
clock edge, but the interrupt request is not recognized until the
next instruction boundary. The system logic can drive INTR
either synchronously or asynchronously. If it is asserted
asynchronously, it must be asserted for a minimum pulse width
of two clocks. In order to be recognized, INTR must remain
asserted until an interrupt acknowledge sequence is complete.
5.31 INV (Invalidation Request)
Pin Attribute Input
Pin Location U-05
Summary During an inquire cycle, the state of INV determines whether
an addressed cache line that is found in the processor’s
instruction or data cache transitions to the invalid state or the
shared state.
If INV is sampled asserted during an inquire cycle, the
processor transitions the cache line (if found) to the invalid
state, regardless of its previous state. If INV is sampled negated
during an inquire cycle, the processor transitions the cache line
(if found) to the shared state. In either case, if the cache line is
found in the modified state, the processor writes it back to
memory before changing its state.
Sampled INV is sampled on the clock edge on which EADS# is sampled
asserted.
Chapter 5 Signal Descriptions 111
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.32 KEN# (Cache Enable)
Pin Attribute Input
Pin Location W-05
Summary If KEN# is sampled asserted, it indicates that the address
presented by the processor is cacheable. If KEN# is sampled
asserted and the processor intends to perform a cache-line fill
(signified by the assertion of CACHE#), the processor executes
a 32-byte burst read cycle and expects to sample BRDY#
asserted a total of four times. If KEN# is sampled negated
during a read cycle, a single-transfer cycle is executed and the
processor does not cache the data. For write cycles, CACHE# is
asserted to indicate the current bus cycle is a modified
cache-line writeback. KEN# is ignored during writebacks.
If PCD is asserted during a bus cycle, the processor does not
cache any data read during that cycle, regardless of the state of
KEN#. See “PCD (Page Cache Disable)” on page 115 for more
details.
If the processor has sampled the state of KEN# during a cycle,
and that cycle is aborted due to the sampling of BOFF#
asserted, the system logic must ensure that KEN# is sampled in
the same state when the processor restarts the aborted cycle.
Sampled KEN# is sampled on the clock edge on which the first BRDY# or
NA# of a read cycle is sampled asserted. If the read cycle is a
burst, KEN# is ignored during the last three assertions of
BRDY#. KEN# is sampled during read cycles only when
CACHE# is asserted.
112 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.33 LOCK# (Bus Lock)
Pin Attribute Output
Pin Location AH-04
Summary The processor asserts LOCK# during a sequence of bus cycles to
ensure that the cycles are completed without allowing other bus
masters to intervene. Locked operations consist of two to five
bus cycles. LOCK# is asserted during the following operations:
■An interrupt acknowledge sequence
■Descriptor Table accesses
■Page Directory and Page Table accesses
■XCHG instruction
■An instruction with an allowable LOCK prefix
In order to ensure that locked operations appear on the bus and
are visible to the entire system, any data operands addressed
during a locked cycle that reside in the processor’s cache are
flushed and invalidated from the cache prior to the locked
operation. If the cache line is in the modified state, it is written
back and invalidated prior to the locked operation. Likewise,
any data read during a locked operation is not cached.
The processor is designed so that LOCK# does not glitch.
Driven and Floated During a locked cycle, LOCK# is asserted off the same clock
edge on which ADS# is asserted and remains asserted until the
last BRDY# of the last bus cycle is sampled asserted. The
processor negates LOCK# for at least one clock between
consecutive sequences of locked operations to allow the system
logic to arbitrate for the bus.
LOCK# is floated off the clock edge on which BOFF# is sampled
asserted and off the clock edge on which the processor asserts
HLDA in response to HOLD. When LOCK# is floated due to
BOFF# sampled asserted, the system logic is responsible for
preserving the lock condition while LOCK# is in the
high-impedance state.
Chapter 5 Signal Descriptions 113
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.34 M/IO# (Memory or I/O)
Pin Attribute Output
Pin Location T-04
Summary The processor drives M/IO# during a bus cycle to indicate
whether it is addressing the memory or I/O space. If M/IO# = 1,
the processor is addressing memory or a memory-mapped I/O
port as the result of an instruction fetch or an instruction that
loads or stores data. If M/IO# = 0, the processor is addressing an
I/O port during the execution of an I/O instruction. In addition,
M/IO# is used to define other bus cycles, including interrupt
acknowledge and special cycles. See Table 23 on page 132 for
more details.
Driven and Floated M/IO# is driven off the same clock edge as ADS# and remains in
the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted. M/IO# is
driven during memory cycles, I/O cycles, special bus cycles, and
interrupt acknowledge cycles.
M/IO# is floated off the clock edge on which BOFF# is sampled
asserted and off the clock edge on which the processor asserts
HLDA in response to HOLD.
114 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.35 NA# (Next Address)
Pin Attribute Input
Pin Location Y-05
Summary System logic asserts NA# to indicate to the processor that it is
ready to accept another bus cycle pipelined into the previous
bus cycle. ADS#, along with address and status signals, can be
asserted as early as one clock edge after NA# is sampled
asserted if the processor is prepared to start a new cycle.
Because the processor allows a maximum of two cycles to be in
progress at a time, the assertion of NA# is sampled while two
cycles are in progress, but ADS# is not asserted until the
completion of the first cycle.
Sampled NA# is sampled every clock edge during bus cycles, starting one
clock edge after the clock edge that negates ADS#, until the last
expected BRDY# of the last executed cycle is sampled asserted
(with the exception of the clock edge after the clock edge that
negates the ADS# for a second pending cycle). Because the
processor latches NA# when sampled, the system logic only
needs to assert NA# for one clock.
5.36 NMI (Non-Maskable Interrupt)
Pin Attribute Input
Pin Location AC-33
Summary When NMI is sampled asserted, the processor jumps to the
interrupt service routine defined by interrupt number 02h.
Unlike the INTR signal, software cannot mask the effect of NMI
if it is sampled asserted by the processor. However, NMI is
temporarily masked upon entering system management mode
(SMM). In addition, an interrupt acknowledge cycle is not
executed because the interrupt number is predefined.
If NMI is sampled asserted while the processor is executing the
interrupt service routine for a previous NMI, the subsequent
NMI remains pending until the completion of the execution of
the IRET instruction at the end of the interrupt service routine.
Chapter 5 Signal Descriptions 115
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Sampled NMI is sampled and latched as a rising edge-sensitive signal.
During normal operation, NMI is sampled on every clock edge
but is not recognized until the next instruction boundary. If it is
asserted synchronously, it can be asserted for a minimum of one
clock. If it is asserted asynchronously, it must have been
negated for a minimum of two clocks, followed by an assertion
of a minimum of two clocks.
5.37 PCD (Page Cache Disable)
Pin Attribute Output
Pin Location AG-05
Summary The processor drives PCD to indicate the operating system’s
specification of cacheability for the page being addressed.
System logic can use PCD to control external caching. If PCD is
asserted, the addressed page is not cached. If PCD is negated,
the cacheability of the addressed page depends upon the state
of CACHE# and KEN#.
The state of PCD depends upon the processor’s operating mode
and the state of certain bits in its control registers and TLB as
follows:
■In real mode, or in protected and virtual-8086 modes while
paging is disabled (PG bit in CR0 is 0):
PCD output = CD bit in CR0
■In protected and virtual-8086 modes while caching is
enabled (CD bit in CR0 is 0) and paging is enabled (PG bit in
CR0 is 1):
•For accesses to I/O space, page directory entries, and
other non-paged accesses:
PCD output = PCD bit in CR3
•For accesses to 4-Kbyte page table entries or 4-Mbyte
pages:
PCD output = PCD bit in page directory entry
•For accesses to 4-Kbyte pages:
PCD output = PCD bit in page table entry
116 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Driven and Floated PCD is driven off the same clock edge as ADS# and remains in
the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted.
PCD is floated off the clock edge that BOFF# is sampled
asserted and off the clock edge that the processor asserts HLDA
in response to HOLD.
5.38 PCHK# (Parity Check)
Pin Attribute Output
Pin Location AF-04
Summary The processor asserts PCHK# during read cycles if it detects an
even parity error on one or more valid bytes of D[63:0] during a
read cycle. (Even parity means that the total number of odd (1)
bits within each byte of data and its respective data parity bit is
even.) The processor checks data parity for the data bytes that
are valid, as defined by BE[7:0]#, the byte enables.
PCHK# is always driven but is only asserted for memory and I/O
read bus cycles and the second cycle of an interrupt
acknowledge sequence. PCHK# is not driven during any type of
write cycles or special bus cycles. The processor does not take
an internal exception as the result of detecting a data parity
error, and system logic must respond appropriately to the
assertion of this signal.
The processor is designed so that PCHK# does not glitch,
enabling the signal to be used as a clocking source for system
logic.
Driven PCHK# is always driven except in the three-state test mode. For
each BRDY# returned to the processor during a read cycle with
a parity error detected on the data bus, PCHK# is asserted for
one clock, one clock edge after BRDY# is sampled asserted.
Chapter 5 Signal Descriptions 117
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.39 PWT (Page Writethrough)
Pin Attribute Output
Pin Location AL-03
Summary The processor drives PWT to indicate the operating system’s
specification of the writeback state or writethrough state for
the page being addressed. PWT, together with WB/WT#,
specifies the data cache-line state during cacheable read misses
and write hits to shared cache lines. See “WB/WT# (Writeback
or Writethrough)” on page 129 for more details.
The state of PWT depends upon the processor’s operating mode
and the state of certain bits in its control registers and TLB as
follows:
■In real mode, or in protected and virtual-8086 modes while
paging is disabled (PG bit in CR0 is 0):
PWT output = 0 (writeback state)
■In protected and virtual-8086 modes while paging is enabled
(PG bit in CR0 is 1):
•For accesses to I/O space, page directory entries, and
other non-paged accesses:
PWT output = PWT bit in CR3
•For accesses to 4-Kbyte page table entries or 4-Mbyte
pages:
PWT output = PWT bit in page directory entry
•For accesses to 4-Kbyte pages:
PWT output = PWT bit in page table entry
Driven and Floated PWT is driven off the same clock edge as ADS# and remains in
the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted.
PWT is floated off the clock edge on which BOFF# is sampled
asserted and off the clock edge on which the processor asserts
HLDA in response to HOLD.
118 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.40 RESET (Reset)
Pin Attribute Input
Pin Location AK-20
Summary When the processor samples RESET asserted, it immediately
flushes and initializes all internal resources and its internal
state including its pipelines and caches, the floating-point
state, the MMX state, the 3DNow! state, and all registers, and
then the processor jumps to address FFFF_FFF0h to start
instruction execution.
The FLUSH# signal is sampled during the falling transition of
RESET to invoke the three-state test mode.
Sampled RESET is sampled as a level-sensitive input on every clock
edge. System logic can drive the signal either synchronously or
asynchronously.
During the initial power-on reset of the processor, RESET must
remain asserted for a minimum of 1.0 ms after CLK and VCC
reach specification before it is negated.
During a warm reset, while CLK and VCC are within their
specification, RESET must remain asserted for a minimum of
15 clocks prior to its negation.
Chapter 5 Signal Descriptions 119
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.41 RSVD (Reserved)
Pin Attribute Not Applicable
Pin Location See “Pin Designations by Functional Grouping” on page 301.
Summary Reserved signals are a special class of pins that can be treated
in one of the following ways:
■As no-connect (NC) pins, in which case these pins are left
unconnected
■As pins connected to the system logic as defined by the
industry-standard Pentium® interface (Socket 7)
■Any combination of NC and Socket 7 pins
In any case, if the RSVD pins are treated accordingly, the
normal operation of the AMD-K6-2E processor is not adversely
affected in any manner.
120 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.42 SCYC (Split Cycle)
Pin Attribute Output
Pin Location AL-17
Summary The processor asserts SCYC during misaligned, locked transfers
on the D[63:0] data bus. The processor generates additional bus
cycles to complete the transfer of misaligned data.
For purposes of bus cycles, the term aligned means:
■Any 1-byte transfers
■2-byte and 4-byte transfers that lie within 4-byte address
boundaries
■8-byte transfers that lie within 8-byte address boundaries
Driven and Floated SCYC is asserted off the same clock edge as ADS#, and negated
off the clock edge on which NA# or the last expected BRDY# of
the entire locked sequence is sampled asserted. SCYC is only
valid during locked memory cycles.
SCYC is floated off the clock edge on which BOFF# is sampled
asserted and off the clock edge on which the processor asserts
HLDA in response to HOLD.
Chapter 5 Signal Descriptions 121
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.43 SMI# (System Management Interrupt)
Pin Attribute Input, Internal Pullup
Pin Location AB-34
Summary The assertion of SMI# causes the processor to enter system
management mode (SMM). Upon recognizing SMI#, the
processor performs the following actions, in the order shown:
1. Flushes its instruction pipelines
2. Completes all pending and in-progress bus cycles
3. Acknowledges the interrupt by asserting SMIACT# after
sampling EWBE# asserted (if EWBE# is masked off, then
SMIACT# is not affected by EWBE#)
4. Saves the internal processor state in SMM memory
5. Disables interrupts by clearing the interrupt flag (IF) in
EFLAGS and disables NMI interrupts
6. Jumps to the entry point of the SMM service routine at the
SMM base physical address which defaults to 0003_8000h in
SMM memory
See “System Management Mode (SMM)” on page 217 for more
details regarding SMM.
Sampled SMI# is sampled and latched as a falling edge-sensitive signal.
SMI# is sampled on every clock edge but is not recognized until
the next instruction boundary. If SMI# is to be recognized on
the instruction boundary associated with a BRDY#, it must be
sampled asserted a minimum of three clock edges before the
BRDY# is sampled asserted. If it is asserted synchronously, it
can be asserted for a minimum of one clock. If it is asserted
asynchronously, it must have been negated for a minimum of
two clocks followed by an assertion of a minimum of two clocks.
A second assertion of SMI# while in SMM is latched but is not
recognized until the SMM service routine is exited.
122 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.44 SMIACT# (System Management Interrupt Active)
Pin Attribute Output
Pin Location AG-03
Summary The processor acknowledges the assertion of SMI# with the
assertion of SMIACT# to indicate that the processor has
entered system management mode (SMM). The system logic
can use SMIACT# to enable SMM memory. See “SMI# (System
Management Interrupt)” on page 121 for more details.
See “System Management Mode (SMM)” on page 217 for more
details regarding SMM.
Driven The processor asserts SMIACT# after the last BRDY# of the last
pending bus cycle is sampled asserted (including all pending
write cycles) and after EWBE# is sampled asserted (if EWBE#
is masked off, then SMIACT# is not affected by EWBE#).
SMIACT# remains asserted until after the last BRDY# of the
last pending bus cycle associated with exiting SMM is sampled
asserted.
SMIACT# remains asserted during any flush, internal snoop, or
writeback cycle due to an inquire cycle.
Chapter 5 Signal Descriptions 123
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.45 STPCLK# (Stop Clock)
Pin Attribute Input, Internal Pullup
Pin Location V-34
Summary The assertion of STPCLK# causes the processor to enter the
Stop Grant state, during which the processor’s internal clock is
stopped. From the Stop Grant state, the processor can
subsequently transition to the Stop Clock state, in which the
bus clock CLK is stopped. Upon recognizing STPCLK#, the
processor performs the following actions, in the order shown:
1. Flushes its instruction pipelines
2. Completes all pending and in-progress bus cycles
3. Acknowledges the STPCLK# assertion by executing a Stop
Grant special bus cycle (see Table 23 on page 132)
4. Stops its internal clock after BRDY# of the Stop Grant
special bus cycle is sampled asserted and after EWBE# is
sampled asserted (if EWBE# is masked off, then entry into
the Stop Grant state is not affected by EWBE#)
5. Enters the Stop Clock state if the system logic stops the bus
clock CLK (optional)
See “Clock Control States” on page 247 for more details
regarding clock control.
Sampled STPCLK# is sampled as a level-sensitive input on every clock
edge but is not recognized until the next instruction boundary.
System logic can drive the signal either synchronously or
asynchronously. If it is asserted asynchronously, it must be
asserted for a minimum pulse width of two clocks.
STPCLK# must remain asserted until recognized, which is
indicated by the completion of the Stop Grant special cycle.
124 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.46 TCK (Test Clock)
Pin Attribute Input, Internal Pullup
Pin Location M-34
Summary TCK is the clock for boundary-scan testing using the Test
Access Port (TAP). See “Boundary-Scan Test Access Port
(TAP)” on page 229 for details regarding the operation of the
TAP controller.
Sampled The processor always samples TCK, except while TRST# is
asserted.
5.47 TDI (Test Data Input)
Pin Attribute Input, Internal Pullup
Pin Location N-35
Summary TDI is the serial test data and instruction input for
boundary-scan testing using the Test Access Port (TAP). See
“Boundary-Scan Test Access Port (TAP)” on page 229 for details
regarding the operation of the TAP controller.
Sampled The processor samples TDI on every rising TCK edge, but only
while in the Shift-IR and Shift-DR states.
5.48 TDO (Test Data Output)
Pin Attribute Output
Pin Location N-33
Summary TDO is the serial test data and instruction output for
boundary-scan testing using the Test Access Port (TAP). See
“Boundary-Scan Test Access Port (TAP)” on page 229 for details
regarding the operation of the TAP controller.
Driven and Floated The processor drives TDO on every falling TCK edge, but only
while in the Shift-IR and Shift-DR states. TDO is floated at all
other times.
Chapter 5 Signal Descriptions 125
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.49 TMS (Test Mode Select)
Pin Attribute Input, Internal Pullup
Pin Location P-34
Summary TMS specifies the test function and sequence of state changes
for boundary-scan testing using the Test Access Port (TAP). See
“Boundary-Scan Test Access Port (TAP)” on page 229 for details
regarding the operation of the TAP controller.
Sampled The processor samples TMS on every rising TCK edge. If TMS is
sampled High for five or more consecutive clocks, the TAP
controller enters its Test-Logic-Reset state, regardless of the
controller state. This action is the same as that achieved by
asserting TRST#.
5.50 TRST# (Test Reset)
Pin Attribute Input, Internal Pullup
Pin Location Q-33
Summary The assertion of TRST# initializes the Test Access Port (TAP) by
resetting its state machine to the Test-Logic-Reset state. See
“Boundary-Scan Test Access Port (TAP)” on page 229 for details
regarding the operation of the TAP controller.
Sampled TRST# is a completely asynchronous input that does not
require a minimum setup and hold time relative to TCK. See
Table 64 on page 280 for the minimum pulse width
requirement.
126 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.51 VCC2DET (VCC2 Detect)
Pin Attribute Output
Pin Location AL-01
Summary VCC2DET is internally tied to VSS (logic level 0) to indicate to
the system logic that it must supply the specified dual-voltage
requirements to the VCC2 and VCC3 pins. The VCC2 pins supply
voltage to the processor core, independent of the voltage
supplied to the I/O buffers on the VCC3 pins. Upon sampling
VCC2DET Low, system logic should sample VCC2H/L# to
identify core voltage requirements.
Driven VCC2DET always equals 0 and is never floated—even during
the three-state test mode.
Chapter 5 Signal Descriptions 127
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.52 VCC2H/L# (VCC2 High/Low)
Pin Attribute Output
Pin Location AN-05
Summary VCC2H/L# is internally tied to VSS (logic level 0) to indicate to
the system logic that it must supply the specified processor core
voltage to the VCC2 pins. The VCC2 pins supply voltage to the
processor core, independent of the voltage supplied to the I/O
buffers on the VCC3 pins.
Upon sampling VCC2DET Low to identify dual-voltage
processor requirements, system logic should sample VCC2H/L#
to identify the core voltage requirements for 2.9V and 3.2V
products (High) or 2.4 V, 2.2 V, and 1.9 V products (Low).
Driven VCC2H/L# always equals 0 and is never floated for 2.4 V, 2.2 V,
and 1.9 V products—even during the three-state test mode. To
ensure proper operation for 2.9V and 3.2V products, system
logic that samples VCC2H/L# should design a weak pullup
resistor for this signal.
The output pin float conditions for VCC2DET and VCC2H/L#
are listed in Table 17.
Table 17. Output Pin Float Conditions
Name Floated At:
VCC2DET1
Notes:
1. All outputs except VCC2DET, VCC2H/L#, and TDO float during the three-state test mode.
Always Driven
VCC2H/L# Always Driven
128 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.53 W/R# (Write/Read)
Pin Attribute Output
Pin Location AM-06
Summary The processor drives W/R# to indicate whether it is performing
a write or a read cycle on the bus. In addition, W/R# is used to
define other bus cycles, including interrupt acknowledge and
special cycles. See Table 23 on page 132 for more details.
Driven and Floated W/R# is driven off the same clock edge as ADS# and remains in
the same state until the clock edge on which NA# or the last
expected BRDY# of the cycle is sampled asserted. W/R# is
driven during memory cycles, I/O cycles, special bus cycles, and
interrupt acknowledge cycles.
W/R# is floated off the clock edge on which BOFF# is sampled
asserted and off the clock edge on which the processor asserts
HLDA in response to HOLD.
Chapter 5 Signal Descriptions 129
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
5.54 WB/WT# (Writeback or Writethrough)
Pin Attribute Input
Pin Location AA-05
Summary WB/WT#, together with PWT, specifies the data cache-line state
during cacheable read misses and write hits to shared cache
lines.
■If WB/WT# = 0 or PWT = 1 during a cacheable read miss or
write hit to a shared cache line, the accessed line is cached
in the shared state. This is referred to as the writethrough
state, because all write cycles to this cache line are driven
externally on the bus.
■If WB/WT# = 1 and PWT = 0 during a cacheable read miss or
a write hit to a shared cache line, the accessed line is cached
in the exclusive state. Subsequent write hits to the same line
cause its state to transition from exclusive to modified. This
is referred to as the writeback state, because the data cache
can contain modified cache lines that are subject to be
written back—referred to as a writeback cycle—as the result
of an inquire cycle, an internal snoop, a flush operation, or
the WBINVD instruction.
Sampled WB/WT# is sampled on the clock edge on which the first BRDY#
or NA# of a bus cycle is sampled asserted. If the cycle is a burst
read, WB/WT# is ignored during the last three assertions of
BRDY#. WB/WT# is sampled during memory read and
non-writeback write cycles and is ignored during all other types
of cycles.
130 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.55 Pin Tables by Type
Table 18. Input Pin Types
Name Type Name Type
A20M#1
Notes:
1. These level-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must be asserted for a minimum pulse width of two clocks.
Asynchronous IGNNE#1Asynchronous
AHOLD Synchronous INIT2
2. These edge-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must have been negated at least two clocks prior to assertion and
must remain asserted at least two clocks.
Asynchronous
BF[2:0]3
3. BF[2:0] are sampled during the falling transition of RESET. They must meet a minimum setup time of 1.0 ms and a minimum
hold time of two clocks relative to the negation of RESET.
Synchronous INTR1Asynchronous
BOFF# Synchronous INV Synchronous
BRDY# Synchronous KEN# Synchronous
BRDYC# Synchronous NA# Synchronous
CLK Clock NMI2Asynchronous
EADS# Synchronous RESET4,5
4. During the initial power-on reset of the processor, RESET must remain asserted for a minimum of 1.0 ms after CLK and VCC
reach specification before it is negated.
5. During a warm reset, while CLK and VCC are within their specification, RESET must remain asserted for a minimum of 15 clocks
prior to its negation.
Asynchronous
EWBE#6
6. When EFER[3] is 1, EWBE# is ignored by the processor.
Synchronous SMI#2Asynchronous
FLUSH#2,7
7. FLUSH# is also sampled during the falling transition of RESET and can be asserted synchronously or asynchronously. To be
sampled on a specific clock edge, setup and hold times must be met relative to the clock edge before the clock edge on which
RESET is sampled negated. If asserted asynchronously, FLUSH# must meet a minimum setup and hold time of two clocks
relative to the negation of RESET.
Asynchronous STPCLK#1Asynchronous
HOLD Synchronous WB/WT# Synchronous
Chapter 5 Signal Descriptions 131
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 19. Output Pin Float Conditions
Name Floated At:1Name Floated At:1
A[4:3]2,3 HLDA, AHOLD, BOFF# HLDA Always Driven
ADS#2HLDA, BOFF# LOCK#2HLDA, BOFF#
ADSC#2HLDA, BOFF# M/IO#2HLDA, BOFF#
APCHK# Always Driven PCD2HLDA, BOFF#
BE[7:0]#2HLDA, BOFF# PCHK# Always Driven
BREQ Always Driven PWT2HLDA, BOFF#
CACHE#2HLDA, BOFF# SCYC2HLDA, BOFF#
D/C#2HLDA, BOFF# SMIACT# Always Driven
FERR# Always Driven VCC2DET Always Driven
HIT# Always Driven VCC2H/L# Always Driven
HITM# Always Driven W/R#2HLDA, BOFF#
Notes:
1. All outputs except VCC2DET, VCC2H/L#, and TDO float during three-state test mode.
2. Floated off the clock edge that BOFF# is sampled asserted and off the clock edge that HLDA is asserted.
3. Floated off the clock edge that AHOLD is sampled asserted.
Table 20. Input/Output Pin Float Conditions
Name Floated At:1
Notes:
1. All outputs except VCC2DET and TDO float during three-state test mode.
A[31:5]2,3
2. Floated off the clock edge that BOFF# is sampled asserted and off the clock edge that
HLDA is asserted.
3. Floated off the clock edge that AHOLD is sampled asserted.
HLDA, AHOLD, BOFF#
AP2,3 HLDA, AHOLD, BOFF#
D[63:0]2HLDA, BOFF#
DP[7:0]2HLDA, BOFF#
Table 21. Test Pins
Name Type Comment
TCK Clock
TDI Input Sampled on the rising edge of TCK
TDO Output Driven on the falling edge of TCK
TMS Input Sampled on the rising edge of TCK
TRST# Input Asynchronous (Independent of TCK)
132 Signal Descriptions Chapter 5
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
5.56 Bus Cycle Definitions
Table 22. Bus Cycle Definition
Bus Cycle Initiated Generated by CPU Generated
by System Logic
M/IO# D/C# W/R# CACHE# KEN#
Code Read, Instruction Cache Line Fill 1 0 0 0 0
Code Read, Noncacheable 1 0 0 1 x
Code Read, Noncacheable 1 0 0 x 1
Encoding for Special Cycle 0011 x
Interrupt Acknowledge 0 0 0 1 x
I/O Read 0 1 0 1 x
I/O Write 0 1 1 1 x
Memory Read, Data Cache Line Fill 1 1 0 0 0
Memory Read, Noncacheable 1 1 0 1 x
Memory Read, Noncacheable 1 1 0 x 1
Memory Write, Data Cache Writeback 1 1 1 0 x
Memory Write, Noncacheable 1 1 1 1 x
Notes:
x means “don’t care”.
Table 23. Special Cycles
Special Cycle
A4
BE7#
BE6#
BE5#
BE4#
BE3#
BE2#
BE1#
BE0#
M/IO#
D/C#
W/R#
CACHE#
KEN#
Stop Grant 111111011 0 0 1 1 x
Flush Acknowledge
(FLUSH# sampled asserted) 011101111 0 01 1 x
Writeback (WBINVD
instruction) 011110111 0 01 1 x
Halt 0 1 1 1 1 1 0 1 1 0 0 1 1 x
Flush (INVD, WBINVD
instruction) 011111101 0 01 1 x
Shutdown 0 1 1 1 1 1 1 1 0 0 0 1 1 x
Notes:
x means “don’t care”.
Chapter 6 Bus Cycles 133
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
6 Bus Cycles
The following sections describe and illustrate the timing and
relationship of bus signals during various types of bus cycles. A
representative set of bus cycles is illustrated.
6.1 Timing Diagrams
The timing diagrams illustrate the signals on the external local
bus as a function of time, as measured by the bus clock (CLK).
Bus Clock (CLK) Throughout this chapter, the term clock refers to a single
bus-clock cycle. A clock extends from one rising CLK edge to
the next rising CLK edge. The processor samples and drives
most signals relative to the rising edge of CLK. The exceptions
to this rule include the following:
■BF[2:0]—Sampled on the falling edge of RESET
■FLUSH#—Sampled on the falling edge of RESET, also
sampled on the rising edge of CLK
■All inputs and outputs are sampled relative to TCK in
boundary-scan test mode. Inputs are sampled on the rising
edge of TCK, outputs are driven off of the falling edge of
TCK.
Waveform
Definitions For each signal in the timing diagrams, the High level
represents 1, the Low level represents 0, and the Middle level
represents the floating (high-impedance) state.
When both the High and Low levels are shown, the meaning
depends on the signal:
■A single signal indicates ‘don’t care’.
■In the case of bus activity, if both High and Low levels are
shown, it indicates that the processor, alternate master, or
system logic is driving a value, but this value may or may not
be valid. (For example, the value on the address bus is valid
only during the assertion of ADS#, but addresses are also
driven on the bus at other times.)
Figure 50 defines the different waveform representations.
134 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Active High Signals For all active High signals, the term asserted means the signal is
in the High-voltage state and the term negated means the signal
is in the Low-voltage state.
Active Low Signals For all active Low signals, the term asserted means the signal is
in the Low-voltage state and the term negated means the signal
is in the High-voltage state.
Figure 50. Waveform Definitions
Waveform
Don’t care or bus is driven
Description
Signal or bus is changing from Low to High
Signal or bus is changing from High to Low
Bus is changing
Bus is changing from valid to invalid
Signal or bus is floating
Denotes multiple clock periods
Chapter 6 Bus Cycles 135
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
6.2 Bus States
Figure 51. Bus State Machine Diagram
Last BRDY#
Last BRDY#
Bus Transition?
No
Yes
Yes
Yes
Yes
No
No
No
No
Pending
Request?
NA# Sampled
Asserted?
NA# Sampled
Asserted?
Asserted?
Asserted?
Addr
Data
Pipe-A
Pipe-D
Idle
Trans
Bus State
Branch Condition
Idle Data
Pipeline
Address
Transition
Address
Pipeline
Data
Note: The processor transitions to the IDLE state on the clock edge on which BOFF# or RESET is sampled asserted.
Data-NA#
Requested
Data-NA#
Last BRDY#
Yes
No
Asserted?
Pending
Request? No
Yes
No
Yes
Yes
NA# Sampled
Asserted?
Yes
No
136 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Idle The processor does not drive the system bus in the Idle state
and remains in this state until a new bus cycle is requested. The
processor enters this state off the clock edge on which the last
BRDY# of a cycle is sampled asserted during the following
conditions:
■The processor is in the Data state
■The processor is in the Data-NA# Requested state and no
internal pending cycle is requested
In addition, the processor is forced into this state when the
system logic asserts RESET or BOFF#. The transition to this
state occurs on the clock edge on which RESET or BOFF# is
sampled asserted.
Address In this state, the processor drives ADS# to indicate the
beginning of a new bus cycle by validating the address and
control signals. The processor remains in this state for one clock
and unconditionally enters the Data state on the next clock
edge.
Data In the Data state, the processor drives the data bus during a
write cycle or expects data to be returned during a read cycle.
The processor remains in this state until either NA# or the last
BRDY# is sampled asserted. If the last BRDY# is sampled
asserted or both the last BRDY# and NA# are sampled asserted
on the same clock edge, the processor enters the Idle state. If
NA# is sampled asserted first, the processor enters the
Data-NA# Requested state.
Data-NA# Requested If the processor samples NA# asserted while in the Data state
and the current bus cycle is not completed (the last BRDY# is
not sampled asserted), it enters the Data-NA# Requested state.
The processor remains in this state until either the last BRDY#
is sampled asserted or an internal pending cycle is requested. If
the last BRDY# is sampled asserted before the processor drives
a new bus cycle, the processor enters the Idle state (no internal
pending cycle is requested) or the Address state (processor has
a internal pending cycle).
Pipeline Address In this state, the processor drives ADS# to indicate the
beginning of a new bus cycle by validating the address and
control signals. In this state, the processor is still waiting for the
current bus cycle to be completed (until the last BRDY# is
Chapter 6 Bus Cycles 137
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
sampled asserted). If the last BRDY# is not sampled asserted,
the processor enters the Pipeline Data state.
If the processor samples the last BRDY# asserted in this state, it
determines if a bus transition is required between the current
bus cycle and the pipelined bus cycle. A bus transition is
required when the data bus direction changes between bus
cycles, such as a memory write cycle followed by a memory read
cycle. If a bus transition is required, the processor enters the
Transition state for one clock to prevent data bus contention. If
a bus transition is not required, the processor enters the Data
state.
The processor does not transition to the Data-NA# Requested
state from the Pipeline Address state because the processor
does not begin sampling NA# until it has exited the Pipeline
Address state.
Pipeline Data Two bus cycles are executing concurrently in this state. The
processor cannot issue any additional bus cycles until the
current bus cycle is completed. The processor drives the data
bus during write cycles or expects data to be returned during
read cycles for the current bus cycle until the last BRDY# of the
current bus cycle is sampled asserted.
If the processor samples the last BRDY# asserted in this state, it
determines if a bus transition is required between the current
bus cycle and the pipelined bus cycle. If the bus transition is
required, the processor enters the Transition state for one clock
to prevent data bus contention. If a bus transition is not
required, the processor enters the Data state (NA# was not
sampled asserted) or the Data-NA# Requested state (NA# was
sampled asserted).
Transition The processor enters the Transition state for one clock during
data bus transitions and enters the Data state on the next clock
edge if NA# is not sampled asserted. The sole purpose of this
state is to avoid bus contention caused by bus transitions during
pipeline operation.
138 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
6.3 Memory Reads and Writes
The AMD-K6-2E processor performs single or burst memory bus
cycles.
■The single-transfer memory bus cycle transfers 1, 2, 4, or 8
bytes and requires a minimum of two clocks.
■Misaligned instructions or operands result in a split cycle,
which requires multiple transactions on the bus.
■A burst cycle consists of four back-to-back 8-byte (64-bit)
transfers on the data bus.
Single-Transfer
Memory Read and
Write
Figure 52 on page 139 shows a single-transfer read from memory,
followed by two single-transfer writes to memory. For the
memory read cycle, the processor asserts ADS# for one clock to
validate the bus cycle and also drives A[31:3], BE[7:0]#, D/C#,
W/R#, and M/IO# to the bus. The processor then waits for the
system logic to return the data on D[63:0] (with DP[7:0] for
parity checking) and assert BRDY#. The processor samples
BRDY# on every clock edge starting with the clock edge after
the clock edge that negates ADS#. See “BRDY# (Burst Ready)”
on page 95.
During the read cycle, the processor drives PCD, PWT, and
CACHE# to indicate its caching and cache-coherency intent for
the access. The system logic returns KEN# and WB/WT# to
either confirm or change this intent. If the processor asserts
PCD and negates CACHE#, the accesses are noncacheable, even
though the system logic asserts KEN# during the BRDY# to
indicate its support for cacheability. The processor (which
drives CACHE#) and the system logic (which drives KEN#) must
agree in order for an access to be cacheable.
The processor can drive another cycle (in this example, a write
cycle) by asserting ADS# off the next clock edge after BRDY# is
sampled asserted. Therefore, an idle clock is guaranteed
between any two bus cycles. The processor drives D[63:0] with
valid data one clock edge after the clock edge on which ADS# is
asserted. To minimize processor idle times, the system logic
stores the address and data in write buffers, returns BRDY#, and
performs the store to memory later. If the processor samples
EWBE# negated during a write cycle, it suspends certain
activities until EWBE# is sampled asserted. See “EWBE#
(External Write Buffer Empty)” on page 102. In Figure 52, the
Chapter 6 Bus Cycles 139
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
second write cycle occurs during the execution of a serializing
instruction. The processor delays the following cycle until
EWBE# is sampled asserted.
Figure 52. Non-Pipelined Single-Transfer Memory Read/Write and Write Delayed by EWBE#
Read Cycle Write Cycle (Next Cycle Delayed by EWBE#)
Write Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
BREQ
D[63:0]
DP[7:0]
CACHE#
EWBE#
KEN#
BRDY#
WB/WT#
ADDR DATA IDLE ADDR ADDRDATA IDLE
DATA DATA DATA IDLE IDLE IDLE IDLE IDLE ADDR
140 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Misaligned
Single-Transfer
Memory Read and
Write
Figure 53 on page 141 shows a misaligned (split) memory read
followed by a misaligned memory write. Any cycle that is not
aligned as defined in “SCYC (Split Cycle)” on page 120 is
considered misaligned. When the processor encounters a
misaligned access, it determines the appropriate pair of bus
cycles—each with its own ADS# and BRDY#— required to
complete the access.
The AMD-K6-2E processor performs misaligned memory reads
and memory writes using least-significant bytes (LSBs) first
followed by most-significant bytes (MSBs). Table 24 shows the
order. In the first memory read cycle in Figure 53, the processor
reads the least-significant bytes. Immediately after the
processor samples BRDY# asserted, it drives the second bus
cycle to read the most-significant bytes to complete the
misaligned transfer.
Similarly, the misaligned memory write cycle in Figure 53
transfers the LSBs to the memory bus first. In the next cycle,
after the processor samples BRDY# asserted, the MSBs are
written to the memory bus.
Table 24. Bus-Cycle Order During Misaligned Memory Transfers
Type of Access First Cycle Second Cycle
Memory Read LSBs MSBs
Memory Write LSBs MSBs
Chapter 6 Bus Cycles 141
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 53. Misaligned Single-Transfer Memory Read and Write
LSB MSB LSB MSB
Memory Read (Misaligned) Memory Write (Misaligned)
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
D[63:0]
BRDY#
ADDR DATA DATA IDLE ADDR DATA DATA IDLE ADDR DATA DATA IDLE
DATA ADDR DATA DATA IDL
E
DATA
142 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Burst Reads and
Pipelined Burst Reads Figure 54 on page 143 shows normal burst read cycles and a
pipelined burst read cycle. The AMD-K6-2E processor drives
CACHE# and ADS# together to specify that the current bus
cycle is a burst cycle. If the processor samples KEN# asserted
with the first BRDY#, it performs burst transfers. During the
burst transfers, the system logic must ignore BE[7:0]# and must
return all eight bytes beginning at the starting address the
processor asserts on A[31:3]. Depending on the starting
address, the system logic must determine the successive
quadword addresses (A[4:3]) for each transfer in a burst, as
shown in Table 25. The processor expects the second, third, and
fourth quadwords to occur in the sequences shown in Table 25.
In Figure 54, the processor drives CACHE# throughout all burst
read cycles. In the first burst read cycle, the processor drives
ADS# and CACHE#, then samples BRDY# on every clock edge
starting with the clock edge after the clock edge that negates
ADS#. The processor samples KEN# asserted on the clock edge
on which the first BRDY# is sampled asserted, executes a
32-byte burst read cycle, and expects a total of four BRDY#
signals. An ideal no-wait state access is shown in Figure 54,
whereas most system logic solutions add wait states between
the transfers.
The second burst read cycle illustrates a similar sequence, but
the processor samples NA# asserted on the same clock edge
that the first BRDY# is sampled asserted. NA# assertion
indicates the system logic is requesting the processor to output
the next address early (also known as a pipeline transfer
request). Without waiting for the current cycle to complete, the
processor drives ADS# and related signals for the next burst
cycle. Pipelining can reduce processor cycle-to-cycle idle times.
Table 25. A[4:3] Address-Generation Sequence During Bursts
Address Driven By
Processor on A[4:3]
A[4:3] Addresses of Subsequent
Quadwords1 Generated by System Logic
Notes:
1. quadword = 8 bytes
Quadword 1 Quadword 2 Quadword 3 Quadword 4
00b 01b 10b 11b
01b 00b 11b 10b
10b 11b 00b 01b
11b 10b 01b 00b
Chapter 6 Bus Cycles 143
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 54. Burst Reads and Pipelined Burst Reads
-NA -ADDR
DATA1 DATA2 DATA3
Burst Read Pipelined Burst Read
ADDR1 ADDR2 ADDR3
Burst Read
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
NA#
D[63:0]
CACHE#
KEN#
BRDY#
ADDR DATA IDLEDATADATADATA ADDR DATA DATA DATA PIPE DATA DATA DATA DATA IDLE
144 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Burst Writeback Figure 55 on page 145 shows a burst read followed by a
writeback transaction. The AMD-K6-2E processor initiates
writebacks under the following conditions:
■Replacement—If a cache-line fill is initiated for a cache line
currently filled with valid entries, the processor selects a
line for replacement based on a least-recently-used (LRU)
algorithm for the instruction cache, and a
least-recently-allocated (LRA) algorithm for the data cache.
Before a replacement is made to an L1 data cache line that is
in the Modified state, the modified line is scheduled to be
written back to memory.
■Internal Snoop—The processor snoops its instruction cache
during read or write misses to its data cache, and it snoops
its data cache during read misses to its instruction cache.
This snooping is performed to determine whether the same
address is stored in both caches, a situation that implies the
occurrence of self-modifying code. If a snoop hits a data
cache line in the Modified state, the line is written back to
memory before being invalidated.
■WBINVD Instruction—When the processor executes a
WBINVD instruction, it writes back all modified lines in the
data cache and then invalidates all lines in both caches.
■Cache Flush—When the processor samples FLUSH#
asserted, it executes a flush acknowledge special cycle and
writes back all modified lines in the data cache and then
invalidates all lines in both caches.
The processor drives writeback cycles during inquire or cache
flush cycles. The writeback shown in Figure 55 is caused by a
cache-line replacement. The processor completes the burst read
cycle that fills the cache line. Immediately following the burst
read cycle is the burst writeback cycle that represents the
modified line to be written back to memory. D[63:0] are driven
one clock edge after the clock edge on which ADS# is asserted
and are subsequently changed off the clock edge on which each
of the four BRDY# signals of the burst cycle are sampled
asserted.
Chapter 6 Bus Cycles 145
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 55. Burst Writeback due to Cache-Line Replacement
Burst Read Burst Writeback from L1 Cache
CLK
A[31:3]
BE[7:0]#
ADS#
CACHE#
M/IO#
D/C#
W/R#
D[63:0]
KEN#
BRDY#
WB/WT#
ADDR DATA IDLE
DATA
DATA DATA ADDR DATA IDLE
DATA
DATA DATA
146 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
6.4 I/O Read and Write
Basic I/O Read and
Write The processor accesses I/O when it executes an I/O instruction
(for example, IN or OUT). Figure 56 shows an I/O read followed
by an I/O write. The processor drives M/IO# Low and D/C# High
during I/O cycles. In this example, the first cycle shows a single
wait state I/O read cycle. It follows the same sequence as a
single-transfer memory read cycle. The processor drives ADS#
to initiate the bus cycle, then it samples BRDY# on every clock
edge starting with the clock edge after the clock edge that
negates ADS#. The system logic must return BRDY# to
complete the cycle. When the processor samples BRDY#
asserted, it can assert ADS# for the next cycle off the next clock
edge. (In this example, an I/O write cycle.)
The I/O write cycle is similar to a memory write cycle, but the
processor drives M/IO# low during an I/O write cycle. The
processor asserts ADS# to initiate the bus cycle. The processor
drives D[63:0] with valid data one clock edge after the clock
edge on which ADS# is asserted. The system logic must assert
BRDY# when the data is properly stored to the I/O destination.
The processor samples BRDY# on every clock edge starting with
the clock edge after the clock edge that negates ADS#. In this
example, two wait states are inserted while the processor waits
for BRDY# to be asserted.
Figure 56. Basic I/O Read and Write
I/O Read Cycle I/O Write Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
D[63:0]
BRDY#
ADDR DATA DATA IDLE ADDR DATA DATA DATA IDLE
Chapter 6 Bus Cycles 147
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Misaligned I/O Read
and Write Table 26 shows the misaligned I/O read and write cycle order
executed by the AMD-K6-2E processor. In Figure 57, the
least-significant bytes (LSBs) are transferred first. Immediately
after the processor samples BRDY# asserted, it drives the
second bus cycle to transfer the most-significant bytes (MSBs)
to complete the misaligned bus cycle.
Figure 57. Misaligned I/O Transfer
Table 26. Bus-Cycle Order During Misaligned I/O Transfers
Type of Access First Cycle Second Cycle
I/O Read LSBs MSBs
I/O Write LSBs MSBs
LSB MSB LSB MSB
Misaligned I/O Read Misaligned I/O Write
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
D[63:0]
BRDY#
ADDR DATA DATA IDLE ADDR DATA DATA IDLE ADDR DATA DATA
DATA IDLE ADDR DATA DATA
DATA IDLE
SCYC
148 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
6.5 Inquire and Bus Arbitration Cycles
The AMD-K6-2E processor provides built-in level-one data and
instruction caches. Each cache is 32 Kbytes and two-way
set-associative. The system logic or other bus master devices
can initiate an inquire cycle to maintain cache/memory
coherency. In response to the inquire cycle, the processor
compares the inquire address with its cache tag addresses in
both caches, and, if necessary, updates the MESI state of the
cache line and performs writebacks to memory.
An inquire cycle can be initiated by asserting AHOLD, BOFF#,
or HOLD. AHOLD is exclusively used to support inquire cycles.
During AHOLD-initiated inquire cycles, the processor only
floats the address bus. BOFF# provides the fastest access to the
bus because it aborts any processor cycle that is in-progress,
whereas AHOLD and HOLD both permit an in-progress bus
cycle to complete. During HOLD-initiated and BOFF#-initiated
inquire cycles, the processor floats all of its bus-driving signals.
Hold and Hold
Acknowledge Cycle The system logic or another bus device can assert HOLD to
initiate an inquire cycle or to gain full control of the bus. When
the AMD-K6-2E processor samples HOLD asserted, it
completes any in-progress bus cycle and asserts HLDA to
acknowledge release of the bus. The processor floats the
following signals off the same clock edge on which HLDA is
asserted:
Figure 58 shows a basic HOLD/HLDA operation. In this
example, the processor samples HOLD asserted during the
memory read cycle. It continues the current memory read cycle
until BRDY# is sampled asserted. The processor drives HLDA
and floats its outputs one clock edge after the last BRDY# of the
cycle is sampled asserted. The system logic can assert HOLD for
as long as it needs to utilize the bus. The processor samples
■A[31:3] ■DP[7:0]
■ADS# ■LOCK#
■AP# ■M/IO#
■BE[7:0]# ■PCD
■CACHE# ■PWT
■D[63:0] ■SCYC
■D/C# ■W/R#
Chapter 6 Bus Cycles 149
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
HOLD on every clock edge but does not assert HLDA until any
in-progress cycle or sequence of locked cycles is completed.
When the processor samples HOLD negated during a hold
acknowledge cycle, it negates HLDA off the next clock edge.
The processor regains control of the bus and can assert ADS#
off the same clock edge on which HLDA is negated.
Figure 58. Basic HOLD/HLDA Operation
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
D[63:0]
HOLD
HLDA
BRDY#
150 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
HOLD-Initiated
Inquire Hit to Shared
or Exclusive Line
Figure 59 on page 151 shows a HOLD-initiated inquire cycle. In
this example, the processor samples HOLD asserted during the
burst memory read cycle. The processor completes the current
cycle (until the last expected BRDY# is sampled asserted),
asserts HLDA, and floats its outputs as described on “Hold and
Hold Acknowledge Cycle” on page 148.
The system logic drives an inquire cycle within the hold
acknowledge cycle. It asserts EADS#, which validates the
inquire address on A[31:5]. If EADS# is sampled asserted
before HOLD is sampled negated, the processor recognizes it as
a valid inquire cycle.
In Figure 59, the processor asserts HIT# and negates HITM# on
the clock edge after the clock edge on which EADS# is sampled
asserted, indicating the current inquire cycle hit a shared or
exclusive cache line. (Shared and exclusive cache lines have not
been modified and do not need to be written back.) During an
inquire cycle, the processor samples INV to determine whether
the addressed cache line found in the processor’s instruction or
data cache transitions to the Invalid state or the Shared state.
In this example, the processor samples INV asserted with
EADS#, which invalidates the cache line.
The system logic can negate HOLD off the same clock edge on
which EADS# is sampled asserted. The processor continues
driving HIT# in the same state until the next inquire cycle.
HITM# is not asserted unless HIT# is asserted.
Chapter 6 Bus Cycles 151
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 59. HOLD-Initiated Inquire Hit to Shared or Exclusive Line
Burst Memory Read Inquire
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
HIT#
HITM#
D[63:0]
KEN#
BRDY#
HOLD
HLDA
EADS#
INV
152 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
HOLD-Initiated
Inquire Hit to
Modified Line
Figure 60 on page 153 shows the same sequence as Figure 59,
but in Figure 60, the inquire cycle hits a modified line and the
processor asserts both HIT# and HITM#. In this example, the
processor performs a writeback cycle immediately after the
inquire cycle. It updates the modified cache line to the external
memory (normally, external cache or DRAM). The processor
uses the address (A[31:5]) that was latched during the inquire
cycle to perform the writeback cycle. The processor asserts
HITM# throughout the writeback cycle and negates HITM# one
clock edge after the last expected BRDY# of the writeback is
sampled asserted.
When the processor samples EADS# during the inquire cycle, it
also samples INV to determine the cache line MESI state after
the inquire cycle. If INV is sampled asserted during an inquire
cycle, the processor transitions the line (if found) to the Invalid
state, regardless of its previous state. The cache line
invalidation operation is not visible on the bus. If INV is
sampled negated during an inquire cycle, the processor
transitions the line (if found) to the Shared state. In Figure 60
the processor samples INV asserted during the inquire cycle.
In a HOLD-initiated inquire cycle, the system logic can negate
HOLD off the same clock edge on which EADS# is sampled
asserted. The processor drives HIT# and HITM# on the clock
edge after the clock edge on which EADS# is sampled asserted.
Chapter 6 Bus Cycles 153
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 60. HOLD-Initiated Inquire Hit to Modified Line
Burst Memory Read Inquire Writeback Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
EADS#
HIT#
HITM#
D[63:0]
KEN#
BRDY#
HOLD
HLDA
INV
154 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
AHOLD-Initiated
Inquire Miss AHOLD can be asserted by the system to initiate one or more
inquire cycles. To allow the system to drive the address bus
during an inquire cycle, the processor floats A[31:3] and AP off
the clock edge on which AHOLD is sampled asserted. The data
bus and all other control and status signals remain under the
control of the processor and are not floated. This functionality
allows a bus cycle in progress when AHOLD is sampled asserted
to continue to completion. The processor resumes driving the
address bus off the clock edge on which AHOLD is sampled
negated.
In Figure 61 on page 155, the processor samples AHOLD
asserted during the memory burst read cycle, and it floats the
address bus off the same clock edge on which it samples AHOLD
asserted. While the processor still controls the bus, it completes
the current cycle until the last expected BRDY# is sampled
asserted. The system logic drives EADS# with an inquire
address on A[31:5] during an inquire cycle. The processor
samples EADS# asserted and compares the inquire address to
its tag address in both the instruction and data caches. In Figure
61, the inquire address misses the tag address in the processor
(both HIT# and HITM# are negated). Therefore, the processor
proceeds to the next cycle when it samples AHOLD negated.
(The processor can drive a new cycle by asserting ADS# off the
same clock edge that it samples AHOLD negated.)
For an AHOLD-initiated inquire cycle to be recognized, the
processor must sample AHOLD asserted for at least two
consecutive clocks before it samples EADS# asserted. If the
processor detects an address parity error during an inquire
cycle, APCHK# is asserted for one clock. The system logic must
respond appropriately to the assertion of this signal.
Chapter 6 Bus Cycles 155
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 61. AHOLD-Initiated Inquire Miss
Read Inquire
CLK
A[31:3]
BE[7:0]#
AP
APCHK#
ADS#
HIT#
HITM#
D[63:0]
KEN#
BRDY#
AHOLD
EADS#
INV
156 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
AHOLD-Initiated
Inquire Hit to Shared
or Exclusive Line
In Figure 62, the processor asserts HIT# and negates HITM# off
the clock edge after the clock edge on which EADS# is sampled
asserted, indicating the current inquire cycle hits either a
shared or exclusive line. (HIT# is driven in the same state until
the next inquire cycle.) The processor samples INV asserted
during the inquire cycle and transitions the line to the Invalid
state regardless of its previous state.
During an AHOLD-initiated inquire cycle, the processor
samples AHOLD on every clock edge until it is negated. In
Figure 62, the processor asserts ADS# off the same clock on
which AHOLD is sampled negated. If the inquire cycle hits a
modified line, the processor performs a writeback cycle before
it drives a new bus cycle. The next section describes the
AHOLD-initiated inquire cycle that hits a modified line.
Chapter 6 Bus Cycles 157
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 62. AHOLD-Initiated Inquire Hit to Shared or Exclusive Line
Burst Memory Read Inquire
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
HIT#
HITM#
D[63:0]
KEN#
BRDY#
AHOLD
INV
EADS#
158 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
AHOLD-Initiated
Inquire Hit to
Modified Line
Figure 63 on page 159 shows an AHOLD-initiated inquire cycle
that hits a modified line. During the inquire cycle in this
example, the processor asserts both HIT# and HITM# on the
clock edge after the clock edge that it samples EADS# asserted.
This condition indicates that the cache line exists in the
processor’s data cache in the Modified state.
If the inquire cycle hits a modified line, the processor performs
a writeback cycle immediately after the inquire cycle to update
the modified cache line to shared memory (normally external
cache or DRAM). In Figure 63, the system logic holds AHOLD
asserted throughout the inquire cycle and the processor
writeback cycle. In this case, the processor is not driving the
address bus during the writeback cycle because AHOLD is
sampled asserted. The system logic writes the data to memory
by using its latched copy of the inquire cycle address. If the
processor samples AHOLD negated before it performs the
writeback cycle, it drives the writeback cycle by using the
address (A[31:5]) that it latched during the inquire cycle.
If INV is sampled asserted during an inquire cycle, the
processor transitions the line (if found) to the Invalid state,
regardless of its previous state (the cache invalidation
operation is not visible on the bus). If INV is sampled negated
during an inquire cycle, the processor transitions the line (if
found) to the Shared state. In either case, if the line is found in
the Modified state, the processor writes it back to memory
before changing its state. Figure 63 shows that the processor
samples INV asserted during the inquire cycle and invalidates
the cache line after the inquire cycle.
Chapter 6 Bus Cycles 159
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 63. AHOLD-Initiated Inquire Hit to Modified Line
Burst Memory Read Inquire Writeback
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
HIT#
HITM#
D[63:0]
KEN#
BRDY#
AHOLD
EADS#
INV
160 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
AHOLD Restriction When the system logic drives an AHOLD-initiated inquire
cycle, it must assert AHOLD for at least two clocks before it
asserts EADS#. This requirement guarantees the processor
recognizes and responds to the inquire cycle properly. The
processor’s 32 address bus drivers turn on almost immediately
after AHOLD is sampled negated. If the processor switches the
data bus (D[63:0] and DP[7:0]) during a write cycle off the same
clock edge that switches the address bus (A[31:3] and AP), the
processor switches 102 drivers simultaneously, which can lead
to ground-bounce spikes. Therefore, before negating AHOLD,
the following restrictions must be observed by the system logic:
■When the system logic negates AHOLD during a write cycle,
it must ensure that AHOLD is not sampled negated on the
clock edge on which BRDY# is sampled asserted (See Figure
64 on page 161).
■When the system logic negates AHOLD during a writeback
cycle, it must ensure that AHOLD is not sampled negated on
the clock edge on which ADS# is negated (See Figure 64).
■When a write cycle is pipelined into a read cycle, AHOLD
must not be sampled negated on the clock edge after the
clock edge on which the last BRDY# of the read cycle is
sampled asserted. This avoids the processor simultaneously
driving the data bus (for the pending write cycle) and the
address bus off this same clock edge.
Chapter 6 Bus Cycles 161
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 64. AHOLD Restriction
The system must ensure that AHOLD is not sampled negated on the clock edge that ADS# is negated.
Legal AHOLD negation during write cycle
Illegal AHOLD negation during write cycle
CLK
ADS#
W/R#
HITM#
EADS#
D[63:0]
BRDY#
AHOLD
The system must ensure that AHOLD is not sampled negated on the clock edge on which BRDY# is sampled asserted.
162 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Bus Backoff (BOFF#) BOFF# provides the fastest response among bus-hold inputs.
Either the system logic or another bus master can assert BOFF#
to gain control of the bus immediately. BOFF# is also used to
resolve potential deadlock problems that arise as a result of
inquire cycles. The processor samples BOFF# on every clock
edge. If BOFF# is sampled asserted, the processor
unconditionally aborts any cycles in progress and transitions to
a Bus Hold state. (See “BOFF# (Backoff)” on page 94.) Figure
65 on page 163 shows a read cycle that is aborted when the
processor samples BOFF# asserted even though BRDY# is
sampled asserted on the same clock edge. The read cycle is
restarted after BOFF# is sampled negated (KEN# must be in
the same state during the restarted cycle as its state during the
aborted cycle).
During a BOFF#-initiated inquire cycle that hits a shared or
exclusive line, the processor samples BOFF# negated and
restarts any bus cycle that was aborted when BOFF# was
asserted. If a BOFF#-initiated inquire cycle hits a modified line,
the processor performs a writeback cycle before it restarts the
aborted cycle.
If the processor samples BOFF# asserted on the same clock
edge that it asserts ADS#, ADS# is floated but the system logic
may erroneously interpret ADS# as asserted. In this case, the
system logic must properly interpret the state of ADS# when
BOFF# is negated.
Chapter 6 Bus Cycles 163
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 65. BOFF# Timing
Read Backoff Cycle Restart Read Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
BOFF#
D[63:0]
BRDY#
164 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Locked Cycles The processor asserts LOCK# during a sequence of bus cycles to
ensure the cycles are completed without allowing other bus
masters to intervene. Locked operations can consist of two to
five cycles. LOCK# is asserted during the following operations:
■An interrupt acknowledge sequence
■Descriptor Table accesses
■Page Directory and Page Table accesses
■XCHG instruction
■An instruction with an allowable LOCK prefix
In order to ensure that locked operations appear on the bus and
are visible to the entire system, any data operands addressed
during a locked cycle that reside in the processor’s cache are
flushed and invalidated from the cache prior to the locked
operation. If the cache line is in the Modified state, it is written
back and invalidated prior to the locked operation. Likewise,
any data read during a locked operation is not cached. The
processor negates LOCK# for at least one clock between
consecutive sequences of locked operations to allow the system
logic to arbitrate for the bus.
The processor asserts SCYC during misaligned locked transfers
on the D[63:0] data bus. The processor generates additional bus
cycles to complete the transfer of misaligned data.
Basic Locked
Operation Figure 66 on page 165 shows a pair of read-write bus cycles. It
represents a typical read-modify-write locked operation. The
processor asserts LOCK# off the same clock edge that it asserts
ADS# of the first bus cycle in the locked operation and holds it
asserted until the last expected BRDY# of the last bus cycle in
the locked operation is sampled asserted. (The processor
negates LOCK# off of the same clock edge.)
Chapter 6 Bus Cycles 165
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 66. Basic Locked Operation
Locked Read Cycle Locked Write Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
LOCK#
M/IO#
D/C#
W/R#
D[63:0]
BRDY#
ADDR DATA DATA DATA IDLE IDLE ADDR DATA DATA DATA IDLE IDLE ADDR
SCYC
166 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Locked Operation
with BOFF#
Intervention
Figure 67 on page 167 shows BOFF# asserted within a locked
read-write pair of bus cycles. In this example, the processor
asserts LOCK# with ADS# to drive a locked memory read cycle
followed by a locked memory write cycle. During the locked
memory write cycle in this example, the processor samples
BOFF# asserted. The processor immediately aborts the locked
memory write cycle and floats all its bus-driving signals,
including LOCK#. The system logic or another bus master can
initiate an inquire cycle or drive a new bus cycle one clock edge
after the clock edge on which BOFF# is sampled asserted. If the
system logic drives a BOFF#-initiated inquire cycle and hits a
modified line, the processor performs a writeback cycle before
it restarts the locked cycle (the processor asserts LOCK# during
the writeback cycle).
In Figure 67, the processor immediately restarts the aborted
locked write cycle by driving the bus off the clock edge on
which BOFF# is sampled negated. The system logic must ensure
the processor results for interrupted and uninterrupted locked
cycles are consistent. That is, the system logic must guarantee
the memory accessed by the processor is not modified during
the time another bus master controls the bus.
Chapter 6 Bus Cycles 167
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 67. Locked Operation with BOFF# Intervention
Locked Read Cycle Aborted Write Cycle Restart Write Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
LOCK#
M/IO#
D/C#
W/R#
BOFF#
D[63:0]
BRDY#
168 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Interrupt
Acknowledge In response to recognizing the system’s maskable interrupt
(INTR), the processor drives an interrupt acknowledge cycle at
the next instruction boundary. During an interrupt
acknowledge cycle, the processor drives a locked pair of read
cycles as shown in Figure 68 on page 169. The first read cycle is
not functional, and the second read cycle returns the interrupt
number on D[7:0] (00h–FFh). Table 27 shows the state of the
signals during an interrupt acknowledge cycle.
The system logic can drive INTR either synchronously or
asynchronously. If it is asserted asynchronously, it must be
asserted for a minimum pulse width of two clocks. To ensure it
is recognized, INTR must remain asserted until an interrupt
acknowledge sequence is complete.
Table 27. Interrupt Acknowledge Operation Definition
Processor
Outputs First Bus Cycle Second Bus Cycle
D/C# Low Low
M/IO# Low Low
W/R# Low Low
BE[7:0]# EFh FEh (low byte enabled)
A[31:3] 0000_0000h 0000_0000h
D[63:0] (ignored) Interrupt number expected from interrupt controller
on D[7:0]
Chapter 6 Bus Cycles 169
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 68. Interrupt Acknowledge Operation
Interrupt Acknowledge Cycles
Interrupt Number
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
LOCK#
INTR
D[63:0]
KEN#
BRDY#
170 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
6.6 Special Bus Cycles
The AMD-K6-2E processor drives special bus cycles that
include the following:
■Stop grant
■Flush acknowledge
■Cache writeback invalidation
■Halt
■Cache invalidation
■Shutdown
During all special cycles, D/C# = 0, M/IO# = 0, and W/R# = 1.
BE[7:0]# and A[31:3] are driven to differentiate among the
special cycles, as shown in Table 28. (See also Table 23 on
page 132.)
Note that the system logic must return BRDY# in response to all
processor special cycles.
Basic Special Bus
Cycle Figure 69 on page 171 shows a basic special bus cycle.
The processor drives D/C# = 0, M/IO# = 0, and W/R# = 1 off the
same clock edge that it asserts ADS#.
In this example, BE[7:0]# = FBh and A[31:3] = 0000_0000h,
which indicates that the special cycle is a halt special cycle (See
Table 28). A halt special cycle is generated after the processor
executes the HLT instruction.
Table 28. Encodings for Special Bus Cycles
BE[7:0]# A[4:3]1
Notes:
1. A[31:5] = 0
Special Bus Cycle Cause
FBh 10b Stop Grant STPCLK# sampled asserted
EFh 00b Flush Acknowledge FLUSH# sampled asserted
F7h 00b Writeback WBINVD instruction
FBh 00b Halt HLT instruction
FDh 00b Flush INVD,WBINVD instruction
FEh 00b Shutdown Triple fault
Chapter 6 Bus Cycles 171
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
If the processor samples FLUSH# asserted, it writes back any
data cache lines that are in the Modified state and invalidates
all lines in the instruction and data cache. The processor then
drives a flush acknowledge special cycle.
If the processor executes a WBINVD instruction, it drives a
writeback special cycle after the processor completes
invalidating and writing back the cache lines.
Figure 69. Basic Special Bus Cycle (Halt Cycle)
Halt Cycle
A[4:3] = 00b
FBh
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
BRDY#
172 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Shutdown Cycle In Figure 70, a shutdown (triple fault) occurs in the first half of
the waveform, and a shutdown special cycle follows in the
second half. The processor enters shutdown when an interrupt
or exception occurs during the handling of a double fault (INT
8), which amounts to a triple fault. When the processor
encounters a triple fault, it stops its activity on the bus and
generates the shutdown special bus cycle (BE[7:0]# = FEh).
The system logic must assert NMI, INIT, RESET, or SMI# to get
the processor out of the Shutdown state.
Figure 70. Shutdown Cycle
Shutdown Occurs
(Triple Fault) Shutdown Special Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
LOCK#
M/IO#
D/C#
W/R#
D[63:0]
KEN#
BRDY#
A[4:3] = 00b
FEh
Chapter 6 Bus Cycles 173
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Stop Grant and Stop
Clock States Figure 71 on page 174 and Figure 72 on page 175 show the
processor transition from normal execution to the Stop Grant
state, then to the Stop Clock state, back to the Stop Grant state,
and finally back to normal execution. The series of transitions
begins when the processor samples STPCLK# asserted. On
recognizing a STPCLK# interrupt at the next instruction
retirement boundary, the processor performs the following
actions, in the order shown:
1. Its instruction pipelines are flushed.
2. All pending and in-progress bus cycles are completed.
3. The STPCLK# assertion is acknowledged by executing a
Stop Grant special bus cycle.
4. Its internal clock is stopped after BRDY# of the Stop Grant
special bus cycle is sampled asserted and after EWBE# is
sampled asserted (if EWBE# is masked off, then entry into
the Stop Grant state is not affected by EWBE#).
5. The Stop Clock state is entered if the system logic stops the
bus clock CLK (optional).
STPCLK# is sampled as a level-sensitive input on every clock
edge but is not recognized until the next instruction boundary.
The system logic drives the signal either synchronously or
asynchronously. If it is asserted asynchronously, it must be
asserted for a minimum pulse width of two clocks. STPCLK#
must remain asserted until recognized, which is indicated by
the completion of the Stop Grant special cycle.
174 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 71. Stop Grant and Stop Clock Modes, Part 1
STPCLK# Sampled Asserted Stop Grant Special Cycle
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
CACHE#
STPCLK#
D[63:0]
KEN#
BRDY#
Stop Clock
A[4:3] = 10b
FBh
Chapter 6 Bus Cycles 175
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 72. Stop Grant and Stop Clock Modes, Part 2
Stop Grant State
(Re-entered after PLL stabilization)
STPCLK# Sampled Negated NormalStop Clock
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
CACHE#
STPCLK#
D[63:0]
KEN#
BRDY#
176 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
INIT-Initiated
Transition from
Protected Mode to
Real Mode
INIT is typically asserted in response to a BIOS interrupt that
writes to an I/O port. This interrupt is often in response to a
Ctrl-Alt-Del keyboard input. The BIOS writes to a port (similar
to port 64h in the keyboard controller) that asserts INIT. INIT is
also used to support 80286 software that must return to real
mode after accessing extended memory in protected mode.
The assertion of INIT causes the processor to empty its
pipelines, initialize most of its internal state, and branch to
address FFFF_FFF0h—the same instruction execution starting
point used after RESET. Unlike RESET, the processor
preserves the contents of its caches, the Floating-Point state,
the MMX state, Model-Specific Registers (MSRs), the CD and
NW bits of the CR0 register, the time stamp counter, and other
specific internal resources.
Figure 73 on page 177 shows an example in which the operating
system writes to an I/O port, causing the system logic to assert
INIT. The sampling of INIT asserted starts an extended
microcode sequence that terminates with a code fetch from
FFFF_FFF0h, the reset location. INIT is sampled on every clock
edge but is not recognized until the next instruction boundary.
During an I/O write cycle, it must be sampled asserted a
minimum of three clock edges before BRDY# is sampled
asserted if it is to be recognized on the boundary between the
I/O write instruction and the following instruction. If INIT is
asserted synchronously, it can be asserted for a minimum of one
clock. If it is asserted asynchronously, it must have been
negated for a minimum of two clocks, followed by an assertion
of a minimum of two clocks.
Chapter 6 Bus Cycles 177
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 73. INIT-Initiated Transition from Protected Mode to Real Mode
Code Fetch
FFFF_FFF0h
INIT Sampled Asserted
CLK
A[31:3]
BE[7:0]#
ADS#
M/IO#
D/C#
W/R#
D[63:0]
KEN#
BRDY#
INIT
178 Bus Cycles Chapter 6
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 7 Power-On Configuration and Initialization 179
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
7 Power-On Configuration and Initialization
On power-on, the system logic must reset the AMD-K6-2E
processor by asserting the RESET signal. When the processor
samples RESET asserted, it immediately flushes and initializes
all internal resources and its internal state, including its
pipelines and caches, the floating-point state, the MMX and
3DNow! states, and all registers. Then, the processor jumps to
address FFFF_FFF0h to start instruction execution.
7.1 Signals Sampled During the Falling Transition of RESET
FLUSH# FLUSH# is sampled on the falling transition of RESET to
determine if the processor begins normal instruction execution
or enters three-state test mode.
■If FLUSH# is High during the falling transition of RESET,
the processor unconditionally runs its Built-In Self Test
(BIST), performs the normal reset functions, then jumps to
address FFFF_FFF0h to start instruction execution. (See
“Built-In Self-Test (BIST)” on page 227 for more details.)
■If FLUSH# is Low during the falling transition of RESET,
the processor enters three-state test mode. (See
“Three-State Test Mode” on page 228 and “FLUSH# (Cache
Flush)” on page 104 for more details.)
BF[2:0] The internal operating frequency of the processor is
determined by the state of the bus frequency signals BF[2:0]
when they are sampled during the falling transition of RESET.
The frequency of the CLK input signal is multiplied internally
by a ratio defined by BF[2:0]. (See “BF[2:0] (Bus Frequency)”
on page 93 for the processor-clock to bus-clock ratios.)
180 Power-On Configuration and Initialization Chapter 7
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
7.2 RESET Requirements
During the initial power-on reset of the processor, RESET must
remain asserted for a minimum of 1.0 ms after CLK and VCC
reach specification. (See “CLK Switching Characteristics” on
page 267 for clock specifications. See “Electrical Data” on page
253 for VCC specifications.)
During a warm reset while CLK and VCC are within
specification, RESET must remain asserted for a minimum of
15 clocks prior to its negation.
7.3 State of Processor After RESET
Output Signals Table 29 shows the state of all processor outputs and
bidirectional signals immediately after RESET is sampled
asserted.
Registers Table 30 on page 181 shows the state of all architecture
registers and Model-Specific Registers (MSRs) after the
processor has completed its initialization due to the recognition
of the assertion of RESET.
Table 29. Output Signal State After RESET
Signal State Signal State
A[31:3], AP Floating LOCK# High
ADS#, ADSC# High M/IO# Low
APCHK# High PCD Low
BE[7:0]# Floating PCHK# High
BREQ Low PWT Low
CACHE# High SCYC Low
D/C# Low SMIACT# High
D[63:0], DP[7:0] Floating TDO Floating
FERR# High VCC2DET Low
HIT# High VCC2H/L# Low
HITM# High W/R# Low
HLDA Low ––
Chapter 7 Power-On Configuration and Initialization 181
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 30. Register State After RESET
Register State (hex)
GDTR base:0000_0000h limit:0FFFFh
IDTR base:0000_0000h limit:0FFFFh
TR 0000h
LDTR 0000h
EIP FFFF_FFF0h
EFLAGS 0000_0002h
EAX10000_0000h
EBX 0000_0000h
ECX 0000_0000h
EDX20000_058Xh
ESI 0000_0000h
EDI 0000_0000h
EBP 0000_0000h
ESP 0000_0000h
CS F000h
SS 0000h
DS 0000h
ES 0000h
FS 0000h
GS 0000h
FPU Stack R7–R030000_0000_0000_0000_0000h
FPU Control Word30040h
FPU Status Word30000h
FPU Tag Word35555h
FPU Instruction Pointer30000_0000_0000h
FPU Data Pointer30000_0000_0000h
FPU Opcode Register3000_0000_0000b
CR046000_0010h
CR2 0000_0000h
CR3 0000_0000h
CR4 0000_0000h
DR7 0000_0400h
DR6 FFFF_0FF0h
DR3 0000_0000h
182 Power-On Configuration and Initialization Chapter 7
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
DR2 0000_0000h
DR1 0000_0000h
DR0 0000_0000h
MCAR30000_0000_0000_0000h
MCTR30000_0000_0000_0000h
TR1230000_0000_0000_0000h
TSC30000_0000_0000_0000h
EFER30000_0000_0000_0002h (Model 8/[F:8])
STAR30000_0000_0000_0000h
WHCR30000_0000_0000_0000h
UWCCR30000_0000_0000_0000h
PSOR3,5 0000_0000_0000_01SBh
PFIR30000_0000_0000_0000h
Notes:
1. The contents of EAX indicate if BIST was successful. If EAX = 0000_0000h, BIST was successful.
If EAX is non-zero, BIST failed.
2. EDX contains the AMD-K6-2E processor signature, where X indicates the processor Stepping
ID.
3. The contents of these registers are preserved following the recognition of INIT.
4. The CD and NW bits of CR0 are preserved following the recognition of INIT.
5. “S” represents the Stepping. “B” represents PSOR[3:0], where PSOR[3] equals 0, and
PSOR[2:0] is equal to the value of the BF[2:0] signals sampled during the falling transition of
RESET.
Table 30. Register State After RESET (continued)
Register State (hex)
Chapter 7 Power-On Configuration and Initialization 183
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
7.4 State of Processor After INIT
The recognition of the assertion of INIT causes the processor to
empty its pipelines, to initialize most of its internal state, and to
branch to address FFFF_FFF0h—the same instruction
execution starting point used after RESET.
Unlike RESET, the processor preserves the contents of its
caches, the floating-point state, the MMX and 3DNow! states,
MSRs, and the CD and NW bits of the CR0 register.
The edge-sensitive interrupts FLUSH# and SMI# are sampled
and preserved during the INIT process and are handled
accordingly after the initialization is complete. However, the
processor resets any pending NMI interrupt upon sampling
INIT asserted.
INIT can be used as an accelerator for 80286 code that requires
a reset to exit from protected mode back to real mode.
184 Power-On Configuration and Initialization Chapter 7
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 8 Cache Organization 185
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
8 Cache Organization
The following sections describe the basic architecture and
resources of the AMD-K6-2E processor internal caches.
The performance of the AMD-K6-2E processor is enhanced by a
writeback level-one (L1) cache. The cache is organized as a
separate 32-Kbyte instruction cache and a 32-Kbyte data cache,
each with two-way set associativity (See Figure 74). The cache
line size is 32 bytes, and lines are prefetched from external
memory using an efficient, pipelined burst transaction.
As the instruction cache is filled, each instruction byte is
analyzed for instruction boundaries using predecode logic.
Predecoding annotates each instruction byte with information
that later enables the decoders to efficiently decode multiple
instructions simultaneously. Translation lookaside buffers
(TLB) are also used to translate linear addresses to physical
addresses. The instruction cache is associated with a 64-entry
TLB, while the data cache is associated with a 128-entry TLB.
Figure 74. Cache Organization
Processor
Core
System Bus
Interface Unit
128-Entry TLB
64-Entry TLB
State
Bit
Tag
RAM Way 0 Way 1 State
Bit
Tag
RAM
32-Kbyte Instruction Cache
32-Kbyte Data Cache
Predecode Instruction Cache
MESI
Bits
Tag
RAM Way 0 Way 1 MESI
Bits
Tag
RAM
186 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The processor cache design takes advantage of a sectored
organization (See Figure 75). Each sector consists of 64 bytes
configured as two 32-byte cache lines. The two cache lines of a
sector share a common tag but have separate MESI (modified,
exclusive, shared, invalid) bits that track the state of each cache
line.
Instruction Cache Line
Data Cache Line
Notes:
Instruction-cache lines have only two coherency states (valid or invalid) rather than the four MESI coherency states of data-cache lines.
Only two states are needed for the instruction cache because these lines are read-only.
Figure 75. Cache Sector Organization
8.1 MESI States in the Data Cache
The state of each line in the caches is tracked by the MESI bits.
The coherency of these states or MESI bits is maintained by
internal processor snoops and external inquiries by the system
logic. The following four states are defined for the data cache:
■Modified—This line has been modified and is different from
main memory.
■Exclusive—This line is not modified and is the same as
external memory. If this line is written to, it becomes
Modified.
■Shared—If a cache line is in the Shared state it means that
the same line can exist in more than one cache system.
■Invalid—The information in this line is not valid.
Tag
Address
Cache Line 0 Byte 31 Predecode Bits Byte 30 Predecode Bits ........ ........ Byte 0 Predecode Bits 1 MESI Bit
Cache Line 1 Byte 31 Predecode Bits Byte 30 Predecode Bits ........ ........ Byte 0 Predecode Bits 1 MESI Bit
Tag
Address
Cache Line 0 Byte 31 Byte 30 ........ ........ Byte 0 2 MESI Bits
Cache Line 1 Byte 31 Byte 30 ........ ........ Byte 0 2 MESI Bits
Chapter 8 Cache Organization 187
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
8.2 Predecode Bits
Decoding x86 instructions is particularly difficult because the
instructions vary in length, ranging from 1 to 15 bytes long.
Predecode logic supplies the predecode bits associated with
each instruction byte.
Predecode bits indicate the number of bytes to the start of the
next x86 instruction. The predecode bits are passed with the
instruction bytes to the decoders, where they assist with
parallel x86 instruction decoding. The predecode bits use
memory separate from the 32-Kbyte instruction cache. The
predecode bits are stored in an extended instruction cache
alongside each x86 instruction byte as shown in Figure 75 on
page 186.
8.3 Cache Operation
The operating modes for the caches are configured by software
using the Not Writethrough (NW) and Cache Disable (CD) bits
of control register 0 (CR0 bits 29 and 30 respectively). These
bits are used in all operating modes.
■When the CD and NW bits are both 0, the cache is fully
enabled. This is the standard operating mode for the cache.
If a read miss occurs when the processor reads from the
cache, a line fill (32-byte burst read) on the system bus
occurs in order to fetch the cache line. Write hits to the
cache are updated, while write misses and writes to shared
lines cause external memory updates. Refer to Table 34,
“Data Cache States for Read and Write Accesses,” on
page 198 for a summary of cache read and write cycles and
the effect of these operations on the cache MESI state.
Note: A write allocate operation can modify the behavior of write
misses to the cache. See “Write Allocate” on page 192.
■When the CD bit is 0 and the NW bit is 1, an invalid mode of
operation exists that causes a general protection fault to
occur.
■When the CD bit is 1 (disabled) and the NW bit is 0, the
cache fill mechanism is disabled but the contents of the
cache are still valid. The processor reads from the cache, and
if a read miss occurs, no line fills take place. Write hits to the
cache are updated, while write misses and writes to shared
188 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
lines cause external memory updates. If PWT is driven Low
and WB/WT# is sampled High, a write hit to a shared line
changes the cache-line state to Exclusive.
■When the CD and NW bits are both 1, the cache is fully
disabled. Even though the cache is disabled, the contents
are not necessarily invalid. The processor reads from the
cache and, if a read miss occurs, no line fills take place. If a
write hit occurs, the cache is updated but an external
memory update does not occur. If a cache line is in the
Exclusive state during a write hit, the cache-line state is
changed to Modified. Cache lines in the Shared state remain
in the Shared state after a write hit. Write misses access
external memory directly.
The operating system can control the cacheability of a page.
The paging mechanism is controlled by CR3, the Page Directory
Entry (PDE), and the Page Table Entry (PTE). Within CR3,
PDE, and PTE are Page Cache Disable (PCD) and Page
Writethrough (PWT) bits. The values of the PCD and PWT bits
used in Table 31 and Table 32 are taken from either the PTE or
PDE. For more information see the descriptions of PCD and
PWT on page 115 and page 117, respectively.
Tables 31 through 33 describe the logic that determines the
cacheability of a cycle and how that cacheability is affected by
the PCD bits, the PWT bits, the PG and CD bits of CR0,
writeback cycles, the Cache Inhibit (CI) bit of Test Register 12
(TR12), and unlocked memory reads.
Table 31 describes how the PWT signal is driven based on the
values of the PWT bits and the PG bit of CR0.
Table 31. PWT Signal Generation
PWT Bit1
Notes:
1. PWT is taken from PTE or PDE.
PG Bit of CR0 PWT Signal
11High
01Low
10Low
00Low
Chapter 8 Cache Organization 189
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 32 describes how the PCD signal is driven based on the
values of the CD bit of CR0, the PCD bits, and the PG bit of
CR0.
Table 33 describes how the CACHE# signal is driven based on
the cycle type, the CI bit of TR12, the PCD signal, and the
UWCCR model-specific register.
Table 32. PCD Signal Generation
CD Bit of CR0 PCD Bit1
Notes:
1. PCD is taken from PTE or PDE.
PG Bit of CR0 PCD Signal
1Don’t care Don’t care High
011High
001Low
010Low
000Low
Table 33. CACHE# Signal Generation
Cycle Type CI Bit of TR12 PCD Signal Access Within
WC/UC Range1
Notes:
1. WC and UC refer to Write-Combining and Uncacheable Memory Ranges as defined in the UWCCR.
CACHE#
Writebacks Don’t care Don’t care Don’t care Low
Unlocked Reads 0 0 0 Low
Locked Reads Don’t care Don’t care Don’t care High
Single Writes Don’t care Don’t care Don’t care High
Any Cycle Except Writebacks 1 Don’t care Don’t care High
Any Cycle Except Writebacks Don’t care 1 Don’t care High
Any Cycle Except Writebacks Don’t care Don’t care 1 High
190 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Cache-Related Signals Complete descriptions of the signals that control cacheability
and cache coherency are given on the following pages:
■CACHE#—page 97
■EADS#—page 101
■FLUSH#—page 104
■HIT#—page 105
■HITM#—page 105
■INV—page 110
■KEN#—page 111
■PCD—page 115
■PWT—page 117
■WB/WT#—page 129
8.4 Cache Disabling and Flushing
To completely disable all cache accesses, the CD bit must be set
to 1 and the cache must be completely flushed. There are three
different methods for flushing the cache. The first method
relies on the system logic, and the other two rely on software.
■For the system logic to flush the cache, the processor must
sample FLUSH# asserted. In this method, the processor
writes back any data cache lines that are in the Modified
state, invalidates all lines in the instruction and data caches,
and then executes a flush acknowledge special cycle (See
Table 23 on page 132).
■The second method relies on software to execute the
WBINVD instruction which causes all modified lines to first
be written back to memory, then marks all cache lines as
invalid. Alternatively, if writing modified lines back to
memory is not necessary, the INVD instruction can be used
to invalidate all cache lines.
■The third method is to make use of the Page Flush/Invalidate
Register (PFIR), which allows cache invalidation and
optional flushing of a specific 4-Kbyte page from the linear
address space (see “Page Flush/Invalidate Register (PFIR)”
on page 200). Unlike the previous two methods of flushing
the cache, this particular method requires the software to be
aware of which specific pages must be flushed and
invalidated.
Chapter 8 Cache Organization 191
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
8.5 Cache-Line Fills
The processor performs a cache-line fill for any area of system
memory defined as cacheable. If an area of system memory is
not explicitly defined as uncacheable by the software or system
logic, or implicitly treated as uncacheable by the processor,
then the memory access is assumed to be cacheable.
Software can prevent caching of certain pages by setting the
PCD bit in the page directory entry (PDE) or page table entry
(PTE). Additionally, software can define regions of memory as
uncacheable or write combinable by programming the MTRRs
in the UC/WC cacheability control register (UWCCR) (see
“Memory Type Range Registers” on page 207). Write-
combinable memory is defined as uncacheable.
The system logic also has control of the cacheability of bus
cycles. If system logic determines the address is not cacheable,
system logic negates the KEN# signal when asserting the first
BRDY# or NA# of a cycle.
The processor does not cache certain memory accesses, such as
locked operations. In addition, the processor does not cache
PDE or PTE memory reads in the L1 cache (referred to as page
table walks).
When the processor needs to read memory, the processor drives
a read cycle onto the bus. If the cycle is cacheable, the
processor asserts CACHE#. If the cycle is not cacheable, a
non-burst, single-transfer read takes place. The processor waits
for the system logic to return the data and assert a single
BRDY# (See Figure 52 on page 139). If the cycle is cacheable,
the processor executes a 32-byte burst read cycle. The processor
expects a total of four BRDY# signals for a burst read cycle to
take place (See Figure 54 on page 143).
Cache-line fills initiate 32-byte burst read cycles from memory
on the system bus for the instruction cache and the data cache.
If a data-cache line being filled replaces a modified line, the
modified contents of the line are copied to a 32-byte writeback
(copyback) buffer in the bus interface unit while the new line is
being read.
192 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
8.6 Cache-Line Replacements
As programs execute and task switches occur, some cache lines
eventually require replacement.
Instruction cache lines are replaced using a Least Recently
Used (LRU) algorithm. If line replacement is required, lines are
replaced when read cache misses occur.
The data cache uses a slightly different approach to line
replacement. If a miss occurs, and a replacement is required,
lines are replaced by using a Least Recently Allocated (LRA)
algorithm.
Two forms of cache misses and associated cache fills can take
place—a tag-miss cache fill and a tag-hit cache fill.
■In the case of a tag-miss cache fill, the miss is due to a tag
mismatch, in which case the required cache line is filled
from external memory, and the cache line within the sector
that was not required is marked as invalid.
■In the case of a tag-hit cache fill, the address matches the
tag, but the requested cache line is marked as invalid. The
required cache line is filled from external memory, and the
cache line within the sector that is not required remains in
the same cache state.
8.7 Write Allocate
Write allocate, if enabled, occurs when the processor has a
pending memory write cycle to a cacheable line and the line
does not currently reside in the data cache. In this case, the
processor performs a 32-byte burst read cycle to fetch the
data-cache line addressed by the pending write cycle. The data
associated with the pending write cycle is merged with the
recently-allocated data-cache line and stored in the processor’s
data cache. The final MESI state of the cache line depends on
the state of the WB/WT# and PWT signals during the burst read
cycle and the subsequent data cache write hit (See Table 34 on
page 198 to determine the cache-line states and the access
types following a cache read miss and cache write hit).
If a data-cache line fetch from memory is attempted because
the write allocate misses the data cache, and KEN# is sampled
Chapter 8 Cache Organization 193
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
negated, the processor does not perform an allocation. In this
case, the pending write cycle is executed as a single write cycle
on the system bus.
During write allocates, a 32-byte burst read cycle is executed in
place of a non-burst write cycle. While the burst read cycle
generally takes longer to execute than the write cycle,
performance gains are realized on subsequent write cycle hits
to the write-allocated cache line. Due to the nature of software,
memory accesses tend to occur in proximity of each other
(principle of locality). The likelihood of additional write hits to
the write-allocated cache line is high.
The following is a description of three mechanisms by which the
AMD-K6-2E processor performs write allocations. A write
allocate is performed when any one or more of these
mechanisms indicates that a pending write is to a cacheable
area of memory.
Write to a Cacheable
Page Every time the processor performs a cache line fill, the address
of the page in which the cache line resides is saved in the
Cacheability Control Register (CCR). The page address of
subsequent write cycles is compared with the page address
stored in the CCR. If the two addresses are equal, then the
processor performs a write allocate because the page has
already been determined to be cacheable.
When the processor performs a cache line fill from a different
page than the address saved in the CCR, the CCR is updated
with the new page address.
Write to a Sector If the address of a pending write cycle matches the tag address
of a valid cache sector, but the addressed cache line within the
sector is marked invalid (a sector hit but a cache line miss),
then the processor performs a write allocate. The pending write
cycle is determined to be cacheable because the sector hit
indicates the presence of at least one valid cache line in the
sector. The two cache lines within a sector are guaranteed by
design to be within the same page.
Write Allocate Limit The AMD-K6-2E processor uses two mechanisms that are
programmable within the Write Handling Control register
(WHCR) to enable write allocations for write cycles that
address a definable or special 1-Mbyte memory area.
194 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Write Handling Control Register (WHCR) . The WHCR register
contains two fields—the Write Allocate Enable Limit
(WAELIM) field, and the Write Allocate Enable 15-to-16-Mbyte
(WAE15M) bit (see Figure 76).
Figure 76. Write Handling Control Register (WHCR)
Write Allocate Enable Limit Field. The WAELIM field is 10 bits wide.
This field, multiplied by 4 Mbytes, defines an upper memory
limit. Any pending write cycle that addresses memory below
this limit causes the processor to perform a write allocate
(assuming the address is not within a range where write
allocates are disallowed). Write allocate is disabled for memory
accesses at and above this limit unless the processor determines
a pending write cycle is cacheable by means of one of the other
write allocate mechanisms—“Write to a Cacheable Page” and
“Write to a Sector.” The maximum value of this limit is ((210–1)
· 4 Mbytes) = 4092 Mbytes. When all the bits in this field are 0,
all memory is above this limit and the write allocate mechanism
is disabled (even if all bits in the WAELIM field are 0, write
allocates can still occur due to the “Write to a Cacheable Page”
and “Write to a Sector” mechanisms).
Write Allocate Enable 15-to-16-Mbyte Bit. The Write Allocate Enable
15-to-16-Mbyte (WAE15M) bit is used to enable write
allocations for the memory write cycles that address the 1
Mbyte of memory between 15 Mbytes and 16 Mbytes. This bit
must be set to 1 to allow write allocate in this memory area. This
bit is provided to account for a small number of uncommon
memory-mapped I/O adapters that use this particular memory
address space. If the system contains one of these peripherals,
the bit should be written to 0 (even if the WAE15M bit is 0,
1522 063
WAELIM
16
Note: Hardware RESET initializes this MSR to all zeros.
W
A
E
1
5
M
Symbol Description Bits
WAELIM Write Allocate Enable Limit 31-22
WAE15M Write Allocate Enable 15-to-16-Mbyte 16
17
213132
Reserved
Chapter 8 Cache Organization 195
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
write allocates can still occur between 15 Mbytes and 16
Mbytes due to the “Write to a Cacheable Page” and “Write to a
Sector” mechanisms). The WAE15M bit is ignored if the value
in the WAELIM field is less than 16 Mbytes.
By definition, a write allocate is never performed in the
memory area between 640 Kbytes and 1 Mbyte unless the
processor determines a pending write cycle is cacheable by
means of one of the other write allocate mechanisms—“Write
to a Cacheable Page” and “Write to a Sector”. It is not
considered safe to perform write allocations between 640
Kbytes and 1 Mbyte (000A_0000h to 000F_FFFFh) because it is
considered a noncacheable region of memory.
If a memory region is defined as write-combinable or
uncacheable by a MTRR, write allocates are not performed in
that region.
Write Allocate Logic
Mechanisms and
Conditions
Figure 77 shows the logic flow for all the mechanisms involved
with write allocate for memory bus cycles. The left side of the
diagram (the text) describes the conditions that need to be true
for the value of that line to be a 1. Items 1–4 of the diagram are
related to general cache operation and items 5–10 are related to
the write allocate mechanisms.
For more information about write allocate, see the
Implementation of Write Allocate in the K86™ Processors
Application Note, order #21326.
Figure 77. Write Allocate Logic Mechanisms and Conditions
1) CD Bit of CR0 Perform
Write Allocate
3) CI Bit of TR12
2) PCD Signal
5) Write to Cacheable Page (CCR)
6) Write to a Sector
8) Between 640 Kbytes and 1 Mbyte
7) Less Than Limit (WAELIM)
9) Between 15–16 Mbytes
10) Write Allocate Enable 15–16 Mbyte (WAE15M)
4) UC or WC
196 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The following list corresponds to the items in Figure 77 on page
195:
1. CD Bit of CR0—When the Cache Disable (CD) bit within
control register 0 (CR0) is 1, the cache fill mechanism for
both reads and writes is disabled, and write allocate does
not occur.
2. PCD Signal—When the PCD (page cache disable) signal is
driven High, caching for that page is disabled even if KEN#
is sampled asserted, and write allocate does not occur.
3. CI Bit of TR12—When the cache inhibit bit of test register 12
is 1, L1 cache fills are disabled, and write allocate does not
occur.
4. UC or WC—If a pending write cycle addresses a region of
memory defined as write combinable or uncacheable by an
MTRR, write allocates are not performed in that region.
5. Write to a Cacheable Page (CCR)—A write allocate is
performed if the processor knows that a page is cacheable.
The CCR is used to store the page address of the last cache
fill for a read miss. See “Write to a Cacheable Page” on page
193 for a detailed description of this condition.
6. Write to a Sector—A write allocate is performed if the
address of a pending write cycle matches the tag address of a
valid cache sector, but the addressed cache line within the
sector is invalid. See “Write to a Sector” on page 193 for a
detailed description of this condition.
7. Less Than Limit (WAELIM)—The write allocate limit
mechanism determines if the memory area being addressed
is less than the limit set in the WAELIM field of WHCR. If
the address is less than the limit, write allocate for that
memory address is performed as long as conditions 8
through 10 do not prevent write allocate (even if conditions
8 and 10 attempt to prevent write allocate, condition 5 or 6
allows write allocates to occur).
8. Between 640 Kbytes and 1 Mbyte—Write allocate is not
performed in the memory area between 640 Kbytes and 1
Mbyte. It is not considered safe to perform write allocations
between 640 Kbytes and 1 Mbyte (000A_0000h to
000F_FFFFh) because this area of memory is considered a
noncacheable region of memory (even if condition 8
attempts to prevent write allocate, condition 5 or 6 allows
write allocate to occur).
Chapter 8 Cache Organization 197
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
9. Between 15–16 Mbytes—If the address of a pending write
cycle is in the 1 Mbyte of memory between 15 Mbytes and 16
Mbytes, and the WAE15M bit is 1, write allocate for this
cycle is enabled.
10.Write Allocate Enable 15–16 Mbytes (WAE15M)—This
condition is associated with the Write Allocate Limit
mechanism and affects write allocate only if the limit
specified by the WAELIM field is greater than or equal to
16 Mbytes. If the memory address is between 15 Mbytes and
16 Mbytes, and the WAE15M bit in the WHCR is 0, write
allocate for this cycle is disabled (even if condition 10
attempts to prevent write allocate, condition 5 or 6 allows
write allocate to occur).
8.8 Prefetching
Hardware
Prefetching The AMD-K6-2E processor conditionally performs cache
prefetching which results in the filling of the required cache
line first, and a prefetch of the second cache line making up the
other half of the sector. From the perspective of the external
bus, the two cache-line fills typically appear as two 32-byte
burst read cycles occurring back-to-back or, if allowed, as
pipelined cycles. The burst read cycles do not occur
back-to-back (wait states occur) if the processor is not ready to
start a new cycle, if higher priority data read or write requests
exist, or if NA# (next address) was sampled negated. Wait states
can also exist between burst cycles if the processor samples
AHOLD or BOFF# asserted.
Software Prefetching The 3DNow! technology includes an instruction called
PREFETCH that allows a cache line to be prefetched into the
L1 data cache. Unlike prefetching under hardware control,
software prefetching only fetches the cache line specified by
the operand of the PREFETCH instruction, and does not
attempt to fetch the other cache line in the sector. The
PREFETCH instruction format is defined in Table 15,
“3DNow!™ Instructions,” on page 81. For more detailed
information, see the 3DNow!™ Technology Manual, order#
21928.
198 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
8.9 Cache States
Table 34 shows all the possible cache-line states before and
after program-generated accesses to individual cache lines. The
table includes the correspondence between MESI states and
Writethrough or Writeback states for lines in the data cache.
Table 34. Data Cache States for Read and Write Accesses
Type Cache State Before
Access Access Type1
Notes:
1. Single read, single write, cache update, and writethrough = 1 to 8 bytes. Line fill = 32-byte burst read.
Cache State After Access
MESI State2
2. The final MESI state assumes that the state of the WB/WT# signal remains the same for all accesses to a particular cache line.
Writeback/
Writethrough State
Cache
Read
Read miss
Invalid Single read from bus Invalid Not applicable or none
Invalid Burst read from bus, fill
cache3
3. If CACHE# is driven Low and KEN# is sampled asserted.
Shared or
exclusive4
4. If PWT is driven Low and WB/WT# is sampled High, the line is cached in the exclusive (writeback) state. If PWT is driven High or
WB/WT# is sampled Low, the line is cached in the shared (writethrough) state.
Writethrough or writeback4
Read hit
Exclusive Not applicable or none Exclusive Writeback
Modified Not applicable or none Modified Writeback
Shared Not applicable or none Shared Writethrough
Cache
Write
Write miss
Invalid Single write to bus5
5. Assumes the write allocate conditions as specified in “Write Allocate” on page 192 are not met.
Invalid Not applicable or none
Invalid Burst read from bus, fill
cache, write to cache6
6. Assumes the write allocate conditions as specified in “Write Allocate” on page 192 are met.
Modified7
7. Assumes PWT is driven Low and WB/WT# is sampled High.
Not applicable or none
Invalid
Burst read from bus, fill
cache, write to cache,
single write to bus6Shared8
8. Assumes PWT is driven High or WB/WT # is sampled Low.
Not applicable or none
Write hit
Exclusive or modified Write to cache Modified Writeback
Shared Write to cache, single write
to bus
Shared or
exclusive4Writethrough or writeback4
Chapter 8 Cache Organization 199
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
8.10 Cache Coherency
Different methods exist to maintain coherency between the
system memory and cache memories. Inquire cycles, internal
snoops, FLUSH#, WBINVD, INVD, and line replacements all
prevent inconsistencies between memories.
Inquire Cycles Inquire cycles are bus cycles initiated by system logic. These
inquiries ensure coherency between the caches and main
memory. In systems with multiple caching masters, system logic
maintains cache coherency by driving inquire cycles to the
processor. System logic initiates inquire cycles by asserting
AHOLD, BOFF#, or HOLD to obtain control of the address bus
and then driving EADS#, INV (optional), and an inquire
address (A[31:5]).
This type of bus cycle causes the processor to compare the tags
for both its instruction and data caches with the inquire
address.
■If there is a hit to a shared or exclusive line in the data cache
or a valid line in the instruction cache, the processor asserts
HIT#.
■If the compare hits a modified line in the data cache, the
processor asserts HIT# and HITM#. If HITM# is asserted, the
processor writes the modified line back to memory.
■If INV was sampled asserted with EADS#, a hit invalidates
the line.
■If INV was sampled negated with EADS#, a hit leaves the
line in the Shared state or transitions it from the Exclusive
or Modified to Shared state.
Table 35 on page 202 shows the effects of inquire cycles—
performed with INV equal to 0 (non-validating) and INV equal
to 1 (invalidating) snoops and invalidations.
Internal Snooping Internal snooping is initiated by the processor (rather than
system logic) during certain cache accesses. It is used to
maintain coherency between the L1 instruction and data
caches.
The processor automatically snoops its instruction cache during
read or write misses to its data cache, and it snoops its data
200 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
cache during read misses to its instruction cache. Table 35
summarizes the actions taken during this internal snooping.
If an internal snoop hits its target, the processor does the
following:
■Data Cache Snoop During an Instruction-Cache Read
Miss—If modified, the line in the data cache is written back
on the system bus to external memory. Regardless of its
state, the data-cache line is invalidated and the instruction
cache performs a burst read cycle from external memory.
■Instruction Cache Snoop During a Data-Cache Miss—The
line in the instruction cache is marked invalid, and the
data-cache read or write is performed from memory.
FLUSH# In response to sampling FLUSH# asserted, the processor writes
back any data cache lines that are in the Modified state and
then marks all lines in the instruction and data caches as
invalid.
Page Flush/Invalidate
Register (PFIR) The AMD-K6-2E processor contains the page flush/invalidate
register (PFIR) (see Figure 78) that allows cache invalidation
and optional flushing of a specific 4-Kbyte page from the linear
address space. When the PFIR is written to (using the WRMSR
instruction), the invalidation and the flushing (optional) begin.
The total amount of cache in the AMD-K6-2E processor is 64
Kbytes. Using this register can result in a much lower cycle
count for flushing particular pages versus flushing the entire
cache.
Figure 78. Page Flush/Invalidate Register (PFIR)—MSR C000_0088h
LINPAGE
10
63
F
/
I
Symbol Description Bit
LINPAGE 20-bit Linear Page Address 31-12
PF Page Fault Occurred 8
F/I Flush/Invalidate Command 0
1131 1232
P
F
987
Reserved
Chapter 8 Cache Organization 201
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
LINPAGE Field. This 20-bit field must be written with bits 31:12 of
the linear address of the 4-Kbyte page that is to be invalidated
and optionally flushed from the L1 cache.
PF Bit. If an attempt to invalidate or flush a page results in a
page fault, the processor sets the PF bit to 1, and the invalidate
or flush operation is not performed (even though invalidate
operations do not normally generate page faults). In this case,
an actual page fault exception is not generated. If the PF bit
equals 0 after an invalidate or flush operation, then the
operation executed successfully. The PF bit must be read after
every write to the PFIR register to determine if the invalidate
or flush operation executed successfully.
F/I Bit. This bit is used to control the type of action that occurs to
the specified linear page. If a 0 is written to this bit, the
operation is a flush, in which case all cache lines in the
modified state within the specified page are written back to
memory, after which the entire page is invalidated. If a 1 is
written to this bit, the operation is an invalidation, in which
case the entire page is invalidated without the occurrence of
any writebacks.
WBINVD and INVD
Instructions These x86 instructions cause all cache lines to be marked as
invalid. WBINVD writes back modified lines before marking all
cache lines invalid. INVD does not write back modified lines.
Cache-Line
Replacement Replacing lines in the instruction or data cache, according to
the line replacement algorithms described in “Cache-Line
Fills” on page 191, ensures coherency between external
memory and the caches.
Table 35 on page 202 shows all possible cache-line states before
and after various cache-related operations.
202 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 35. Cache States for Inquire Cycles, Snoops, Flushes, and Invalidation
Type of Operation Cache State
Before Operation Memory Access
Cache State After Operation
MESI State Writeback/
Writethrough State
Inquire Cycle
Shared or exclusive Not applicable or none INV=0 Shared Writethrough
INV=1 Invalid Invalid
Modified Writeback to bus INV=0 Shared Writethrough
INV=1 Invalid Invalid
Internal Snoop Shared or exclusive Not applicable or none Invalid Invalid
Modified Writeback to bus
FLUSH# Signal
Shared or
exclusive Not applicable or none Invalid Invalid
Modified Writeback to bus
PFIR
(F/I = 0)
Shared or exclusive Not applicable or none Invalid Invalid
Modified Writeback to bus
PFIR
(F/I = 1) Not applicable or none Not applicable or none Invalid Invalid
WBINVD Instruction Shared or exclusive Not applicable or none Invalid Invalid
Modified Writeback to bus
INVD Instruction Not applicable or none Not applicable or none Invalid Invalid
Notes:
All writebacks are 32-byte burst write cycles.
Chapter 8 Cache Organization 203
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Cache Snooping Table 36 shows the conditions under which snooping occurs in
the AMD-K6-2E processor and the resources that are snooped.
Table 36. Snoop Action
Type of Event Type of Access Snooping Action
Instruction Cache Data Cache
Inquire Cycle System Logic Yes1
Notes:
1. The processor’s response to an inquire cycle depends on the state of the INV input signal and the state of the cache line as
follows:
For the instruction cache, if INV is sampled negated, the line remains invalid or valid, but if INV is sampled asserted, the line is
invalidated.
For the data cache, if INV is sampled negated, valid lines remain in or transition to the Shared state, a modified data cache line
is written back before the line is marked shared (with HITM# asserted), and invalid lines remain invalid. For the data cache, if
INV is sampled asserted, the line is marked invalid. Modified lines are written back before invalidation.
Yes1
Internal Snoop
Instruction Cache
Read
Miss Not applicable Yes2
2. If an internal snoop hits a modified line in the data cache, the line is written back and invalidated. Then the instruction cache
performs a burst read from memory.
Read
Hit Not applicable No
Data Cache
Read
Miss Yes3
3. If an internal snoop hits a line in the instruction cache, the instruction cache line is invalidated and the data-cache read or write
is performed from memory.
Not applicable
Read
Hit No Not applicable
Write
Miss Yes3Not applicable
Write
Hit No Not applicable
204 Cache Organization Chapter 8
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
8.11 Writethrough and Writeback Coherency States
The terms writethrough and writeback apply to two related
concepts in a read-write cache like the AMD-K6-2E processor’s
L1 data cache. The following conditions apply to both the
writethrough and writeback modes:
■Memory Writes—A relationship exists between external
memory writes and their concurrence with cache updates:
•An external memory write that occurs concurrently with
a cache update to the same location is a writethrough.
Writethroughs are driven as single cycles on the bus.
•An external memory write that occurs after the processor
has modified a cache line is a writeback. Writebacks are
driven as burst cycles on the bus.
■Coherency State—A relationship exists between MESI
coherency states and writethrough-writeback coherency
states of lines in the cache as follows:
•Shared and invalid MESI lines are in writethrough state.
•Modified and exclusive MESI lines are in writeback state.
8.12 A20M# Masking of Cache Accesses
Although the processor samples A20M# as a level-sensitive
input on every clock edge, it should only be asserted in real
mode. The processor applies the A20M# masking to its tags,
through which all programs access the caches. Therefore,
assertion of A20M# affects all addresses (cache and external
memory), including the following:
■Cache-line fills (caused by read misses or write allocates)
■Cache writethroughs (caused by write misses or write hits to
lines in the Shared state)
However, A20M# does not mask writebacks or invalidations
caused by the following actions:
■Internal snoops
■Inquire cycles
■The FLUSH# signal
■The WBINVD instruction
■Writing to the page flush/invalidate register (PFIR)
Chapter 9 Write Merge Buffer 205
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
9 Write Merge Buffer
The AMD-K6-2E processor contains an 8-byte write merge
buffer that allows the processor to conditionally combine data
from multiple noncacheable write cycles into this merge buffer.
The merge buffer operates in conjunction with the Memory
Type Range Registers (MTRRs). Refer to “Memory Type Range
Registers” on page 207 for a description of the MTRRs.
Merging multiple write cycles into a single write cycle reduces
processor bus utilization and processor stalls, thereby
increasing the overall system performance.
9.1 EWBE# Control
The presence of the merge buffer creates the potential to
perform out-of-order write cycles relative to the processor’s L1
cache. In general, the ordering of write cycles that are driven
externally on the system bus and those that hit the processor’s
cache can be controlled by the EWBE# signal. See “EWBE#
(External Write Buffer Empty)” on page 102 for more
information.
If EWBE# is sampled negated, the processor delays the
commitment of write cycles to cache lines in the Modified state
or Exclusive state in the processor’s cache. Therefore, the
system logic can enforce strong ordering by negating EWBE#
until the external write cycle is complete, thereby ensuring that
a subsequent write cycle that hits the cache does not complete
ahead of the external write cycle.
However, the addition of the write merge buffer introduces the
potential for out-of-order write cycles to occur between writes
to the merge buffer and writes to the processor’s cache. Because
these writes occur entirely within the processor and are not
sent out to the processor bus, the system logic is not able to
enforce strong ordering with the EWBE# signal.
206 Write Merge Buffer Chapter 9
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The EWBE control (EWBEC) bits in the EFER register provide
a mechanism for enforcing three different levels of write
ordering in the presence of the write merge buffer:
■EFER[3] is defined as the Global EWBE# Disable
(GEWBED). When GEWBED equals 1, the processor does
not attempt to enforce any write ordering internally or
externally (the EWBE# signal is ignored). This is the
maximum performance setting.
■EFER[2] is defined as the Speculative EWBE# Disable
(SEWBED). SEWBED only affects the processor when
GEWBED equals 0. If GEWBED equals 0 and SEWBED
equals 1, the processor enforces strong ordering for all
internal write cycles with the exception of write cycles
addressed to a range of memory defined as uncacheable
(UC) or write-combining (WC) by the MTRRs. In addition,
the processor samples the EWBE# signal. If EWBE# is
sampled negated, the processor delays the commitment of
write cycles to processor cache lines in the Modified state or
Exclusive state until EWBE# is sampled asserted.
This setting provides performance comparable to, but
slightly less than, the performance obtained when
GEWBED equals 1 because some degree of write ordering is
maintained.
■If GEWBED equals 0 and SEWBED equals 0, the processor
enforces strong ordering for all internal and external write
cycles. In this setting, the processor assumes, or speculates,
that strong order must be maintained between writes to the
merge buffer and writes that hit the processor’s cache. Once
the merge buffer is written out to the processor’s bus, the
EWBE# signal is sampled. If EWBE# is sampled negated, the
processor delays the commitment of write cycles to
processor cache lines in the Modified state or Exclusive
state until EWBE# is sampled asserted.
This setting is the default after RESET and provides the
lowest performance of the three settings because full write
ordering is maintained.
Chapter 9 Write Merge Buffer 207
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 37 summarizes the three settings of the EWBEC field for
the EFER register, along with the effect of write ordering and
performance. For more information on the EFER register, see
“Extended Feature Enable Register (EFER)” on page 43.
9.2 Memory Type Range Registers
The AMD-K6-2E processor provides two variable-range Memory
Type Range registers (MTRRs), MTRR0 and MTRR1, each of
which specifies a range of memory. Each range can be defined
as one of the following memory types:
■Uncacheable (UC) Memory—Memory read cycles are
sourced directly from the specified memory address and the
processor does not allocate a cache line. Memory write
cycles are targeted at the specified memory address and a
write allocation does not occur.
■Write-Combining (WC) Memory—Memory read cycles are
sourced directly from the specified memory address and the
processor does not allocate a cache line. The processor
conditionally combines data from multiple noncacheable
write cycles that are addressed within this range into a
merge buffer. Merging multiple write cycles into a single
write cycle reduces processor bus utilization and processor
stalls, thereby increasing the overall system performance.
This memory type is applicable for linear video frame
buffers.
Table 37. EWBEC Settings
EFER[3]
(GEWBED) EFER[2]
(SEWBED) Write Ordering Performance
1 0 or 1 None Best
0 1 All except UC/WC Close-to-Best
0 0 All Slowest
208 Write Merge Buffer Chapter 9
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
UC/WC Cacheability
Control Register
(UWCCR)
The MTRRs are accessed by addressing the 64-bit MSR known
as the UC/WC Cacheability Control Register (UWCCR). The
MSR address of the UWCCR is C000_0085h. Following reset, all
bits in the UWCCR register are 0. MTRR0 (lower 32 bits of the
UWCCR register) defines the size and memory type of range 0
and MTRR1 (upper 32 bits) defines the size and memory type of
range 1 (see Figure 79 on page 208).
.
Figure 79. UC/WC Cacheability Control Register (UWCCR)—MSR C000_0085h
Physical Base Address n (n=0, 1). This address is the 15 most-
significant bits of the physical base address of the memory
range. The least-significant 17 bits of the base address are not
needed because the base address is by definition always aligned
on a 128-Kbyte boundary.
Physical Address Mask n (n=0, 1). This value is the 15 most-
significant bits of a physical address mask that is used to define
the size of the memory range. This mask is logically ANDed
with both the physical base address field of the UWCCR
register and the physical address generated by the processor. If
the results of the two AND operations are equal, then the
generated physical address is considered within the range.
That is, if:
Mask & Physical Base Address = Mask & Physi cal Addres s Generat ed
then, the physical address generated by the processor is in the
range.
16 063
Physical Address Mask 0
1731
Physical Base Address 0
12
Physical Address Mask 1Physical Base Address 1
3233344849
U
C
0
W
C
0
U
C
1
W
C
1
MTRR1 MTRR0
Symbol Description Bits
UC0 Uncacheable Memory Type 0
WC0 Write-Combining Memory Type 1
Symbol Description Bits
UC1 Uncacheable Memory Type 32
WC1 Write-Combining Memory Type 33
Chapter 9 Write Merge Buffer 209
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
WCn (n=0, 1). When set to 1, this memory range is defined as
write combinable (see Table 38 on page 209). Write-combinable
memory is uncacheable.
UCn (n=0, 1). When set to 1, this memory range is defined as
uncacheable (see Table 38).
9.3 Memory-Range Restrictions
The following rules regarding the address alignment and size of
each range must be adhered to when programming the physical
base address and physical address mask fields of the UWCCR
register:
■The minimum size of each range is 128 Kbytes.
■The physical base address must be aligned on a 128-Kbyte
boundary.
■The physical base address must be range-size aligned. For
example, if the size of the range is 1 Mbyte, then the
physical base address must be aligned on a 1-Mbyte
boundary.
■All bits set to 1 in the physical address mask must be
contiguous. Likewise, all bits cleared to 0 in the physical
address mask must be contiguous. For example:
111_1111_1100_0000b is a valid physical address mask
111_1111_1101_0000b is invalid
Table 38. WC/UC Memory Type
WCn UCn Memory Type
0 0 No effect on cacheability or write combining
1 0 Write-combining memory range (uncacheable)
0 or 1 1 Uncacheable memory range
210 Write Merge Buffer Chapter 9
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 39 lists the valid physical address masks and the resulting
range sizes that can be programmed in the UWCCR register.
Table 39. Valid Masks and Range Sizes
Masks Size
111_1111_1111_1111b 128 Kbytes
111_1111_1111_1110b 256 Kbytes
111_1111_1111_1100b 512 Kbytes
111_1111_1111_1000b 1 Mbyte
111_1111_1111_0000b 2 Mbytes
111_1111_1110_0000b 4 Mbytes
111_1111_1100_0000b 8 Mbytes
111_1111_1000_0000b 16 Mbytes
111_1111_0000_0000b 32 Mbytes
111_1110_0000_0000b 64 Mbytes
111_1100_0000_0000b 128 Mbytes
111_1000_0000_0000b 256 Mbytes
111_0000_0000_0000b 512 Mbytes
110_0000_0000_0000b 1 Gbyte
100_0000_0000_0000b 2 Gbytes
000_0000_0000_0000b 4 Gbytes
Chapter 9 Write Merge Buffer 211
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
9.4 Examples
Suppose that the range of memory from 16 Mbytes to 32 Mbytes
is uncacheable, and the 8-Mbyte range of memory on top of 1
Gbyte is write-combinable. Range 0 is defined as the
uncacheable range, and range 1 is defined as the write-
combining range.
■Extracting the 15 most-significant bits of the 32-bit physical
base address that corresponds to 16 Mbytes (0100_0000h)
yields a physical base address 0 field of
000_0000_1000_0000b. Because the uncacheable range size
is 16 Mbytes, the physical mask value 0 field is
111_1111_1000_0000b, according to Table 39. Bit 1 of the
UWCCR register (WC0) is cleared to 0 and bit 0 of the
UWCCR register is set to 1 (UC0).
■Extracting the 15 most-significant bits of the 32-bit physical
base address that corresponds to 1 Gbyte (4000_0000h)
yields a physical base address 1 field of
010_0000_0000_0000b. Because the write-combining range
size is 8 Mbytes, the physical mask value 1 field is
111_1111_1100_0000b, according to Table 39. Bit 33 of the
UWCCR register (WC1) is set to 1 and bit 32 of the UWCCR
register is cleared to 0 (UC1).
212 Write Merge Buffer Chapter 9
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 10 Floating-Point and Multimedia Execution Units 213
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
10 Floating-Point and Multimedia Execution Units
10.1 Floating-Point Execution Unit
The AMD-K6-2E processor contains an IEEE 754-compatible
and IEEE 854-compatible floating-point execution unit
designed to accelerate the performance of software that utilizes
the x86 floating-point instruction set. Floating-point software is
typically written to manipulate numbers that are very large or
very small, that require a high degree of precision, or that result
from complex mathematical operations such as
transcendentals. Applications that take advantage of
floating-point operations include geometric calculations for
graphics acceleration, scientific, statistical, and engineering
applications, and business applications that use large amounts
of high-precision data.
The high-performance floating-point execution unit contains an
adder unit, a multiplier unit, and a divide/square root unit.
These low-latency units can execute floating-point instructions
in as few as two processor clocks. To increase performance, the
processor is designed to simultaneously decode most
floating-point instructions with most short-decodeable
instructions.
See “Software Environment” on page 23 for a description of the
floating-point data types, registers, and instructions.
Handling
Floating-Point
Exceptions
The AMD-K6-2E processor provides the following two types of
exception handling for floating-point exceptions:
■If the numeric error (NE) bit in CR0 is 1, the processor
invokes the interrupt 10h handler. In this manner, the
floating-point exception is completely handled by software.
■If the NE bit in CR0 is 0, the processor requires external
logic to generate an interrupt on the INTR signal in order to
handle the exception.
External Logic
Support of
Floating-Point
Exceptions
The processor provides the FERR# (Floating-Point Error) and
IGNNE# (Ignore Numeric Error) signals to allow the external
logic to generate the interrupt in a manner consistent with
PC/AT-compatible systems. The assertion of FERR# indicates
the occurrence of an unmasked floating-point exception
resulting from the execution of a floating-point instruction.
214 Floating-Point and Multimedia Execution Units Chapter 10
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
IGNNE# is used by the external hardware to control the effect
of an unmasked floating-point exception. Under certain
circumstances, if IGNNE# is sampled asserted, the processor
ignores the floating-point exception.
Figure 80 illustrates an implementation of external logic for
supporting floating-point exceptions. The following example
explains the operation of the external logic in Figure 80:
1. As the result of a floating-point exception, the processor
asserts FERR#.
2. The assertion of FERR# and the sampling of IGNNE#
negated indicates the processor has stopped instruction
execution and is waiting for an interrupt.
3. The assertion of FERR# leads to the assertion of INTR by
the interrupt controller.
4. The processor acknowledges the interrupt and jumps to the
corresponding interrupt service routine in which an I/O
write cycle to address port F0h leads to the assertion of
IGNNE#.
5. When IGNNE# is sampled asserted, the processor ignores
the floating-point exception and continues instruction
execution.
6. When the processor negates FERR#, the external logic
negates IGNNE#.
See “FERR# (Floating-Point Error)” on page 103 and “IGNNE#
(Ignore Numeric Exception)” on page 108 for more details.
Chapter 10 Floating-Point and Multimedia Execution Units 215
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 80. External Logic for Supporting Floating-Point Exceptions
10.2 Multimedia and 3DNow!™ Execution Units
The multimedia and 3DNow! execution units of the AMD-K6-2E
processor are designed to accelerate the performance of
software written using the industry-standard MMX instructions
and the 3DNow! instructions. Applications that can take
advantage of the MMX and 3DNow! instructions include
graphics, video and audio compression and decompression,
speech recognition, and telephony applications.
The multimedia execution unit can execute MMX instructions
in a single processor clock. All MMX and 3DNow! arithmetic
instructions are pipelined for higher performance. To increase
performance, the processor is designed to simultaneously
decode all MMX and 3DNow! instructions with most other
instructions.
For more information on MMX instructions, see the AMD-K6®
Processor Multimedia Technology Manual, order #20726. For
more information on 3DNow! instructions, see the 3DNow!™
Technology Manual, order #21928.
FERR#
Flip-Flop
CLOCKQ
DATAQ
CLEAR
IRQ13
Interrupt
Controller
I/O Address
Port F0h
AMD-K6™-2E
Processor
FERR#
INTR
IGNNE#
RESET
“1”
IGNNE#
Flip-Flop
CLOCKQ
DATAQ
CLEAR
216 Floating-Point and Multimedia Execution Units Chapter 10
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
10.3 Floating-Point and MMX™/3DNow!™ Instruction Compatibility
Registers The eight 64-bit MMX registers (which are also utilized by
3DNow! instructions) are mapped on the floating-point stack.
This enables backward compatibility with all existing software.
For example, the register saving event that is performed by
operating systems during task switching requires no changes to
the operating system. The same support provided in an
operating system’s interrupt 7 handler (Device Not Available)
for saving and restoring the floating-point registers also
supports saving and restoring the MMX registers.
Exceptions There are no new exceptions defined for supporting the MMX
and 3DNow! instructions. All exceptions that occur while
decoding or executing an MMX or 3DNow! instruction are
handled in existing exception handlers without modification.
FERR# and IGNNE# MMX instructions and 3DNow! instructions do not generate
floating-point exceptions. However, if an unmasked
floating-point exception is pending, the processor asserts
FERR# at the instruction boundary of the next floating-point
instruction, MMX instruction, 3DNow! instruction or WAIT
instruction.
The sampling of IGNNE# asserted only affects processor
operation during the execution of an error-sensitive
floating-point instruction, MMX instruction, 3DNow!
instruction or WAIT instruction when the NE bit in CR0 is
cleared to 0.
Chapter 11 System Management Mode (SMM) 217
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
11 System Management Mode (SMM)
SMM is an alternate operating mode entered by way of a system
management interrupt (SMI) and handled by an interrupt
service routine. SMM is designed for system control activities
such as power management. These activities appear
transparent to conventional operating systems like DOS and
Windows. SMM is primarily targeted for use by the Basic Input
Output System (BIOS) and specialized low-level device drivers.
The code and data for SMM are stored in the SMM memory
area, which is isolated from main memory.
The processor enters SMM by the system logic’s assertion of the
SMI# interrupt and the processor’s acknowledgment by the
assertion of SMIACT#. At this point the processor saves its state
into the SMM memory state-save area and jumps to the SMM
service routine. The processor returns from SMM when it
executes the resume (RSM) instruction from within the SMM
service routine. Subsequently, the processor restores its state
from the SMM save area, negates SMIACT#, and resumes
execution with the instruction following the point where it
entered SMM.
The following sections summarize the SMM state-save area,
entry into and exit from SMM, exceptions and interrupts in
SMM, memory allocation and addressing in SMM, and the SMI#
and SMIACT# signals.
11.1 SMM Operating Mode and Default Register Values
The software environment within SMM has the following
characteristics:
■Addressing and operation in real mode
■4-Gbyte segment limits
■Default 16-bit operand, address, and stack sizes, although
instruction prefixes can override these defaults
■Control transfers that do not override the default operand
size truncate the EIP to 16 bits
■Far jumps or calls cannot transfer control to a segment with
a base address requiring more than 20 bits, as in real mode
segment-base addressing
218 System Management Mode (SMM) Chapter 11
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
■A20M# is masked
■Interrupt vectors use the real-mode interrupt vector table
■The IF flag in EFLAGS is cleared (INTR not recognized)
■The TF flag in EFLAGS is cleared
■The NMI and INIT interrupts are disabled
■Debug register DR7 is cleared (debug traps disabled)
Figure 81 shows the default map of the SMM memory area. It
consists of a 64-Kbyte area, between 0003_0000h and
0003_FFFFh, of which the top 32 Kbytes (0003_8000h to
0003_FFFFh) must be populated with RAM. The default
code-segment (CS) base address for the area—called the SMM
base address—is at 0003_0000h. The top 512 bytes
(0003_FE00h to 0003_FFFFh) contain a fill-down SMM
state-save area. The default entry point for the SMM service
routine is 0003_8000h.
Figure 81. SMM Memory
SMM
State-Save
Area
SMM Base Address (CS)
Service Routine Entry Point
Fill Down
SMM
Service Routine
32-Kbyte
Minimum RAM
0003_8000h
0003_FE00h
0003_FFFFh
0003_0000h
Chapter 11 System Management Mode (SMM) 219
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 40 shows the initial state of registers when entering SMM.
11.2 SMM State-Save Area
When the processor acknowledges an SMI# interrupt by
asserting SMIACT#, it saves its state in a 512-byte SMM
state-save area shown in Table 41. The save begins at the top of
the SMM memory area (SMM base address + FFFFh) and fills
down to SMM base address + FE00h.
Table 41 shows the offsets in the SMM state-save area relative
to the SMM base address. The SMM service routine can alter
any of the read/write values in the state-save area.
Table 40. Initial State of Registers in System Management Mode (SMM)
Register SMM Initial State
General-Purpose Registers Unmodified
EFLAGs 0000_0002h
CR0 PE, EM, TS, and PG are cleared (bits 0, 2, 3, and 31).
The other bits are unmodified.
DR7 0000_0400h
GDTR, LDTR, IDTR, TSSR, DR6 Unmodified
EIP 0000_8000h
CS 0003_0000h
DS, ES, FS, GS, SS 0000_0000h
Table 41. SMM State-Save Area Map
Address Offset Contents Saved
FFFCh CR0
FFF8h CR3
FFF4h EFLAGS
FFF0h EIP
FFECh EDI
FFE8h ESI
FFE4h EBP
FFE0h ESP
FFDCh EBX
FFD8h EDX
FFD4h ECX
FFD0h EAX
220 System Management Mode (SMM) Chapter 11
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
FFCCh DR6
FFC8h DR7
FFC4h TR
FFC0h LDTR Base
FFBCh GS
FFB8h FS
FFB4h DS
FFB0h SS
FFACh CS
FFA8h ES
FFA4h I/O Trap Doubleword
FFA0h No data dump at that address
FF9Ch I/O Trap EIP1
FF98h No data dump at that address
FF94h No data dump at that address
FF90h IDT Base
FF8Ch IDT Limit
FF88h GDT Base
FF84h GDT Limit
FF80h TSS Attr
FF7Ch TSS Base
FF78h TSS Limit
FF74h No data dump at that address
FF70h LDT High
FF6Ch LDT Low
FF68h GS Attr
FF64h GS Base
FF60h GS Limit
FF5Ch FS Attr
FF58h FS Base
FF54h FS Limit
FF50h DS Attr
FF4Ch DS Base
FF48h DS Limit
FF44h SS Attr
FF40h SS Base
Table 41. SMM State-Save Area Map (continued)
Address Offset Contents Saved
Chapter 11 System Management Mode (SMM) 221
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
11.3 SMM Revision Identifier
The SMM revision identifier at offset FEFCh in the SMM
state-save area specifies the version of SMM and the extensions
that are available on the processor. The SMM revision identifier
fields are as follows:
■Bits 31–18—Reserved
■Bit 17—SMM base address relocation (1 = enabled)
■Bit 16—I/O trap restart (1 = enabled)
■Bits 15–0—SMM revision level for the AMD-K6-2E processor
= 0002h
FF3Ch SS Limit
FF38h CS Attr
FF34h CS Base
FF30h CS Limit
FF2Ch ES Attr
FF28h ES Base
FF24h ES Limit
FF20h No data dump at this address
FF1Ch No data dump at this address
FF18h No data dump at this address
FF14h CR2
FF10h CR4
FF0Ch I/O restart ESI1
FF08h I/O restart ECX1
FF04h I/O restart EDI1
FF02h HALT Restart Slot
FF00h I/O Trap Restart Slot
FEFCh SMM RevID
FEF8h SMM BASE
FEF7h–FE00h No data dump at these addresses
Notes:
1. Only contains information if SMI# is asserted during a valid I/O bus cycle.
Table 41. SMM State-Save Area Map (continued)
Address Offset Contents Saved
222 System Management Mode (SMM) Chapter 11
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 42 shows the format of the SMM revision identifier.
11.4 SMM Base Address
During RESET, the processor sets the base address of the
code-segment (CS) for the SMM memory area—the SMM base
address—to its default, 0003_0000h. The SMM base address at
offset FEF8h in the SMM state-save area can be changed by the
SMM service routine to any address that is aligned to a
32-Kbyte boundary. (Locations not aligned to a 32-Kbyte
boundary cause the processor to enter the Shutdown state when
executing the RSM instruction.)
In some operating environments it may be desirable to relocate
the 64-Kbyte SMM memory area to a high memory area in order
to provide more low memory for legacy software. During system
initialization, the base of the 64-Kbyte SMM memory area is
relocated by the BIOS. To relocate the SMM base address, the
system enters the SMM handler at the default address. This
handler changes the SMM base address location in the SMM
state-save area, copies the SMM handler to the new location,
and exits SMM.
The next time SMM is entered, the processor saves its state at
the new base address. This new address is used for every SMM
entry until the SMM base address in the SMM state-save area is
changed or a hardware reset occurs.
11.5 Halt Restart Slot
During entry into SMM, the halt restart slot at offset FF02h in
the SMM state-save area indicates if SMM was entered from the
Halt state. Before returning from SMM, the halt restart slot
(offset FF02h) can be written to by the SMM service routine to
specify whether the return from SMM takes the processor back
to the Halt state or to the next instruction after the HLT
instruction.
Table 42. SMM Revision Identifier
31–18 17 16 15–0
Reserved SMM Base Relocation I/O Trap Extension SMM Revision Level
0 1 1 0002h
Chapter 11 System Management Mode (SMM) 223
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Upon entry into SMM, the halt restart slot is defined as follows:
■Bits 15–1—Reserved
■Bit 0—Point of entry to SMM:
1 = entered from Halt state
0 = not entered from Halt state
After entry into the SMI handler and before returning from
SMM, the halt restart slot can be written using the following
definition:
■Bits 15–1—Reserved
■Bit 0—Point of return when exiting from SMM:
1 = return to Halt state
0 = return to next instruction after the HLT instruction
If the return from SMM takes the processor back to the Halt
state, the HLT instruction is not re-executed, but the Halt
special bus cycle is driven on the bus after the return.
11.6 I/O Trap Doubleword
If the assertion of SMI# is recognized during the execution of an
I/O instruction, the I/O trap doubleword at offset FFA4h in the
SMM state-save area contains information about the
instruction. The fields of the I/O trap doubleword are
configured as follows:
■Bits 31–16—I/O port address
■Bits 15–4—Reserved
■Bit 3—REP (repeat) string operation (1 = REP string, 0 = not
a REP string)
■Bit 2—I/O string operation (1 = I/O string, 0 = not a I/O
string)
■Bit 1—Valid I/O instruction (1 = valid, 0 = invalid)
■Bit 0—Input or output instruction (1 = INx, 0 = OUTx)
224 System Management Mode (SMM) Chapter 11
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 43 shows the format of the I/O trap doubleword.
The I/O trap doubleword is related to the I/O trap restart slot
(see “I/O Trap Restart Slot” on page 224). If bit 1 of the I/O
trap doubleword is set by the processor, it means that SMI#
was asserted during the execution of an I/O instruction. The
SMI handler tests bit 1 to see if there is a valid I/O instruction
trapped. If the I/O instruction is valid, the SMI handler is
required to ensure the I/O trap restart slot is set properly. The
I/O trap restart slot informs the CPU whether it should
re-execute the I/O instruction after the RSM or execute the
instruction following the trapped I/O instruction.
Note: If SMI# is sampled asserted during an I/O bus cycle a mini-
mum of three clock edges before BRDY# is sampled asserted,
the associated I/O instruction is guaranteed to be trapped by
the SMI handler.
11.7 I/O Trap Restart Slot
The I/O trap restart slot at offset FF00h in the SMM state-save
area specifies whether the trapped I/O instruction should be
re-executed on return from SMM. This slot in the state-save area
is called the I/O instruction restart function. Re-executing a
trapped I/O instruction is useful, for example, if an I/O write
occurs to a disk that is powered down. The system logic
monitoring such an access can assert SMI#. Then the SMM
service routine would query the system logic, detect a failed I/O
write, take action to power-up the I/O device, enable the I/O
trap restart slot feature, and return from SMM.
The fields of the I/O trap restart slot are defined as follows:
■Bits 31–16—Reserved
■Bits 15–0—I/O instruction restart on return from SMM:
0000h = execute the next instruction after the trapped
I/O instruction
00FFh = re-execute the trapped I/O instruction
Table 43. I/O Trap Doubleword Configuration
31—16 15—4 3 2 1 0
I/O Port
Address
Reserved REP String
Operation
I/O String
Operation
Valid I/O
Instruction
Input or
Output
Chapter 11 System Management Mode (SMM) 225
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 44 shows the format of the I/O trap restart slot.
The processor initializes the I/O trap restart slot to 0000h upon
entry into SMM. If SMM was entered due to a trapped I/O
instruction, the processor indicates the validity of the I/O
instruction by setting or clearing bit 1 of the I/O trap
doubleword at offset FFA4h in the SMM state-save area. The
SMM service routine should test bit 1 of the I/O trap
doubleword to determine if a valid I/O instruction was being
executed when entering SMM and before writing the I/O trap
restart slot. If the I/O instruction is valid, the SMM service
routine can safely rewrite the I/O trap restart slot with the value
00FFh, which causes the processor to re-execute the trapped I/O
instruction when the RSM instruction is executed. If the I/O
instruction is invalid, writing the I/O trap restart slot has
undefined results.
If a second SMI# is asserted and a valid I/O instruction was
trapped by the first SMM handler, the CPU services the second
SMI# prior to re-executing the trapped I/O instruction. The
second entry into SMM never has bit 1 of the I/O trap
doubleword set, and the second SMM service routine must not
rewrite the I/O trap restart slot.
During a simultaneous SMI# I/O instruction trap and debug
breakpoint trap, the AMD-K6-2E processor first responds to the
SMI# and postpones recognizing the debug exception until
after returning from SMM via the RSM instruction. If the debug
registers DR3–DR0 are used while in SMM, they must be saved
and restored by the SMM handler. The processor automatically
saves and restores DR7–DR6. If the I/O trap restart slot in the
SMM state-save area contains the value 00FFh when the RSM
instruction is executed, the debug trap does not occur until
after the I/O instruction is re-executed.
Table 44. I/O Trap Restart Slot
31–16 15–0
Reserved I/O Instruction restart on return from SMM:
■0000h = Execute the next instruction after the trapped I/O
■00FFh = Re-execute the trapped I/O instruction
226 System Management Mode (SMM) Chapter 11
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
11.8 Exceptions, Interrupts, and Debug in SMM
During an SMI# I/O trap, the exception/interrupt priority of the
AMD-K6-2E processor changes from its normal priority. The
normal priority places the debug traps at a priority higher than
the sampling of the FLUSH# or SMI# signals. However, during
an SMI# I/O trap, the sampling of the FLUSH# or SMI# signals
takes precedence over debug traps.
The processor recognizes the assertion of NMI within SMM
immediately after the completion of an IRET instruction. Once
NMI is recognized within SMM, NMI recognition remains
enabled until SMM is exited, at which point NMI masking is
restored to the state it was in before entering SMM.
Chapter 12 Test and Debug 227
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
12 Test and Debug
The AMD-K6-2E processor implements various test and debug
modes to enable the functional and manufacturing testing of
systems and boards that use the processor. In addition, the
debug features of the processor allow designers to debug the
instruction execution of software components. This chapter
describes the following test and debug features:
■Built-In Self-Test (BIST)—The BIST, which is invoked after
the falling transition of RESET, runs internal tests that
exercise most on-chip RAM structures.
■Three-State Test Mode—A test mode that causes the
processor to float its output and bidirectional pins.
■Boundary-Scan Test Access Port (TAP)—The Joint Test
Action Group (JTAG) test access function defined by the
IEEE Standard Test Access Port and Boundary-Scan
Architecture (IEEE 1149.1-1990) specification.
■Level-One (L1) Cache Inhibit—A feature that disables the
processor’s internal L1 instruction and data caches.
■Debug Support—Consists of all x86-compatible software
debug features, including the debug extensions.
12.1 Built-In Self-Test (BIST)
Following the falling transition of RESET, the processor
unconditionally runs its built-in self test (BIST). The internal
resources tested during BIST include the following:
■L1 instruction and data caches
■Instruction and Data Translation Lookaside Buffers (TLBs)
The contents of the EAX general-purpose register after the
completion of reset indicate if the BIST was successful.
■If EAX contains 0000_0000h, then BIST was successful.
■If EAX is non-zero, the BIST failed.
Following the completion of the BIST, the processor jumps to
address FFFF_FFF0h to start instruction execution, regardless
of the outcome of the BIST. The BIST takes approximately
295,000 processor clocks to complete.
228 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
12.2 Three-State Test Mode
The three-state test mode causes the processor to float its
output and bidirectional pins, which is useful for board-level
manufacturing testing. In this mode, the processor is
electrically isolated from other components on a system board,
allowing automated test equipment (ATE) to test components
that drive the same signals as those the processor floats.
If the FLUSH# signal is sampled Low during the falling
transition of RESET, the processor enters the three-state test
mode. (See “FLUSH# (Cache Flush)” on page 104 for the
specific sampling requirements.) The signals floated in the
three-state test mode are as follows:
The VCC2DET, VCC2H/L#, and TDO signals are the only
outputs not floated in the three-state test mode.
■VCC2DET and VCC2H/L# must remain Low to ensure the
system continues to supply the specified processor core
voltage to the VCC2 pins.
■TDO is never floated because the boundary-scan Test Access
Port must remain enabled at all times, including during the
three-state test mode.
The three-state test mode is exited when the processor samples
RESET asserted.
■A[31:3] ■D/C# ■M/IO#
■ADS# ■D[63:0] ■PCD
■ADSC# ■DP[7:0] ■PCHK#
■AP ■FERR# ■PWT
■APCHK# ■HIT# ■SCYC
■BE[7:0]# ■HITM# ■SMIACT#
■BREQ ■HLDA ■W/R#
■CACHE# ■LOCK#
Chapter 12 Test and Debug 229
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
12.3 Boundary-Scan Test Access Port (TAP)
The boundary-scan Test Access Port (TAP) is an IEEE standard
that defines synchronous scanning test methods for complex
logic circuits, such as boards containing a processor. The
AMD-K6-2E processor supports the TAP standard defined in
the IEEE Standard Test Access Port and Boundary-Scan
Architecture (IEEE 1149.1-1990) specification.
Boundary scan testing uses a shift register consisting of the
serial interconnection of boundary-scan cells that correspond to
each I/O buffer of the processor. This non-inverting register
chain, called a Boundary Scan register (BSR), can be used to
capture the state of every processor pin and to drive every
processor output and bidirectional pin to a known state.
Each BSR of every component on a board that implements the
boundary-scan architecture can be serially interconnected to
enable component interconnect testing.
Test Access Port The Test Access Port (TAP) consists of the following:
■Test Access Port (TAP) Controller—The TAP controller is a
synchronous, finite state machine that uses the TMS and
TDI input signals to control a sequence of test operations.
See “TAP Controller State Machine” on page 236 for a list
of TAP states and their definition.
■Instruction Register (IR)—The IR contains the instructions
that select the test operation to be performed and the Test
Data Register (TDR) to be selected. See “TAP Registers” on
page 231 for more details on the IR.
■Test Data Registers (TDR)—The three TDRs are used to
process the test data. Each TDR is selected by an
instruction in the Instruction Register (IR). See “TAP
Registers” on page 231 for a list of these registers and their
functions.
230 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
TAP Signals The test signals associated with the TAP controller are as
follows:
■TCK—The Test Clock for all TAP operations. The rising
edge of TCK is used for sampling TAP signals, and the
falling edge of TCK is used for asserting TAP signals. The
state of the TMS signal sampled on the rising edge of TCK
causes the state transitions of the TAP controller to occur.
TCK can be stopped in the logic 0 or 1 state.
■TDI—The Test Data Input represents the input to the most
significant bit of all TAP registers, including the IR and all
test data registers. Test data and instructions are serially
shifted by one bit into their respective registers on the rising
edge of TCK.
■TDO—The Test Data Output represents the output of the
least significant bit of all TAP registers, including the IR and
all test data registers. Test data and instructions are serially
shifted by one bit out of their respective registers on the
falling edge of TCK.
■TMS—The Test Mode Select input specifies the test
function and sequence of state changes for boundary-scan
testing. If TMS is sampled High for five or more consecutive
clocks, the TAP controller enters its reset state.
■TRST#—The Test Reset signal is an asynchronous reset that
unconditionally causes the TAP controller to enter its reset
state.
Refer to “Electrical Data” on page 253 and “Signal Switching
Characteristics” on page 267 to obtain the electrical
specifications of the test signals.
Chapter 12 Test and Debug 231
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
TAP Registers The AMD-K6-2E processor provides an Instruction register (IR)
and three Test Data registers (TDR) to support the
boundary-scan architecture. The IR and one of the TDRs—the
Boundary-Scan register (BSR)—consist of a shift register and
an output register. The shift register is loaded in parallel in the
Capture states. (See “TAP Controller State Machine” on page
236 for a description of the TAP controller states.) In addition,
the shift register is loaded and shifted serially in the Shift
states. The output register is loaded in parallel from its
corresponding shift register in the Update states.
Instruction Register (IR). The IR is a 5-bit register, without parity,
that determines which instruction to run and which test data
register to select. When the TAP controller enters the
Capture-IR state, the processor loads the following bits into the
IR shift register:
■01b—Loaded into the two least significant bits, as specified
by the IEEE 1149.1 standard
■000b—Loaded into the three most significant bits
Loading 00001b into the IR shift register during the Capture-IR
state results in loading the SAMPLE/PRELOAD instruction.
For each entry into the Shift-IR state, the IR shift register is
serially shifted by one bit toward the TDO pin. During the shift,
the most significant bit of the IR shift register is loaded from
the TDI pin.
The IR output register is loaded from the IR shift register in the
Update-IR state, and the current instruction is defined by the IR
output register. See “TAP Instructions” on page 235 for a list and
definition of the instructions supported by the AMD-K6-2E
processor.
Boundary Scan Register (BSR). The Boundary Scan Register is a Test
Data register consisting of the interconnection of 152
boundary-scan cells. Each output and bidirectional pin of the
processor requires a two-bit cell, where one bit corresponds to
the pin and the other bit is the output enable for the pin. When
a 0 is shifted into the enable bit of a cell, the corresponding pin
is floated, and when a 1 is shifted into the enable bit, the pin is
driven valid. Each input pin requires a one-bit cell that
corresponds to the pin. The last cell of the BSR is reserved and
does not correspond to any processor pin.
232 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The total number of bits that comprise the BSR is 281. Table 45
on page 233 lists the order of these bits, where TDI is the input
to bit 280, and TDO is driven from the output of bit 0. The
entries listed as pin_E (where pin is an output or bidirectional
signal) are the enable bits.
If the BSR is the register selected by the current instruction
and the TAP controller is in the Capture-DR state, the processor
loads the BSR shift register as follows:
■If the current instruction is SAMPLE/PRELOAD, then the
current state of each input, output, and bidirectional pin is
loaded. A bidirectional pin is treated as an output if its
enable bit equals 1, and it is treated as an input if its enable
bit equals 0.
■If the current instruction is EXTEST, then the current state
of each input pin is loaded. A bidirectional pin is treated as
an input, regardless of the state of its enable.
While in the Shift-DR state, the BSR shift register is serially
shifted toward the TDO pin. During the shift, bit 280 of the BSR
is loaded from the TDI pin.
The BSR output register is loaded with the contents of the BSR
shift register in the Update-DR state. If the current instruction
is EXTEST, the processor’s output pins, as well as those
bidirectional pins that are enabled as outputs, are driven with
their corresponding values from the BSR output register.
Chapter 12 Test and Debug 233
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 45. Boundary Scan Bit Definitions1
Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable
280 D35_E 247 D19 214 BF1 181 A24 148 A14 115 BE7#
279 D35 246 D16_E 213 BF2 180 A18_E 147 A17_E 114 PCD_E
278 D29_E 245 D16 212 RESET 179 A18 146 A17 113 PCD
277 D29 244 D17_E 211 BF0 178 A5_E 145 A16_E 112 DC_E
276 D33_E 243 D17 210 FLUSH# 177 A5 144 A16 111 D/C#
275 D33 242 D15_E 209 INTR 176 EADS# 143 HIT_E 110 WR_E
274 D27_E 241 D15 208 NMI 175 A22_E 142 HIT# 109 W/R#
273 D27 240 DP1_E 207 SMI# 174 A22 141 ADS_E 108 NA#
272 DP0_E 239 DP1 206 A25_E 173 AHOLD 140 ADS# 107 PWT_E
271 DP0 238 D13_E 205 A25 172 HITM_E 139 CLK 106 PWT
270 DP3_E 237 D13 204 A26_E 171 HITM# 138 ADSC_E 105 CACHE_E
269 DP3 236 D6_E 203 A26 170 A4_E 137 ADSC# 104 CACHE#
268 D25_E 235 D6 202 A29_E 169 A4 136 BE0_E 103 WB/WT#
267 D25 234 D14_E 201 A29 168 A9_E 135 BE0# 102 MIO_E
266 D0_E 233 D14 200 A28_E 167 A9 134 AP_E 101 M/IO#
265 D0 232 D11_E 199 A28 166 A8_E 133 AP 100 BREQ_E
264 D30_E 231 D11 198 A23_E 165 A8 132 BE1_E 99 BREQ
263 D30 230 D1_E 197 A23 164 A19_E 131 BE1# 98 SCYC_E
262 DP2_E 229 D1 196 A27_E 163 A19 130 BE2_E 97 SCYC
261 DP2 228 D12_E 195 A27 162 BOFF# 129 BE2# 96 LOCK_E
260 D2_E 227 D12 194 A11_E 161 A6_E 128 BRDY# 95 LOCK#
259 D2 226 D10_E 193 A11 160 A6 127 BE3_E 94 APCHK_E
258 D28_E 225 D10 192 A3_E 159 A20_E 126 BE3# 93 APCHK#
257 D28 224 D7_E 191 A3 158 A20 125 BE4_E 92 PCHK_E
256 D24_E 223 D7 190 A31_E 157 A13_E 124 BE4# 91 PCHK#
255 D24 222 D8_E 189 A31 156 A13 123 BRDYC# 90 EWBE#
254 D26_E 221 D8 188 A21_E 155 A12_E 122 BE5_E 89 SMIACT_E
253 D26 220 D9_E 187 A21 154 A12 121 BE5# 88 SMIACT#
252 D21_E 219 D9 186 A30_E 153 A10_E 120 BE6_E 87 FERR_E
251 D21 218 HOLD 185 A30 152 A10 119 BE6# 86 FERR#
250 D18_E 217 STPCLK# 184 A7_E 151 A15_E 118 KEN# 85 D20_E
249 D18 216 INIT 183 A7 150 A15 117 INV 84 D20
248 D19_E 215 IGNNE# 182 A24_E 149 A14_E 116 BE7_E 83 D22_E
234 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Device Identification Register (DIR). The DIR is a 32-bit Test Data
register selected during the execution of the IDCODE
instruction. The fields of the DIR and their values are shown in
Table 46 and are defined as follows:
■Version Code—This 4-bit field is incremented by AMD
manufacturing for each major revision of silicon.
■Part Number—This 16-bit field identifies the specific
processor model.
■Manufacturer—This 11-bit field identifies the manufacturer
of the component (AMD).
■LSB—The least significant bit (LSB) of the DIR is always 1,
as specified by the IEEE 1149.1 standard.
82 D22 68 D54_E 54 D47_E 40 D62_E 26 D38_E 12 D3_E
81 D23_E 67 D54 53 D47 39 D62 25 D38 11 D3
80 D23 66 D50_E 52 D59_E 38 D49_E 24 D58_E 10 D39_E
79 A20M# 65 D50 51 D59 37 D49 23 D58 9 D39
78 HLDA_E 64 D56_E 50 D51_E 36 DP4_E 22 D42_E 8 D32_E
77 HLDA 63 D56 49 D51 35 DP4 21 D42 7 D32
76 DP7_E 62 D55_E 48 D45_E 34 D4_E 20 D36_E 6 D5_E
75DP7 61D55 47D45 33D4 19D36 5 D5
74 D63_E 60 D48_E 46 D61_E 32 D46_E 18 D60_E 4 D37_E
73 D63 59 D48 45 D61 31 D46 17 D60 3 D37
72 D52_E 58 D57_E 44 DP5_E 30 D41_E 16 D40_E 2 D31_E
71 D52 57 D57 43 DP5 29 D41 15 D40 1 D31
70 DP6_E 56 D53_E 42 D43_E 28 D44_E 14 D34_E 0 Reserved
69 DP6 55 D53 41 D43 27 D44 13 D34
Notes:
1. TDI is the input to bit 280, and TDO is driven from the output of bit 0. The entries listed as pin_E (where pin is an output or
bidirectional signal) are the enable bits.
Table 45. Boundary Scan Bit Definitions1 (continued)
Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable Bit Pin/Enable
Table 46. Device Identification Register
Version Code
(Bits 31–28) Part Number
(Bits 27–12) Manufacturer
(Bits 11–1) LSB
(Bit 0)
Xh 0580h 00000000001b 1b
Chapter 12 Test and Debug 235
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Bypass Register (BR). The BR is a Test Data register consisting of a
1-bit shift register that provides the shortest path between TDI
and TDO. When the processor is not involved in a test
operation, the BR can be selected by an instruction to allow the
transfer of test data through the processor without having to
serially scan the test data through the BSR. This functionality
preserves the state of the BSR and significantly reduces test
time.
The BR register is selected by the BYPASS and HIGHZ
instructions as well as by any instructions not supported by the
AMD-K6-2E processor.
TAP Instructions The processor supports the three instructions required by the
IEEE 1149.1 standard—EXTEST, SAMPLE/PRELOAD, and
BYPASS—as well as two additional optional instructions—
IDCODE and HIGHZ.
Table 47 shows the complete set of TAP instructions supported
by the processor along with the 5-bit Instruction Register
encoding and the register selected by each instruction.
EXTEST Instruction. When the EXTEST instruction is executed,
the processor loads the BSR shift register with the current state
of the input and bidirectional pins in the Capture-DR state and
drives the output and bidirectional pins with the corresponding
values from the BSR output register in the Update-DR state.
Table 47. Supported Test Access Port (TAP) Instructions
Instruction Encoding Register Description
EXTEST100000b BSR Sample inputs and drive outputs
SAMPLE / PRELOAD 00001b BSR Sample inputs and outputs, then load the BSR
IDCODE 00010b DIR Read DIR
HIGHZ 00011b BR Float outputs and bidirectional pins
BYPASS200100b–11110b BR Undefined instruction, execute the BYPASS instruction
BYPASS311111b BR Connect TDI to TDO to bypass the BSR
Notes:
1. Following the execution of the EXTEST instruction, the processor must be reset to return to normal, non-test operation.
2. These instruction encodings are undefined on the AMD-K6-2E processor and default to the BYPASS instruction.
3. Because the TDI input contains an internal pullup, the BYPASS instruction is executed if the TDI input is not connected or open
during an instruction scan operation. The BYPASS instruction does not affect the normal operational state of the processor.
236 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
SAMPLE/PRELOAD Instruction. The SAMPLE/PRELOAD instruction
performs two functions. These functions are as follows:
■During the Capture-DR state, the processor loads the BSR
shift register with the current state of every input, output,
and bidirectional pin.
■During the Update-DR state, the BSR output register is
loaded from the BSR shift register in preparation for the
next EXTEST instruction.
The SAMPLE/PRELOAD instruction does not affect the normal
operational state of the processor.
BYPASS Instruction. The BYPASS instruction selects the BR
register, which reduces the boundary-scan length through the
processor from 281 to one (TDI to BR to TDO). The BYPASS
instruction does not affect the normal operational state of the
processor.
IDCODE Instruction. The IDCODE instruction selects the DIR
register, allowing the device identification code to be shifted
out of the processor. This instruction is loaded into the IR when
the TAP controller is reset. The IDCODE instruction does not
affect the normal operational state of the processor.
HIGHZ Instruction. The HIGHZ instruction forces all output and
bidirectional pins to be floated. During this instruction, the BR
is selected and the normal operational state of the processor is
not affected.
TAP Controller State
Machine The TAP controller state diagram is shown in Figure 82 on page
237. State transitions occur on the rising edge of TCK. The logic
0 or 1 next to the states represents the value of the TMS signal
sampled by the processor on the rising edge of TCK.
Chapter 12 Test and Debug 237
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 82. TAP State Diagram
Test-Logic-Reset
Shift-DR
Pause-DR
Update-DR Update-IR
0
1
0
0
0
0
0
11
1
1 1
0
0
0
0
0
11
00
11
0
11
0
0
1
11
1
Run-Test/Idle
Exit2-IR
Exit1-IR
Pause-IR
Shift-IR
Select-DR-Scan Select-IR-Scan
Capture-DR Capture-IR
Exit1-DR
Exit2-DR
IEEE Std 1149.1-1990, Copyright © 1990. IEEE. All rights reserved
238 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
The states of the TAP controller are described as follows:
Test-Logic-Reset. This state represents the initial reset state of the
TAP controller and is entered when the processor samples
RESET asserted, when TRST# is asynchronously asserted, and
when TMS is sampled High for five or more consecutive clocks.
In addition, this state can be entered from the Select-IR-Scan
state. The IR is initialized with the IDCODE instruction, and
the processor’s normal operation is not affected in this state.
Capture-DR. During the SAMPLE/PRELOAD instruction, the
processor loads the BSR shift register with the current state of
every input, output, and bidirectional pin. During the EXTEST
instruction, the processor loads the BSR shift register with the
current state of every input and bidirectional pin.
Capture-IR. When the TAP controller enters the Capture-IR state,
the processor loads 01b into the two least significant bits of the
IR shift register and loads 000b into the three most significant
bits of the IR shift register.
Shift-DR. While in the Shift-DR state, the selected TDR shift
register is serially shifted toward the TDO pin. During the shift,
the most significant bit of the TDR is loaded from the TDI pin.
Shift-IR. While in the Shift-IR state, the IR shift register is
serially shifted toward the TDO pin. During the shift, the most
significant bit of the IR is loaded from the TDI pin.
Update-DR. During the SAMPLE/PRELOAD instruction, the BSR
output register is loaded with the contents of the BSR shift
register. During the EXTEST instruction, the output pins, as
well as those bidirectional pins defined as outputs, are driven
with their corresponding values from the BSR output register.
Update-IR. In this state, the IR output register is loaded from the
IR shift register, and the current instruction is defined by the
IR output register.
Chapter 12 Test and Debug 239
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The following states have no effect on the normal or test
operation of the processor other than as shown in Figure 82 on
page 237:
■Run-Test/Idle—This state is an idle state between scan
operations.
■Select-DR-Scan—This is the initial state of the test data
register state transitions.
■Select-IR-Scan—This is the initial state of the Instruction
Register state transitions.
■Exit1-DR—This state is entered to terminate the shifting
process and enter the Update-DR state.
■Exit1-IR—This state is entered to terminate the shifting
process and enter the Update-IR state.
■Pause-DR—This state is entered to temporarily stop the
shifting process of a test data register.
■Pause-IR—This state is entered to temporarily stop the
shifting process of the instruction register.
■Exit2-DR—This state is entered in order to either terminate
the shifting process and enter the Update-DR state or to
resume shifting following the exit from the Pause-DR state.
■Exit2-IR—This state is entered in order to either terminate
the shifting process and enter the Update-IR state or to
resume shifting following the exit from the Pause-IR state.
12.4 L1 Cache Inhibit
The AMD-K6-2E processor provides a means for inhibiting the
normal operation of its L1 instruction and data caches while
still supporting an external level-2 (L2) cache. This capability
allows system designers to disable the L1 cache during the
testing and debug of an L2 cache.
If the Cache Inhibit bit (bit 3) of test register 12 (TR12) is 0, the
processor’s L1 cache is enabled and operates as described in
“Cache Organization” on page 185. If the Cache Inhibit bit is 1,
the L1 cache is disabled and no new cache lines are allocated.
Even though new allocations do not occur, valid L1 cache lines
remain valid and are read by the processor when a requested
address hits a cache line. In addition, the processor continues to
support inquire cycles initiated by the system logic, including
240 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
the execution of writeback cycles when a modified cache line is
hit.
While the L1 is inhibited, the processor continues to drive the
PCD output signal appropriately, which system logic can use to
control external L2 caching.
In order to completely disable the L1 cache so that no valid
lines exist in the cache, the Cache Inhibit bit must be set to 1
and the cache must be flushed in one of the following ways:
■By asserting the FLUSH# input signal
■By executing the WBINVD instruction
■By executing the INVD instruction (modified cache lines are
not written back to memory)
■By using the Page Flush/Invalidate register (PFIR) (see
“Page Flush/Invalidate Register (PFIR)” on page 200)
12.5 Debug
The AMD-K6-2E processor implements the standard x86 debug
functions, registers, and exceptions. In addition, the processor
supports the I/O breakpoint debug extension. The debug
feature assists programmers and system designers during
software execution tracing by generating exceptions when one
or more events occur during processor execution. The exception
handler, or debugger, can be written to perform various tasks,
such as displaying the conditions that caused the breakpoint to
occur, displaying and modifying register or memory contents, or
single-stepping through program execution.
The following sections describe the debug registers and the
various types of breakpoints and exceptions that the processor
supports.
Chapter 12 Test and Debug 241
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Debug Registers Figures 83 through 86 show the 32-bit debug registers
supported by the processor.
Figure 83. Debug Register DR7
9876543210101112131415
L
2L
1
L
3
G
3
G
EL
EL
0
G
0
G
1
G
2
G
D
25 24 23 22 21 20 19 18 17 16262728293031
R/W
3
LEN
3R/W
2
LEN
2R/W
1
LEN
1R/W
0
LEN
0
Reserved
Symbol Description Bits
LEN 3 Length of Breakpoint #3 31–30
R/W 3 Type of Transaction(s) to Trap 29–28
LEN 2 Length of Breakpoint #2 27–26
R/W 2 Type of Transaction(s) to Trap 25–24
LEN 1 Length of Breakpoint #1 23–22
R/W 1 Type of Transaction(s) to Trap 21–20
LEN 0 Length of Breakpoint #0 19–18
R/W 0 Type of Transaction(s) to Trap 17–16
Symbol Description Bit
GD General Detect Enabled 13
GE Global Exact Breakpoint Enabled 9
LE Local Exact Breakpoint Enabled 8
G3 Global Exact Breakpoint # 3 Enabled 7
L3 Local Exact Breakpoint # 3 Enabled 6
G2 Global Exact Breakpoint # 2 Enabled 5
L2 Local Exact Breakpoint # 2 Enabled 4
G1 Global Exact Breakpoint # 1 Enabled 3
L1 Local Exact Breakpoint # 1 Enabled 2
G0 Global Exact Breakpoint # 0 Enabled 1
L0 Local Exact Breakpoint # 0 Enabled 0
242 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 84. Debug Register DR6
Figure 85. Debug Registers DR5 and DR4
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
B
1
B
2
B
SB
0
B
TB
DB
3
Reserved
Symbol Description Bit
BT Breakpoint Task Switch 15
BS Breakpoint Single Step 14
BD Breakpoint Debug Access Detected 13
B3 Breakpoint #3 Condition Detected 3
B2 Breakpoint #2 Condition Detected 2
B1 Breakpoint #1 Condition Detected 1
B0 Breakpoint #0 Condition Detected 0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Reserved
DR5
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Reserved
DR4
Chapter 12 Test and Debug 243
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 86. Debug Registers DR3, DR2, DR1, and DR0
DR3–DR0. The processor allows the setting of up to four
breakpoints. DR3–DR0 contain the linear addresses for
breakpoint 3 through breakpoint 0, respectively, and are
compared to the linear addresses of processor cycles to
determine if a breakpoint occurs. Debug register DR7 defines
the specific type of cycle that must occur in order for the
breakpoint to occur.
DR5–DR4. When debugging extensions are disabled (bit 3 of CR4
is 0), the DR5 and DR4 registers are mapped to DR7 and DR6,
respectively, in order to be software compatible with previous
generations of x86 processors. When debugging extensions are
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 3 32-bit Linear Address
DR3
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 0 32-bit Linear Address
DR0
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 2 32-bit Linear Address
DR2
987654321010111213141516171819202131 30 29 28 27 26 25 24 23 22
Breakpoint 1 32-bit Linear Address
DR1
244 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
enabled (bit 3 of CR4 is 1), any attempt to load DR5 or DR4
results in an undefined opcode exception. Likewise, any
attempt to store DR5 or DR4 also results in an undefined
opcode exception.
DR6. If a breakpoint is enabled in DR7, and the breakpoint
conditions as defined in DR7 occur, then the corresponding B
bit (B3–B0) in DR6 is set to 1. In addition, any other breakpoints
defined using these particular breakpoint conditions are
reported by the processor by setting the appropriate B bits in
DR6, regardless of whether these breakpoints are enabled or
disabled. However, if a breakpoint is not enabled, a debug
exception does not occur for that breakpoint.
If the processor decodes an instruction that writes or reads DR7
through DR0, the BD bit (bit 13) in DR6 is set to 1 (if enabled in
DR7) and the processor generates a debug exception. This
operation allows control to pass to the debugger prior to debug
register access by software.
If the Trap Flag (bit 8) of the EFLAGS register is 1, the
processor generates a debug exception after the successful
execution of every instruction (single-step operation) and sets
the BS bit (bit 14) in DR6 to indicate the source of the
exception.
When the processor switches to a new task and the debug trap
bit (T bit) in the corresponding Task State Segment (TSS) is 1,
the processor sets the BT bit (bit 15) in DR6 and generates a
debug exception.
DR7. When set to 1, L3–L0 locally enable breakpoints 3 through
0, respectively. L3–L0 are cleared to 0 whenever the processor
executes a task switch. Clearing L3–L0 to 0 disables the
breakpoints and ensures that these particular debug exceptions
are only generated for a specific task.
When set to 1, G3–G0 globally enable breakpoints 3 through 0,
respectively. Unlike L3–L0, G3–G0 are not cleared to 0
whenever the processor executes a task switch. Not clearing
G3–G0 to 0 allows breakpoints to remain enabled across all
tasks. If a breakpoint is enabled globally but disabled locally,
the global enable overrides the local enable.
Chapter 12 Test and Debug 245
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The LE (bit 8) and GE (bit 9) bits in DR7 have no effect on the
operation of the processor and are provided to be software-
compatible with previous generations of x86 processors.
When set to 1, the GD bit in DR7 (bit 13) enables the debug
exception associated with the BD bit (bit 13) in DR6. This bit is
cleared to 0 when a debug exception is generated.
LEN3–LEN0 and RW3–RW0 are two-bit fields in DR7 that
specify the length and type of each breakpoint as defined in
Table 48.
Debug Exceptions A debug exception is categorized as either a debug trap or a
debug fault.
■A debug trap calls the debugger following the execution of
the instruction that caused the trap.
■A debug fault calls the debugger prior to the execution of the
instruction that caused the fault.
All debug traps and faults generate either an Interrupt 01h or
an Interrupt 03h exception.
Table 48. DR7 LEN and RW Definitions
LEN Bits1
Notes:
1. LEN bits equal to 10b is undefined.
RW Bits Breakpoint
00b 00b2
2. When RW equals 00b, LEN must be equal to 00b.
Instruction Execution
00b 01b One-byte Data Write
01b Two-byte Data Write
11b Four-byte Data Write
00b 10b3
3. When RW equals 10b, debugging extensions (DE) must be enabled (bit 3 of CR4 must be set
to 1). If DE is cleared to 0, then RW equal to 10b is undefined.
One-byte I/O Read or Write
01b Two-byte I/O Read or Write
11b Four-byte I/O Read or Write
00b 11b One-byte Data Read or Write
01b Two-byte Data Read or Write
11b Four-byte Data Read or Write
246 Test and Debug Chapter 12
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Interrupt 01h. The following events are considered debug traps
that cause the processor to generate an Interrupt 01h
exception:
■Enabled breakpoints for data and I/O cycles
■Single-step trap
■Task-switch trap
The following events are considered debug faults that cause the
processor to generate an Interrupt 01h exception:
■Enabled breakpoints for instruction execution
■BD bit in DR6 set to 1
Interrupt 03h. The INT 3 instruction is defined in the x86
architecture as a breakpoint instruction. This instruction
causes the processor to generate an Interrupt 03h exception.
This exception is a debug trap because the debugger is called
following the execution of the INT 3 instruction.
The INT 3 instruction is a one-byte instruction (opcode CCh)
typically used to insert a breakpoint in software by writing CCh
to the address of the first byte of the instruction to be trapped
(the target instruction). Following the trap, if the target
instruction is to be executed, the debugger must replace the
INT 3 instruction with the first byte of the target instruction.
Chapter 13 Clock Control 247
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
13 Clock Control
13.1 Clock Control States
The AMD-K6-2E processor supports five modes of clock control.
The processor can transition between these modes to maximize
performance, to minimize power dissipation, or to provide a
balance between performance and power. (See “Power
Dissipation” on page 258 for the maximum power dissipation of
the AMD-K6-2E processor within the normal and the
reduced-power states.)
The five clock-control states supported are:
■Normal State—The processor is running in real mode,
virtual-8086 mode, protected mode, or system management
mode (SMM). In this state, all clocks are running— including
the external bus clock, CLK, and the internal processor
clock—and the full features and functions of the processor
are available.
■Halt State—This low-power state is entered following the
successful execution of the HLT instruction. During this
state, the internal processor clock is stopped.
■Stop Grant State—This low-power state is entered following
the recognition of the assertion of the STPCLK# signal.
During this state, the internal processor clock is stopped.
■Stop Grant Inquire State—This state is entered from the
Halt state and the Stop Grant state as the result of a
system-initiated inquire cycle.
■Stop Clock State—This low-power state is entered from the
Stop Grant state when the CLK signal is stopped.
Figure 87 on page 248 illustrates the clock control state
transitions. Each of the four reduced-power states are described
in the following sections.
248 Clock Control Chapter 13
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 87. Clock Control State Transitions
EADS# Asserted EADS# Asserted
HLT Instruction
Stop Grant
State
Normal Mode
– Real
– Virtual-8086
– Protected
– SMM
Halt
State
Stop Clock
State
RESET, SMI#, INIT,
or INTR Asserted
Stop Grant
Inquire
State
STPCLK# Asserted
STPCLK# Negated,
or RESET Asserted
CLK
Started
CLK
Stopped
Writeback
Completed
Writeback
Completed
Chapter 13 Clock Control 249
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
13.2 Halt State
Enter Halt State During the execution of the HLT instruction, the AMD-K6-2E
processor executes a Halt special cycle. After BRDY# is
sampled asserted during this cycle, and then EWBE# is also
sampled asserted (if not masked off), the processor enters the
Halt state in which the processor disables most of its internal
clock distribution.
To support the following operations, the internal phase-lock
loop (PLL) continues to run, and some internal resources are
still clocked in the Halt state:
■Inquire Cycles—The processor continues to sample AHOLD,
BOFF#, and HOLD to support inquire cycles that are
initiated by the system logic. The processor transitions to
the Stop Grant Inquire state during the inquire cycle. After
returning to the Halt state following the inquire cycle, the
processor does not execute another Halt special cycle.
■Flush Cycles—The processor continues to sample FLUSH#.
If FLUSH# is sampled asserted, the processor performs the
flush operation in the same manner as it is performed in the
Normal state. Upon completing the flush operation, the
processor executes the Halt special cycle which indicates
the processor is in the Halt state.
■Time Stamp Counter (TSC)—The TSC continues to count in
the Halt state.
■Signal Sampling—The processor continues to sample INIT,
INTR, NMI, RESET, and SMI#.
After entering the Halt state, all signals driven by the processor
retain their state as they existed following the completion of
the Halt special cycle.
Exit Halt State The AMD-K6-2E processor remains in the Halt state until it
samples INIT, INTR (if interrupts are enabled), NMI, RESET, or
SMI# asserted. If any of these signals is sampled asserted, the
processor returns to the Normal state and performs the
corresponding operation. All of the normal requirements for
recognition of these input signals apply within the Halt state.
250 Clock Control Chapter 13
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
13.3 Stop Grant State
Enter Stop Grant
State After recognizing the assertion of STPCLK#, the AMD-K6-2E
processor flushes its instruction pipelines, completes all
pending and in-progress bus cycles, and acknowledges the
STPCLK# assertion by executing a Stop Grant special bus cycle.
After BRDY# is sampled asserted during this cycle, and after
EWBE# is also sampled asserted (if not masked off), the
processor enters the Stop Grant state.
The Stop Grant state is like the Halt state in that the processor
disables most of its internal clock distribution in the Stop Grant
state.
In order to support the following operations, the internal PLL
still runs, and some internal resources are still clocked in the
Stop Grant state:
■Inquire cycles—The processor transitions to the Stop Grant
Inquire state during an inquire cycle. After returning to the
Stop Grant state following the inquire cycle, the processor
does not execute another Stop Grant special cycle.
■Time Stamp Counter (TSC)—The TSC continues to count in
the Stop Grant state.
■Signal Sampling—The processor continues to sample INIT,
INTR, NMI, RESET, and SMI#.
FLUSH# is not recognized in the Stop Grant state (unlike while
in the Halt state).
Upon entering the Stop Grant state, all signals driven by the
processor retain their state as they existed following the
completion of the Stop Grant special cycle.
Exit Stop Grant State The AMD-K6-2E processor remains in the Stop Grant state until
it samples STPCLK# negated or RESET asserted. If STPCLK#
is sampled negated, the processor returns to the Normal state in
less than 10 bus clock (CLK) periods. After the transition to the
Normal state, the processor resumes execution at the
instruction boundary on which STPCLK# was initially
recognized.
If STPCLK# is recognized as negated in the Stop Grant state
and subsequently sampled asserted prior to returning to the
Chapter 13 Clock Control 251
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Normal state, a minimum of one instruction is executed prior to
re-entering the Stop Grant state.
If INIT, INTR (if interrupts are enabled), FLUSH#, NMI, or
SMI# are sampled asserted in the Stop Grant state, the
processor latches the edge-sensitive signals (INIT, FLUSH#,
NMI, and SMI#), but otherwise does not exit the Stop Grant
state to service the interrupt. When the processor returns to the
Normal state due to sampling STPCLK# negated, any pending
interrupts are recognized after returning to the Normal state.
To ensure their recognition, all of the normal requirements for
these input signals apply within the Stop Grant state.
If RESET is sampled asserted in the Stop Grant state, the
processor immediately returns to the Normal state and the
reset process begins.
13.4 Stop Grant Inquire State
Enter Stop Grant
Inquire State The Stop Grant Inquire state is entered from the Stop Grant
state or the Halt state when EADS# is sampled asserted during
an inquire cycle initiated by the system logic. The AMD-K6-2E
processor responds to an inquire cycle in the same manner as in
the Normal state by driving HIT# and HITM#. If the inquire
cycle hits a modified data cache line, the processor performs a
writeback cycle.
Exit Stop Grant
Inquire State Following the completion of any writeback, the processor
returns to the state from which it entered the Stop Grant
Inquire state.
252 Clock Control Chapter 13
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
13.5 Stop Clock State
Enter Stop Clock
State If the CLK signal is stopped while the AMD-K6-2E processor is
in the Stop Grant state, the processor enters the Stop Clock
state. Because all internal clocks and the PLL are not running
in the Stop Clock state, the Stop Clock state represents the
minimum-power state of all clock control states. The CLK signal
must be held Low while it is stopped.
The Stop Clock state cannot be entered from the Halt state.
INTR is the only input signal that is allowed to change states
while the processor is in the Stop Clock state. However, INTR is
not sampled until the processor returns to the Stop Grant state.
All other input signals must remain unchanged in the Stop
Clock state.
Exit Stop Clock State The AMD-K6-2E processor returns to the Stop Grant state from
the Stop Clock state after the CLK signal is started and the
internal PLL has stabilized. PLL stabilization is achieved after
the CLK signal has been running within its specification for a
minimum of 1.0 ms.
The frequency of CLK when exiting the Stop Clock state can be
different than the frequency of CLK when entering the Stop
Clock state.
The state of the BF[2:0] signals when exiting the Stop Clock
state is ignored because the BF[2:0] signals are only sampled
during the falling transition of RESET.
Chapter 14 Electrical Data 253
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
14 Electrical Data
This chapter includes specifications for the operating ranges,
absolute ratings, and DC characteristics of the AMD-K6-2E
embedded processor. Typical and maximum power dissipation
values for the AMD-K6-2E processor during normal and
reduced power states are listed, as are example power derating
values based on lower CPU frequencies. The derating data may
be of special interest to embedded customers who want to use
AMD’s standard- or low-power devices at lower CPU
frequencies. The chapter concludes with a discussion of power
and grounding requirements and I/O buffer characteristics.
254 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.1 Operating Ranges
The AMD-K6-2E processor is designed to provide functional
operation if the voltage and temperature parameters are within
the limits defined in Table 49.
Table 49. Operating Ranges
Parameter Parameter Description Minimum Typical Maximum
VCC21
Notes:
1. VCC2 and VCC3 are referenced from VSS.
Core Supply Voltage—Low Power2
2. VCC2 specification for 1.9-V component.
1.8 V 1.9 V 2.0 V
Core Supply Voltage—Standard Power3
3. VCC2 specification for 2.2-V component.
2.1 V 2.2 V 2.3 V
VCC31I/O Supply Voltage—Standard and Low Power 3.135 V 3.3 V 3.6 V
TCASE Case Temperature—Low Power4
4. Case temperature range required for AMD-K6-2E/xxxAMZ valid ordering part number combinations, where xxx represents the
processor core frequency.
0C– 85
C
Case Temperature—Standard Power5
5. Case temperature range required for AMD-K6-2E/xxxAFR valid ordering part number combinations, where xxx represents the
processor core frequency.
0C– 70
C
Chapter 14 Electrical Data 255
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
14.2 Absolute Ratings
The AMD-K6-2E processor is not designed to be operated
beyond the operating ranges listed in Table 49. Exposure to
conditions outside these operating ranges for extended periods
of time can affect long-term reliability. Permanent damage can
occur if the absolute ratings listed in Table 50 are exceeded.
Note: If the AMD-K6-2E processor shows a “7” after the date code,
refer to the numbers in the last (rightmost) column of Table
50. The AMD-K6®-2 Revision Guide (order #21641)
available on AMD’s web site contains package marking
details, including the location of the date code
Table 50. Absolute Ratings
Parameter Minimum
Maximum for OPN
Suffixes: 233AFR,
233AMZ, 266AFR,
266AMZ, 300AFR1
Notes:
1. The data in this column applies to OPN suffixes 233AFR, 233AMZ, 266AFR, 266AMZ, and
300AFR, provided that the processor is not marked with “7” following the date code (i.e., is
blank).
Maximum for All OPNs2
2. The data in this column applies to all OPNs listed in Table 73, “Valid Ordering Part Number
Combinations,” on page 306 (including 233AFR, 233AMZ, 266AFR, 266AMZ, and 300AFR
when the processor is marked with a “7” following the date code).
VCC2 –0.5 V 2.6 V 2.4 V
VCC3 –0.5 V 3.6 V 3.6 V
VPIN3
3. VPIN (the voltage on any I/O pin) must not be greater than 0.5 V above the voltage being
applied to VCC3. In addition, the VPIN voltage must never exceed 4.0 V.
–0.5 V VCC3 + 0.5 V and < 4.0 V VCC3 + 0.5 V and < 4.0 V
TCASE (under bias) –65C +110C +110C
TSTORAGE –65C +150C +150C
256 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.3 DC Characteristics
The DC characteristics of the AMD-K6-2E processor are shown
in Table 51.
Table 51. DC Characteristics
Symbol Parameter Description Min Max Comments
VIL Input Low Voltage –0.3 V +0.8 V
VIH1 Input High Voltage 2.0 V VCC3+0.3 V
VOL Output Low Voltage 0.4 V IOL = 4.0-mA load
VOH Output High Voltage 2.4 V IOH = 3.0-mA load
ICC2
Low Power 1.9 V Power Supply Current2
4.75 A 233 MHz2,3
5.35 A 266 MHz2,3
5.50 A 300 MHz2,3,5
5.65 A 333 MHz2,3,4
6.25 A 350 MHz2,5
ICC2
Standard
Power
2.2 V Power Supply Current6
6.50 A 233 MHz3,6
7.35 A 266 MHz3,6
8.45 A 300 MHz3,5,6
9.40 A 333 MHz3,4,6
9.85 A 350 MHz5,6
10.00 A 400 MHz3,5,6
ICC3
Standard
and
Low Power
3.3 V Power Supply Current7
0.52 A 233 MHz3,7
0.54 A 266 MHz3,7
0.56 A 300 MHz3,5,7
0.58 A 333 MHz3,4,7
0.60 A 350 MHz5,7
0.62 A 400 MHz3,5,7
ILI8Input Leakage Current 15 mA
ILO8Output Leakage Current 15 mA
IIL9Input Leakage Current Bias with Pullup –400 mA
Chapter 14 Electrical Data 257
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
IIH10 Input Leakage Current Bias with Pulldown 200 mA
CIN Input Capacitance 10 pF
COUT Output Capacitance 15 pF
COUT I/O Capacitance 20 pF
CCLK CLK Capacitance 10 pF
CTIN Test Input Capacitance (TDI, TMS, TRST#) 10 pF
CTOUT Test Output Capacitance (TDO) 15 pF
CTCK TCK Capacitance 10 pF
Notes:
1. VCC3 refers to the voltage being applied to VCC3 during functional operation.
2. VCC2=2.0 V —The maximum power supply current must be taken into account when designing a power supply.
3. This specification applies to components using a CLK frequency of 66 MHz.
4. This specification applies to components using a CLK frequency of 95 MHz.
5. This specification applies to components using a CLK frequency of 100 MHz.
6. VCC2=2.3 V —The maximum power supply current must be taken into account when designing a power supply.
7. VCC3=3.6 V —The maximum power supply current must be taken into account when designing a power supply.
8. Refers to inputs and I/O without an internal pullup resistor and 0 VIN VCC3.
9. Refers to inputs with an internal pullup and VIL=0.4 V.
10. Refers to inputs with an internal pulldown and VIH=2.4 V.
Table 51. DC Characteristics (continued)
Symbol Parameter Description Min Max Comments
258 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.4 Power Dissipation
Table 52 and Table 53 list the typical and maximum power
dissipation of the AMD-K6-2E processor during normal and
reduced power states.
.
Table 52. Typical and Maximum Power Dissipation for OPN Suffix AMZ (Low-Power Devices)
Clock Control State 233 MHz1
Notes:
1. This specification applies to components using a CLK frequency of 66 MHz.
266 MHz1300 MHz1,3333 MHz1,2
2. This specification applies to components using a CLK frequency of 95 MHz.
350 MHz3
3. This specification applies to components using a CLK frequency of 100 MHz.
Thermal Power
(Maximum)4,5
4. The maximum power dissipated in the normal clock control state must be taken into account when designing a solution for
thermal dissipation for the AMD-K6-2E processor.
5. Maximum power is determined for the worst-case instruction sequence or function for the listed clock control states with
VCC2 = 1.9 V, and VCC3 = 3.3 V.
9.00 W 10.00 W 10.00 W 10.00 W 11.00 W
Thermal Power
(Typical)6
6. Typical power is determined for the typical instruction sequences or functions associated with normal system operation with
VCC2 = 1.9 V, and VCC3 = 3.3 V.
6.30 W 7.00 W 7.00 W 7.00 W 7.70 W
Stop Grant/Halt
(Maximum)7
7. The CLK signal and the internal PLL are still running but most internal clocking has stopped.
1.20 W 1.20 W 1.20 W 1.20 W 1.20 W
Stop Clock
(Maximum)8
8. The CLK signal, the internal PLL, and all internal clocking has stopped.
1.00 W 1.00 W 1.00 W 1.00 W 1.00 W
Chapter 14 Electrical Data 259
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 53. Typical and Maximum Power Dissipation for OPN Suffix AFR (Standard-Power Devices)
Clock Control State 233 MHz1
Notes:
1. This specification applies to components using a CLK frequency of 66 MHz.
266 MHz1300 MHz1,3 333 MHz1,2
2. This specification applies to components using a CLK frequency of 95 MHz.
350 MHz3
3. This specification applies to components using a CLK frequency of 100 MHz.
400 MHz1,3
Thermal Power
(Maximum)4,5
4. The maximum power dissipated in the normal clock control state must be taken into account when designing a solution for
thermal dissipation for the AMD-K6-2E processor.
5. Maximum power is determined for the worst-case instruction sequence or function for the listed clock control states with
VCC2 = 2.2 V, and VCC3 = 3.3 V.
13.50 W 14.70 W 17.20 W 19.00 W 19.95 W 16.90 W
Thermal Power
(Typical)6
6. Typical power is determined for the typical instruction sequences or functions associated with normal system operation with
VCC2 = 2.2 V, and VCC3 = 3.3 V.
8.10 8.85 W 10.35 W 11.40 W 11.98 W 10.15 W
Stop Grant/Halt
(Maximum)7
7. The CLK signal and the internal PLL are still running but most internal clocking has stopped.
2.46 W 2.48 W 2.50 W 3.94 W 3.96 W 4.40 W
Stop Clock
(Maximum)8
8. The CLK signal, the internal PLL, and all internal clocking has stopped.
2.25 W 2.25 2.25 W 3.50 W 3.50 W 4.00 W
260 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.5 Power Derating Based on Lower CPU Frequencies
This section provides standard- and low-power product
specification power derating based on lower CPU frequencies.
Note: The 166-MHz and 200-MHz data provided in Table 54 and
Table 55 was derived from the standard- and low-power
AMD-K6-2E/233AFR (and AMZ) and AMD-K6-2E/266AFR
(and AMZ) products and is solely for informational
purposes. The guaranteed electrical and thermal
specification data for the AMD-K6-2E/233AFR (and AMZ)
and AMD-K6-2E/266AFR (and AMZ) is also included here
for convenience. The derating data may be of interest to
customers who want to use AMD’s standard- or low-power
devices at lower CPU frequencies. AMD does not guarantee
the 166-MHz and 200-MHz CPU frequency derating data
and does not provide devices with these lower CPU
frequencies. Only devices listed in Table 73, “Valid Ordering
Part Number Combinations,” on page 306 are available.
Table 54. Power Derating Specification for Standard-Power Devices (AMD-K6-2E/233AFR and 266AFR)
Clock Control State 166.7 MHz1
Notes:
1. This specification applies to components using a clock and bus frequency of 66 MHz.
200.0 MHz1233.3 MHz1266.7 MHz1
Maximum ICC2 (core)
VCC2 = 2.2 V ± 100 mV2
2. The maximum ICC2 specification is taken at VCC2 = 2.3 V. (The maximum power supply current must be taken into account when
designing a power supply.)
5.31 A 6.00 A 6.50 A 7.35 A
Maximum ICC3 (I/O)
VCC3 = 3.3 V +300 mV, –165 mV3
3. The maximum ICC3 specification is taken at VCC3 = 3.6 V. (The maximum power supply current must be taken into account when
designing a power supply.)
0.48 A 0.50 A 0.52 A 0.54 A
Thermal Power (Maximum)4
4. Maximum thermal power is determined for the worst-case instruction sequence or functions for the listed clock control states
with VCC2 = 2.2 V and VCC3 = 3.3 V.
11.10 W 12.50 W 13.50 W 14.70 W
Thermal Power (Typical)5
5. Typical thermal power is determined for the typical instruction sequence or functions associated with normal system operation
with VCC2 = 2.2 V and VCC3 = 3.3 V.
6.65 W 7.50 W 8.10 W 8.85 W
Clock Multiple 2.5x 3.0x 3.5x 4.0x
BF[2:0] Inputs 100b 101b 111b 010b
Chapter 14 Electrical Data 261
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 55. Power Derating Specification for Low-Power Devices (AMD-K6-2E/233AMZ and 266AMZ)
Clock Control State 166.7 MHz1
Notes:
1. This specification applies to components using a clock and bus frequency of 66 MHz.
200.0 MHz1233.3 MHz1266.7 MHz1
Maximum ICC2 (core)
VCC2 = 1.9 V ± 100 mV2
2. The maximum ICC2 specification is taken at VCC2 = 2.0 V. (The maximum power supply current must be taken into account when
designing a power supply.)
3.79 A 4.37 A 4.75 A 5.35 A
Maximum ICC3 (I/O)
VCC3 = 3.3 V +300 mV, –165 mV3
3. The maximum ICC3 specification is taken at VCC3 = 3.6 V. (The maximum power supply current must be taken into account when
designing a power supply.)
0.48 A 0.50 A 0.52 A 0.54 A
Thermal Power (Maximum)4
4. Maximum thermal power is determined for the worst-case instruction sequence or functions for the listed clock control states
with VCC2 = 1.9 V and VCC3 = 3.3 V.
7.07 W 8.10 W 9.00 W 10.00 W
Thermal Power (Typical)5
5. Typical thermal power is determined for the typical instruction sequence or functions associated with normal system operation
with VCC2 = 1.9 V and VCC3 = 3.3 V.
4.95 W 5.67 W 6.30 W 7.00 W
Clock Multiple 2.5x 3.0x 3.5x 4.0x
BF[2:0] Inputs 100b 101b 111b 010b
262 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
14.6 Power and Grounding
Power Connections The AMD-K6-2E processor is a dual voltage device. Two
separate supply voltages are required: VCC2 and VCC3.
■VCC2 provides the core voltage for the processor.
■VCC3 provides the I/O voltage.
See “Electrical Data” on page 253 for the value and range of
VCC2 and VCC3.
There are 28 VCC2, 32 VCC3, and 68 VSS pins on the AMD-K6-2E
processor. (See Chapter 17, “Pin Designation Diagrams” on
page 299 for all power and ground pin designations.) The large
number of power and ground pins are provided to ensure that
the processor and package maintain a clean and stable power
distribution network.
For proper operation and functionality, all VCC2, VCC3, and VSS
pins must be connected to the appropriate planes in the circuit
board. The power planes have been arranged in a pattern to
simplify routing and minimize crosstalk on the circuit board.
The isolation region between two voltage planes must be at
least 0.254 mm if they are in the same layer of the circuit board.
(See Figure 88 on page 263.) To maintain low-impedance
current sink and reference, the ground plane must never be
split.
Although the AMD-K6-2E processor has two separate supply
voltages, there are no special power sequencing requirements.
The best procedure is to minimize the time between which VCC2
and VCC3 are either both on or both off.
Chapter 14 Electrical Data 263
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 88. Suggested Component Placement
Decoupling
Recommendations In addition to the isolation region mentioned in “Power
Connections” on page 262, adequate decoupling capacitance is
required between the two system power planes and the ground
plane to minimize ringing and to provide a low-impedance path
for return currents. Suggested decoupling capacitor placement
is shown in Figure 88.
Surface mounted capacitors should be used as close as possible
to the processor to minimize resistance and inductance in the
lead lengths while maintaining minimal height.
For recommendations about the specific value, quantity, and
location of the capacitors illustrated in Figure 88, see the AMD-
K6® Processor Power Supply Design Application Note, order
#21103.
0.254mm (min.) for
isolation region
VCC2 (Core) PlaneVCC3 (I/O) Plane
C1
CC5
CC3
C2 +
+
+
+
C5
C6
C7
C11
C12
C13
C17
C18
C19
C20
C21
C22
C23
C24
C25
C26
C27
C28
C29
C30
C31
CC4 +
CC6
CC10
CC1 CC2
CC9CC8
CC7
C8
C9
C10
C14
C15
C16
264 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Pin Connection
Requirements For proper operation, the following requirements for signal pin
connections must be met:
■Do not drive address and data signals into large capacitive
loads at high frequencies. If necessary, use buffer chips to
drive large capacitive loads.
■Leave all NC (no-connect) pins unconnected.
■Unused inputs should always be connected to an
appropriate signal level.
•Active Low inputs that are not being used should be
connected to VCC3 through a 20-kW pullup resistor.
•Active High inputs that are not being used should be
connected to GND through a pulldown resistor.
■Reserved signals can be treated in one of the following ways:
•As no-connect (NC) pins, in which case these pins are left
unconnected
•As pins connected to the system logic as defined by the
industry-standard Socket 7 and Super7 interfaces
•Any combination of NC and Socket 7 pins
■Keep trace lengths to a minimum.
14.7 I/O Buffer Characteristics
All of the AMD-K6-2E processor inputs, outputs, and
bidirectional buffers are implemented using a 3.3V buffer
design. AMD has developed a model that represents the
characteristics of the actual I/O buffer to allow system
designers to perform analog simulations of AMD-K6-2E
processor signals that interface with the system logic. Analog
simulations are used to determine a signal’s time of flight from
source to destination and to ensure that the system’s signal
quality requirements are met. Signal quality measurements
include overshoot, undershoot, slope reversal, and ringing.
I/O Buffer Model AMD provides a model of the AMD-K6-2E processor I/O buffer
for system designers to use in board-level simulations. This I/O
buffer model conforms to the I/O Buffer Information
Specification (IBIS). The I/O model contains voltage versus
current (V/I) and voltage versus time (V/T) data tables for
accurate modeling of I/O buffer behavior.
Chapter 14 Electrical Data 265
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
The following list characterizes the properties of the I/O buffer
model:
■All data tables contain minimum, typical, and maximum
values to allow for worst-case, typical, and best-case
simulations, respectively.
■The pullup, pulldown, power clamp, and ground clamp
device V/I tables contain enough data points to accurately
represent the nonlinear nature of the V/I curves. In addition,
the voltage ranges provided in these tables extend beyond
the normal operating range of the AMD-K6-2E processor for
those simulators that yield more accurate results based on
this wider range.
■The rising and falling ramp rates are specified.
■The min/typ/max VCC3 operating range is specified as
3.135V, 3.3V, and 3.6V, respectively.
■VIL = 0.8V, VIH = 2.0V, and VMEAS = 1.5V.
■The R/L/C of the package is modeled.
■The capacitance of the silicon die is modeled.
■The model assumes a test load resistance of 50 W.
I/O Model
Application Note For the AMD-K6-2E processor I/O Buffer IBIS Models and their
application, refer to the AMD-K6® Processor I/O Model (IBIS)
Application Note, order #21084.
I/O Buffer AC and DC
Characteristics See “Signal Switching Characteristics” on page 267 for the
AMD-K6-2E processor AC timing specifications.
Use this chapter for the AMD-K6-2E processor DC
specifications.
266 Electrical Data Chapter 14
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 15 Signal Switching Characteristics 267
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
15 Signal Switching Characteristics
The AMD-K6-2E processor signal switching characteristics are
presented in tables 56 through 65. Valid delay, float, setup, and
hold timing specifications are listed. These specifications are
provided for the system designer to determine if the timings
necessary for the processor to interface with the system logic
are met.
■Table 56 and Table 57 on page 268 contain the switching
characteristics of the CLK input.
■Table 58 through Table 61, beginning on page 270, contain
the timings for the normal operation signals.
■Table 62 on page 278 and Table 63 on page 279 contain the
timings for RESET and the configuration signals.
■Table 64 and Table 65 on page 280 contain the timings for
the test operation signals.
All signal timings provided are:
■Measured between CLK, TCK, or RESET at 1.5 V and the
corresponding signal at 1.5 V—this applies to input and
output signals that are switching from Low to High, or from
High to Low
■Based on input signals applied at a slew rate of 1 V/ns
between 0 V and 3 V (rising) and 3 V to 0 V (falling)
■Valid within the operating ranges given in “Operating
Ranges” on page 254
■Based on a load capacitance (CL) of 0 pF
15.1 CLK Switching Characteristics
Table 56 and Table 57 contain the switching characteristics of
the CLK input to the AMD-K6-2E processor for 100-MHz and
66-MHz bus operation, respectively, as measured at the voltage
levels indicated by Figure 89 on page 269.
The CLK Period Stability parameter specifies the variance
(jitter) allowed between successive periods of the CLK input
measured at 1.5 V. This parameter must be considered as one of
the elements of clock skew between the AMD-K6-2E and the
system logic.
268 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.2 Clock Switching Characteristics for 100-MHz Bus Operation
15.3 Clock Switching Characteristics for 66-MHz Bus Operation
Table 56. CLK Switching Characteristics for 100-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure Comments
Min Max
Frequency 33.3 MHz 100 MHz – In Normal Mode
t1CLK Period 10.0 ns – 89 In Normal Mode
t2CLK High Time 3.0 ns – 89 –
t3CLK Low Time 3.0 ns – 89 –
t4CLK Fall Time 0.15 ns 1.5 ns 89 –
t5CLK Rise Time 0.15 ns 1.5 ns 89 –
CLK Period Stability – 250 ps – Note
Notes:
The jitter frequency power spectrum peaking must occur at frequencies greater than (Frequency of CLK)/3 or less than 500 kHz.
Table 57. CLK Switching Characteristics for 66-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure Comments
Min Max
Frequency 33.3 MHz 66.6 MHz In Normal Mode
t1CLK Period 15.0 ns 30.0 ns 89 In Normal Mode
t2CLK High Time 4.0 ns 89
t3CLK Low Time 4.0 ns 89
t4CLK Fall Time 0.15 ns 1.5 ns 89
t5CLK Rise Time 0.15 ns 1.5 ns 89
CLK Period Stability 250 ps Note
Notes:
The jitter frequency power spectrum peaking must occur at frequencies greater than (Frequency of CLK)/3 or less than 500 KHz.
Chapter 15 Signal Switching Characteristics 269
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 89. CLK Waveform
15.4 Valid Delay, Float, Setup, and Hold Timings
Valid Delay and Float
Timing The maximum valid delay timings are provided to allow a
system designer to determine if setup times to the system logic
can be met. Likewise, the minimum valid delay timings are used
to analyze hold times to the system logic.
■Valid delay and float timings are given for output signals
during functional operation and are given relative to the
rising edge of CLK.
■During boundary-scan testing, valid delay and float timings
for output signals are with respect to the falling edge of
TCK.
Setup and Hold
Timing The setup and hold time requirements for the AMD-K6-2E
processor input signals must be met by the system logic to
assure the proper operation of the processor.
■The setup and hold timings during functional and
boundary-scan test mode are given relative to the rising
edge of CLK and TCK, respectively.
t5
2.0 V
1.5 V
0.8 V
t2
t3
t4
t1
270 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.5 Output Delay Timings for 100-MHz Bus Operation
Table 58. Output Delay Timings for 100-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t6A[31:3] Valid Delay 1.1 ns 4.0 ns 91
t7A[31:3] Float Delay 7.0 ns 92
t8ADS# Valid Delay 1.0 ns 4.0 ns 91
t9ADS# Float Delay 7.0 ns 92
t10 ADSC# Valid Delay 1.0 ns 4.0 ns 91
t11 ADSC# Float Delay 7.0 ns 92
t12 AP Valid Delay 1.0 ns 5.5 ns 91
t13 AP Float Delay 7.0 ns 92
t14 APCHK# Valid Delay 1.0 ns 4.5 ns 91
t15 BE[7:0]# Valid Delay 1.0 ns 4.0 ns 91
t16 BE[7:0]# Float Delay 7.0 ns 92
t17 BREQ Valid Delay 1.0 ns 4.0 ns 91
t18 CACHE# Valid Delay 1.0 ns 4.0 ns 91
t19 CACHE# Float Delay 7.0 ns 92
t20 D/C# Valid Delay 1.0 ns 4.0 ns 91
t21 D/C# Float Delay 7.0 ns 92
t22 D[63:0] Write Data Valid Delay 1.3 ns 4.5 ns 91
t23 D[63:0] Write Data Float Delay 7.0 ns 92
t24 DP[7:0] Write Data Valid Delay 1.3 ns 4.5 ns 91
t25 DP[7:0] Write Data Float Delay 7.0 ns 92
t26 FERR# Valid Delay 1.0 ns 4.5 ns 91
t27 HIT# Valid Delay 1.0 ns 4.0 ns 91
t28 HITM# Valid Delay 1.1 ns 4.0 ns 91
t29 HLDA Valid Delay 1.0 ns 4.0 ns 91
t30 LOCK# Valid Delay 1.1 ns 4.0 ns 91
t31 LOCK# Float Delay 7.0 ns 92
t32 M/IO# Valid Delay 1.0 ns 4.0 ns 91
t33 M/IO# Float Delay 7.0 ns 92
t34 PCD Valid Delay 1.0 ns 4.0 ns 91
Chapter 15 Signal Switching Characteristics 271
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
t35 PCD Float Delay 7.0 ns 92
t36 PCHK# Valid Delay 1.0 ns 4.5 ns 91
t37 PWT Valid Delay 1.0 ns 4.0 ns 91
t38 PWT Float Delay 7.0 ns 92
t39 SCYC Valid Delay 1.0 ns 4.0 ns 91
t40 SCYC Float Delay 7.0 ns 92
t41 SMIACT# Valid Delay 1.0 ns 4.0 ns 91
t42 W/R# Valid Delay 1.0 ns 4.0 ns 91
t43 W/R# Float Delay 7.0 ns 92
Table 58. Output Delay Timings for 100-MHz Bus Operation (continued)
Symbol Parameter Description Preliminary Data Figure
Min Max
272 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.6 Input Setup and Hold Timings for 100-MHz Bus Operation
Table 59. Input Setup and Hold Timings for 100-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t44 A[31:5] Setup Time 3.0 ns 93
t45 A[31:5] Hold Time 1.0 ns 93
t461A20M# Setup Time 3.0 ns 93
t471A20M# Hold Time 1.0 ns 93
t48 AHOLD Setup Time 3.5 ns 93
t49 AHOLD Hold Time 1.0 ns 93
t50 AP Setup Time 1.7 ns 93
t51 AP Hold Time 1.0 ns 93
t52 BOFF# Setup Time 3.5 ns 93
t53 BOFF# Hold Time 1.0 ns 93
t54 BRDY# Setup Time 3.0 ns 93
t55 BRDY# Hold Time 1.0 ns 93
t56 BRDYC# Setup Time 3.0 ns 93
t57 BRDYC# Hold Time 1.0 ns 93
t58 D[63:0] Read Data Setup Time 1.7 ns 93
t59 D[63:0] Read Data Hold Time 1.5 ns 93
t60 DP[7:0] Read Data Setup Time 1.7 ns 93
t61 DP[7:0] Read Data Hold Time 1.5 ns 93
t62 EADS# Setup Time 3.0 ns 93
t63 EADS# Hold Time 1.0 ns 93
t64 EWBE# Setup Time 1.7 ns 93
t65 EWBE# Hold Time 1.0 ns 93
t662FLUSH# Setup Time 1.7 ns 93
t672FLUSH# Hold Time 1.0 ns 93
t68 HOLD Setup Time 1.7 ns 93
t69 HOLD Hold Time 1.5 ns 93
t701IGNNE# Setup Time 1.7 ns 93
t711IGNNE# Hold Time 1.0 ns 93
Chapter 15 Signal Switching Characteristics 273
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
t722INIT Setup Time 1.7 ns 93
t732INIT Hold Time 1.0 ns 93
t741INTR Setup Time 1.7 ns 93
t751INTR Hold Time 1.0 ns 93
t76 INV Setup Time 1.7 ns 93
t77 INV Hold Time 1.0 ns 93
t78 KEN# Setup Time 3.0 ns 93
t79 KEN# Hold Time 1.0 ns 93
t80 NA# Setup Time 1.7 ns 93
t81 NA# Hold Time 1.0 ns 93
t822NMI Setup Time 1.7 ns 93
t832NMI Hold Time 1.0 ns 93
t842SMI# Setup Time 1.7 ns 93
t852SMI# Hold Time 1.0 ns 93
t861STPCLK# Setup Time 1.7 ns 93
t871STPCLK# Hold Time 1.0 ns 93
t88 WB/WT# Setup Time 1.7 ns 93
t89 WB/WT# Hold Time 1.0 ns 93
Notes:
1. These level-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must be asserted for a minimum pulse width of two clocks.
2. These edge-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must have been negated at least two clocks prior to assertion and
must remain asserted at least two clocks.
Table 59. Input Setup and Hold Timings for 100-MHz Bus Operation (continued)
Symbol Parameter Description Preliminary Data Figure
Min Max
274 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.7 Output Delay Timings for 66-MHz Bus Operation
Table 60. Output Delay Timings for 66-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t6A[31:3] Valid Delay 1.1 ns 6.3 ns 91
t7A[31:3] Float Delay 10.0 ns 92
t8ADS# Valid Delay 1.0 ns 6.0 ns 91
t9ADS# Float Delay 10.0 ns 92
t10 ADSC# Valid Delay 1.0 ns 7.0 ns 91
t11 ADSC# Float Delay 10.0 ns 92
t12 AP Valid Delay 1.0 ns 8.5 ns 91
t13 AP Float Delay 10.0 ns 92
t14 APCHK# Valid Delay 1.0 ns 8.3 ns 91
t15 BE[7:0]# Valid Delay 1.0 ns 7.0 ns 91
t16 BE[7:0]# Float Delay 10.0 ns 92
t17 BREQ Valid Delay 1.0 ns 8.0 ns 91
t18 CACHE# Valid Delay 1.0 ns 7.0 ns 91
t19 CACHE# Float Delay 10.0 ns 92
t20 D/C# Valid Delay 1.0 ns 7.0 ns 91
t21 D/C# Float Delay 10.0 ns 92
t22 D[63:0] Write Data Valid Delay 1.3 ns 7.5 ns 91
t23 D[63:0] Write Data Float Delay 10.0 ns 92
t24 DP[7:0] Write Data Valid Delay 1.3 ns 7.5 ns 91
t25 DP[7:0] Write Data Float Delay 10.0 ns 92
t26 FERR# Valid Delay 1.0 ns 8.3 ns 91
t27 HIT# Valid Delay 1.0 ns 6.8 ns 91
t28 HITM# Valid Delay 1.1 ns 6.0 ns 91
t29 HLDA Valid Delay 1.0 ns 6.8 ns 91
t30 LOCK# Valid Delay 1.1 ns 7.0 ns 91
t31 LOCK# Float Delay 10.0 ns 92
t32 M/IO# Valid Delay 1.0 ns 5.9 ns 91
t33 M/IO# Float Delay 10.0 ns 92
t34 PCD Valid Delay 1.0 ns 7.0 ns 91
Chapter 15 Signal Switching Characteristics 275
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
t35 PCD Float Delay 10.0 ns 92
t36 PCHK# Valid Delay 1.0 ns 7.0 ns 91
t37 PWT Valid Delay 1.0 ns 7.0 ns 91
t38 PWT Float Delay 10.0 ns 92
t39 SCYC Valid Delay 1.0 ns 7.0 ns 91
t40 SCYC Float Delay 10.0 ns 92
t41 SMIACT# Valid Delay 1.0 ns 7.3 ns 91
t42 W/R# Valid Delay 1.0 ns 7.0 ns 91
t43 W/R# Float Delay 10.0 ns 92
Table 60. Output Delay Timings for 66-MHz Bus Operation (continued)
Symbol Parameter Description Preliminary Data Figure
Min Max
276 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.8 Input Setup and Hold Timings for 66-MHz Bus Operation
Table 61. Input Setup and Hold Timings for 66-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t44 A[31:5] Setup Time 6.0 ns 93
t45 A[31:5] Hold Time 1.0 ns 93
t461A20M# Setup Time 5.0 ns 93
t471A20M# Hold Time 1.0 ns 93
t48 AHOLD Setup Time 5.5 ns 93
t49 AHOLD Hold Time 1.0 ns 93
t50 AP Setup Time 5.0 ns 93
t51 AP Hold Time 1.0 ns 93
t52 BOFF# Setup Time 5.5 ns 93
t53 BOFF# Hold Time 1.0 ns 93
t54 BRDY# Setup Time 5.0 ns 93
t55 BRDY# Hold Time 1.0 ns 93
t56 BRDYC# Setup Time 5.0 ns 93
t57 BRDYC# Hold Time 1.0 ns 93
t58 D[63:0] Read Data Setup Time 2.8 ns 93
t59 D[63:0] Read Data Hold Time 1.5 ns 93
t60 DP[7:0] Read Data Setup Time 2.8 ns 93
t61 DP[7:0] Read Data Hold Time 1.5 ns 93
t62 EADS# Setup Time 5.0 ns 93
t63 EADS# Hold Time 1.0 ns 93
t64 EWBE# Setup Time 5.0 ns 93
t65 EWBE# Hold Time 1.0 ns 93
t662FLUSH# Setup Time 5.0 ns 93
t672FLUSH# Hold Time 1.0 ns 93
t68 HOLD Setup Time 5.0 ns 93
t69 HOLD Hold Time 1.5 ns 93
t701IGNNE# Setup Time 5.0 ns 93
t711IGNNE# Hold Time 1.0 ns 93
Chapter 15 Signal Switching Characteristics 277
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
t722INIT Setup Time 5.0 ns 93
t732INIT Hold Time 1.0 ns 93
t741INTR Setup Time 5.0 ns 93
t751INTR Hold Time 1.0 ns 93
t76 INV Setup Time 5.0 ns 93
t77 INV Hold Time 1.0 ns 93
t78 KEN# Setup Time 5.0 ns 93
t79 KEN# Hold Time 1.0 ns 93
t80 NA# Setup Time 4.5 ns 93
t81 NA# Hold Time 1.0 ns 93
t822NMI Setup Time 5.0 ns 93
t832NMI Hold Time 1.0 ns 93
t842SMI# Setup Time 5.0 ns 93
t852SMI# Hold Time 1.0 ns 93
t861STPCLK# Setup Time 5.0 ns 93
t871STPCLK# Hold Time 1.0 ns 93
t88 WB/WT# Setup Time 4.5 ns 93
t89 WB/WT# Hold Time 1.0 ns 93
Notes:
1. These level-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must be asserted for a minimum pulse width of two clocks.
2. These edge-sensitive signals can be asserted synchronously or asynchronously. To be sampled on a specific clock edge, setup
and hold times must be met. If asserted asynchronously, they must have been negated at least two clocks prior to assertion and
must remain asserted at least two clocks.
Table 61. Input Setup and Hold Timings for 66-MHz Bus Operation (continued)
Symbol Parameter Description Preliminary Data Figure
Min Max
278 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
15.9 RESET and Test Signal Timing
Table 62. RESET and Configuration Signals for 100-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t90 RESET Setup Time 1.7 ns 94
t91 RESET Hold Time 1.0 ns 94
t92 RESET Pulse Width, VCC and CLK Stable 15 clocks 94
t93 RESET Active After VCC and CLK Stable 1.0 ms 94
t941
Notes:
1. BF[2:0] must meet a minimum setup time of 1.0 ms and a minimum hold time of two clocks relative to the negation of RESET.
BF[2:0] Setup Time 1.0 ms 94
t951BF[2:0] Hold Time 2 clocks 94
t96 intentionally left blank
t97 intentionally left blank
t98 intentionally left blank
t992
2. To be sampled on a specific clock edge, setup and hold times must be met the clock edge before the clock edge on which RESET
is sampled negated.
FLUSH# Setup Time 1.7 ns 94
t1003FLUSH# Hold Time 1.0 ns 94
t1013
3. If asserted asynchronously, these signals must meet a minimum setup and hold time of two clocks relative to the negation of
RESET.
FLUSH# Setup Time 2 clocks 94
t1023FLUSH# Hold Time 2 clocks 94
Chapter 15 Signal Switching Characteristics 279
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Table 63. RESET and Configuration Signals for 66-MHz Bus Operation
Symbol Parameter Description Preliminary Data Figure
Min Max
t90 RESET Setup Time 5.0 ns 94
t91 RESET Hold Time 1.0 ns 94
t92 RESET Pulse Width, VCC and CLK Stable 15 clocks 94
t93 RESET Active After VCC and CLK Stable 1.0 ms 94
t941
Notes:
1. BF[2:0] must meet a minimum setup time of 1.0 ms and a minimum hold time of two clocks relative to the negation of RESET.
BF[2:0] Setup Time 1.0 ms 94
t951BF[2:0] Hold Time 2 clocks 94
t96 intentionally left blank
t97 intentionally left blank
t98 intentionally left blank
t992
2. To be sampled on a specific clock edge, setup and hold times must be met the clock edge before the clock edge on which RESET
is sampled negated.
FLUSH# Setup Time 5.0 ns 94
t1002FLUSH# Hold Time 1.0 ns 94
t1013
3. If asserted asynchronously, these signals must meet a minimum setup and hold time of two clocks relative to the negation of
RESET.
FLUSH# Setup Time 2 clocks 94
t1023FLUSH# Hold Time 2 clocks 94
280 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 64. TCK Waveform and TRST# Timing at 25 MHz
Symbol Parameter Description Preliminary Data Figure
Min Max
TCK Frequency 25 MHz 95
t103 TCK Period 40.0 ns 95
t104 TCK High Time 14.0 ns 95
t105 TCK Low Time 14.0 ns 95
t1061,2
Notes:
1. Rise/Fall times can be increased by 1.0 ns for each 10 MHz that TCK is run below its maximum frequency of 25 MHz.
2. Rise/Fall times are measured between 0.8 V and 2.0 V.
TCK Fall Time 5.0 ns 95
t1071,2 TCK Rise Time 5.0 ns 95
t1083
3. Asynchronous.
TRST# Pulse Width 30.0 ns 96
Table 65. Test Signal Timing at 25 MHz
Symbol Parameter Description Preliminary Data Figure
Min Max
t1091
Notes:
1. Parameter is measured from the TCK rising edge.
TDI Setup Time 5.0 ns 97
t1101TDI Hold Time 9.0 ns 97
t1111TMS Setup Time 5.0 ns 97
t1121TMS Hold Time 9.0 ns 97
t1132
2. Parameter is measured from the TCK falling edge.
TDO Valid Delay 3.0 ns 13.0 ns 97
t1142TDO Float Delay 16.0 ns 97
t1152All Outputs (Non-Test) Valid Delay 3.0 ns 13.0 ns 97
t1162All Outputs (Non-Test) Float Delay 16.0 ns 97
t1171All Inputs (Non-Test) Setup Time 5.0 ns 97
t1181All Inputs (Non-Test) Hold Time 9.0 ns 97
Chapter 15 Signal Switching Characteristics 281
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
15.10 Timing Diagrams
Figure 90. Key to Timing Diagrams
Figure 91. Output Valid Delay Timing
Must be steady
Can change from
High to Low
Can change
from Low to High
(Does not apply)
Don’t care, any
change permitted
Steady
Changing from High to Low
Changing from Low to High
Changing, State Unknown
Center line is high
impedance state
WAVEFORM INPUTS OUTPUTS
Min
Max
Valid n +1
tv
Valid n
CLK
Output Signal
Tx
Tx
1.5 V
Note: For symbols tv listed in Table 58 on page 270 and Table 60 on p ag e 274, where:
v = 6, 8, 10, 12, 14, 15, 17, 18, 20, 22, 24, 26, 27, 28, 29, 30, 32, 34, 36, 37, 39, 41, 42
282 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 92. Maximum Float Delay Timing
Figure 93. Input Setup and Hold Timing
TxTxTx
Valid
Tx
tv
Min
Output Signal
tf
CLK 1.5 V
Note: For symbols tv and tf listed in T able 58 on page 270 and T able 60 on page 274, where:
v = 6, 8, 10, 12, 15, 18, 20, 22, 24, 30, 32, 34, 37, 39, 42
f = 7, 9, 11, 13, 16, 19, 21, 23, 25, 31, 33, 35, 38, 40, 43
CLK
TxTxTxTx
Input Signal
tsth
1.5 V
Note: For symbols ts and th listed in T able 59 on page 272 and T able 61 on page 276, where:
s = 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64, 66, 68, 70, 72, 74, 76, 78, 80, 82, 84, 86, 88
h = 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65, 67, 69, 71, 73, 75, 77, 79, 81, 83, 85, 87, 89
Chapter 15 Signal Switching Characteristics 283
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 94. Reset and Configuration Timing
Tx
CLK
RESET
Tx
t90
FLUSH#
(Synchronous)
1.5 V
1.5 V
1.5 V • • •
t92, 93
t91
t99 t100
• • •
BF[2:0]
(Asynchronous) t94
• • •
t95
FLUSH#
(Asynchronous) t101 t102
• • •
• • •
284 Signal Switching Characteristics Chapter 15
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 95. TCK Timing
Figure 96. TRST# Timing
Figure 97. Test Signal Timing
t107
2.0 V
1.5 V
0.8 V
t105
t106
t103
t104
1.5 V
t108
TCK
TDI, TMS
TDO
Output
Signals
Input
Signals
t103
t109, 111 t110, 112
t113
t115 t116
t117 t118
t114
1.5 V
Chapter 16 Thermal Design 285
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
16 Thermal Design
16.1 Package Thermal Specifications
The AMD-K6-2E processor operating specification calls for the
case temperature (TC) to be in the range of 0°C to 70°C (for the
standard-power 2.2-V component) or 0°C to 85°C (for the low-
power 1.9-V component). The ambient temperature (TA) is not
specified as long as the case temperature is not violated. The
case temperature must be measured on the top center of the
package. Table 66 and Table 67 show the package thermal
specifications for the AMD-K6-2E processor.
Table 66. Package Thermal Specification for OPN Suffix AMZ (Low-Power Devices)
qJC
Junction-Case
Maximum Thermal Power
233 MHz 266 MHz 300 MHz 333 MHz 350 MHz
1.0 °C/W 9.00 W 10.00 W 10.00 W 10.00 W 11.00 W
Stop Grant Mode 1.20 W 1.20 W 1.20 W 1.20 W 1.20 W
Stop Clock Mode 1.00 W 1.00 W 1.00 W 1.00 W 1.00 W
TC Case Temperature 0°C–85°C
Table 67. Package Thermal Specification for OPN Suffix AFR (Standard-Power Devices)
qJC
Junction-Case
Maximum Thermal Power
233 MHz 266 MHz 300 MHz 333 MHz 350 MHz 400 MHz
1.0 °C/W 13.50 W 14.70 W 17.20 W 19.00 W 19.95 W 16.90 W
Stop Grant Mode 2.46 W 2.48 W 2.50 W 3.94 W 3.96 W 4.40 W
Stop Clock Mode 2.25 W 2.25 W 2.25 W 3.50 W 3.50 W 4.00 W
TC Case Temperature 0°C–70°C
286 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 98 shows the thermal model of a processor with a passive
thermal solution. The ambient temperature (TA) is guaranteed
as long as TC is not violated. The case-to-ambient temperature
(TCA) can be calculated from the following equation:
TCA = PMAX • qCA
= PMAX • ( qIF + qSA)
Where:
PMAX = Maximum Power Consumption
qCA = Case-to-Ambient Thermal Resistance
qIF = Interface Material Thermal Resistance
qSA = Sink-to-Ambient Thermal Resistance
Figure 98. Thermal Model
Figure 99 illustrates the case-to-ambient temperature
(TCA=TC–T
A) in relation to the power consumption (X-axis)
and the thermal resistance (Y-axis). If the power consumption
and case temperature are known, the thermal resistance (qCA)
requirement can be calculated for a given ambient temperature
(TA) value.
Temperature Thermal
qSA qCA
qIF
(°C/W)(Ambient)
Case
Sink
TCA
Resistance
Chapter 16 Thermal Design 287
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 99. Power Consumption v. Thermal Resistance
The thermal resistance of a heatsink is determined by the heat
dissipation surface area, the material and shape of the
heatsink, and the airflow volume across the heatsink. In
general, the larger the surface area the lower the thermal
resistance.
The required thermal resistance of a heatsink (qSA) can be
calculated using the following example:
If:
TC = 70°C
TA = 45°C
PMAX = 19.95 W at 350 MHz (Standard-power AMD-K6-2E/350AFR)
Then:
: : : : : : : : : :
3RZHU&RQVXPSWLRQ:DWWV
GHJ&
GHJ&
GHJ&
GHJ&
&DVHWR$PELHQW7HPSHUDWXUH
7
&
7
$
7KHUPDO5HVLVWDQFH&:
qCA
TCTA
–
PMAX
--------------------
ÍÝ
ÌÜ
ËÛ
25C
19.95W
--------------------- 1.25 CW()
==
288 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Thermal grease is recommended as interface material because
it provides the lowest thermal resistance (@ 0.20°C/W). The
required thermal resistance (qSA) of the heatsink in this
example is calculated as follows:
qSA = qCA – qIF = 1.25 – 0.20 = 1.05(°C/W)
Heat Dissipation Path Figure 100 illustrates the heat dissipation path of the processor.
Due to the lower thermal resistance between the processor die
junction and case, most of the heat generated by the processor
is transferred from the top surface of the case. The small
amount of heat generated from the bottom side of the processor
where the processor socket blocks the convection can be safely
ignored.
Figure 100. Processor Heat Dissipation Path
16.2 Measuring Case Temperature
The processor case temperature is measured to ensure that the
thermal solution meets the processor’s operational
specification. This temperature should be measured on the top
center of the package, where most of the heat is dissipated.
Figure 101 on page 289 shows the correct location for measuring
the case temperature. The tip of the thermocouple should be
secured to the package surface with a small amount of
thermally conductive epoxy. A second location along the
thermocouple should also be secured to avoid any movement
during testing.
If a heatsink is installed while measuring, the thermocouple
must be installed into the heatsink via a small hole drilled
through the heatsink base (e.g., 1/16 of an inch). Secure the
thermocouple to the base of the heatsink by filling the small
hole with thermal epoxy, allowing the tip of the thermocouple
to protrude the epoxy and touch the top of the processor case.
Thin Lid
Case temperature
Ambient Temperature
Chapter 16 Thermal Design 289
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 101. Measuring Case Temperature
16.3 Sample Heatsink Measured Data
The example measured data (provided in tables 69 through 72
and figures 105 through 108) shows case-to-ambient thermal
resistance and maximum ambient temperature for both
socketed and soldered processors using three passive heatsink
samples with different height profiles (see Table 68) and
different levels of airflow in the system. This data is based on
the following specification assumptions:
TC = 85°C = TC(MAX) for 1.9-V low-power K6-2E processors
Pd = 10 W
qCA = qSA + qIF
TA(MAX) = TC(MAX) – (qCA ·Pd)
TA(MAX) = 85 - (qSA + qIF)·Pd
Note: AMD does not guarantee the example measured heatsink
data provided in tables 69 through 72 and figures 105
through 108. This data is supplied for informational
purposes only to facilitate the selection of an appropriate
thermal solution.
Example Heatsinks Table 68 describes the three heatsinks tested. Figure 102,
Figure 103, and Figure 104 show the actual heatsinks.
Thermocouple
Thermally Conductive Epoxy
Table 68. Passive Heatsink Samples
Identifier Size X (mm) Size Y (mm) Size Z Height (mm)
A52 50 15
B52 50 20
C50 50 30
290 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 102. Heatsink A (15 mm height)
Figure 103. Heatsink B (20 mm height)
Figure 104. Heatsink C (30 mm height)
Chapter 16 Thermal Design 291
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Socketed Installation:
Case-to-Ambient
Thermal Resistance
Table 69 and Figure 105 show the measured values for the case-
to-ambient thermal resistance (qCA in °C/watt) at different
levels of airflow in the system for the low-power embedded
AMD-K6-2E processors with a maximum thermal power
dissipation of 10.0 watts (OPNs: AMD-K6-2E/266AMZ,
AMD-K6-2E/300AMZ, and AMD-K6-2E/333AMZ) and a
maximum case temperature rating of 85°C, in the 321-pin
Ceramic Pin Grid Array package (CPGA) when installed into a
socket.
Figure 105. Measured Thermal Resistance v. Airflow (Socketed 321-Pin CPGA Package)
Table 69. Socketed CPGA Package: Measured Thermal Resistance (°C/W) qJC and qCA
Heatsink Type qJC (°C/W) qCA (°C/W) v. Airflow (m/s)
0.0 m/s 1.0 m/s 2.0 m/s 3.0 m/s 4.0 m/s
No Heatsink 1.09.17.56.14.74.0
A (15 mm) 1.0 6.5 4.3 3.1 2.6 2.3
B (20 mm) 1.0 5.7 3.8 2.7 2.3 2.0
C (30 mm) 1.0 4.7 2.8 1.8 1.5 1.3
0
1
2
3
4
5
6
7
8
9
10
01234
Airflow Rate (m/s)
q
ca
(°C/W)
None
A
B
C
292 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Socketed Installation:
Maximum Ambient
Temperature
Table 70 and Figure 106 show the measured maximum ambient
temperature (TA) values in °C at different levels of airflow in
the system for the low-power embedded AMD-K6-2E processors
with a maximum thermal power dissipation of 10.0 watts (for
OPNs: AMD-K6-2E/266AMZ, AMD-K6-2E/300AMZ, and
AMD-K6-2E/333AMZ) and a maximum case temperature rating
of 85 °C, in the 321-pin Ceramic Pin Grid Array package (CPGA)
when installed into a socket.
Figure 106. Measured Maximum Ambient Temperature (Socketed 321-Pin CPGA Package)
Table 70. Socketed CPGA Package: Measured Maximum Ambient Temperature (°C)
Heatsink Type
Maximum TA (°C) v. Airflow (m/s)
0.0 m/s 1.0 m/s 2.0 m/s 3.0 m/s 4.0 m/s
No Heatsink -10°C 6°C 21°C 36°C 43°C
A (15 mm) 20°C 42°C 54°C 59°C 62°C
B (20 mm) 28°C 47°C 58°C 62°C 65°C
C (30mm) 38°C 57°C 67°C 70°C 72°C
TC = 85°C , P d = 10 W
-20
-10
0
10
20
30
40
50
60
70
80
01234
Airflow Rate (m/s)
Maximum Ambien
t
Temperature (°C)
1RQH
$
%
&
Chapter 16 Thermal Design 293
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Soldered Installation:
Case-to-Ambient
Thermal Resistance
Table 71 and Figure 107 show the measured values for the case-
to-ambient thermal resistance (qCA in °C/watt) at different
levels of airflow in the system for the low-power embedded
AMD-K6-2E processors with a maximum thermal power
dissipation of 10.0 watts (OPNs: AMD-K6-2E/266AMZ,
AMD-K6-2E/300AMZ, and AMD-K6-2E/333AMZ) and a
maximum case temperature rating of 85°C, in the 321-pin
Ceramic Pin Grid Array package (CPGA) when soldered into a
printed circuit board.
.
Figure 107. Measured Thermal Resistance v. Airflow (Soldered 321-Pin CPGA Package)
Table 71. Soldered CPGA Package: Measured Thermal Resistance (°C/W) qJC and qCA
Heatsink Type qJC (°C/W)
qCA (°C/W) v. Airflow (m/s)
0.0 m/s 1.0 m/s 2.0 m/s 3.0 m/s 4.0 m/s
No Heatsink 1.0 7.0 5.2 4.2 3.3 2.7
A (15 mm) 1.0 5.2 3.4 2.4 1.9 1.7
B (20 mm) 1.0 4.9 2.9 2.0 1.5 1.3
C (30 mm) 1.0 4.3 2.5 1.6 1.4 1.2
0
1
2
3
4
5
6
7
8
01234
Airflow Rate (m/s)
qca (°C/W)
1RQH
$
%
&
294 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Soldered Installation:
Maximum Ambient
Temperature
Table 72 and Figure 108 show the measured maximum ambient
temperature (TA) values in °C at different levels of airflow in
the system for the low-power embedded AMD-K6-2E processors
with a maximum thermal power dissipation of 10.0 watts (for
OPNs: AMD-K6-2E/266AMZ, AMD-K6-2E/300AMZ, and
AMD-K6-2E/333AMZ) and a maximum case temperature rating
of 85 °C, in the 321-pin Ceramic Pin Grid Array package (CPGA)
when soldered into a printed circuit board.
.
Figure 108. Measured Maximum Ambient Temperature (Soldered, 321-Pin CPGA Package)
Table 72. Soldered CPGA Package: Measured Maximum Ambient Temperature (°C)
Heatsink Type
Maximum TA (°C) v. Airflow (m/s)
0.0 m/s 1.0 m/s 2.0 m/s 3.0 m/s 4.0 m/s
No Heatsink 14°C 31°C 42°C 53°C 58°C
A (15 mm) 33°C 51°C 61°C 66°C 68°C
B (20 mm) 36°C 56°C 65°C 70°C 72°C
C (30 mm) 42°C 60°C 69°C 71°C 73°C
TC = 85° C , P d = 10 W
0
10
20
30
40
50
60
70
80
01234
Airflow Rate (m/s)
Maximum Ambien
t
Temperature (°C)
1RQH
$
%
&
Chapter 16 Thermal Design 295
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
16.4 Layout and Airflow Considerations
Voltage Regulator A voltage regulator is required to support the lower voltage
(3.3 V and lower) to the processor. In most applications, the
voltage regulator is designed with power transistors. As a
result, additional heatsinks are required to dissipate the heat
from the power transistors. Figure 109 shows the voltage
regulator placed parallel to the processor with the airflow
aligned with the devices. With this alignment, the heat
generated by the voltage regulator has minimal effect on the
processor.
Figure 109. Voltage Regulator Placement
A heatsink and fan combination can deliver much better
thermal performance than a heatsink alone. More importantly,
with a fan/sink, the airflow requirements in a system design are
not as critical. A unidirectional heatsink with a fan moves air
from the top of the heatsink to the side. In this case, the best
location for the voltage regulator is on the side of the processor
in the path of the airflow exiting the fan sink (see Figure 110 on
page 296). This location guarantees that the heatsinks on both
the processor and the regulator receive adequate air
circulation.
Processor Airflow
Voltage Regulator
296 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 110. Airflow for a Heatsink with Fan
Airflow Management
in a System Design Complete airflow management in a system is important. In
addition to the volume of air, the path of the air is also
important. Figure 111 shows the airflow in a dual-fan system.
The fan in the front end pulls cool air into the system through
intake slots in the chassis. The power supply fan forces the hot
air out of the chassis. The thermal performance of the heatsink
can be maximized if it is located in the shaded area, where it
receives greatest benefit from this air exchange system.
Figure 111. Airflow Path in a Dual-Fan System
Airflow
Ideal areas for voltage regulator
Drive Bays
P/S
Vents
V
e
n
t
s
Fan
Fan
Main Board
Front
Chapter 16 Thermal Design 297
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Figure 112 shows the airflow management in a system using the
ATX form-factor. The orientation of the power supply fan and
the motherboard are modified in the ATX platform design. The
power supply fan pulls cool air through the chassis and across
the processor. The processor is located near the power supply
fan, where it can receive adequate airflow without an auxiliary
fan. The arrangement significantly improves the airflow across
the processor with minimum installation cost.
Figure 112. Airflow Path in an ATX Form-Factor System
For more information about thermal design considerations, see
the AMD-K6® Processor Thermal Solution Design Application
Note, order #21085.
P/S
Main Board
Drive Bays
F
a
n
298 Thermal Design Chapter 16
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 17 Pin Designation Diagrams 299
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
17 Pin Designation Diagrams
Figure 113. AMD-K6™-2E Processor Connection Diagram (Top-Side View CPGA)
300 Pin Designation Diagrams Chapter 17
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Figure 114. AMD-K6™-2E Processor Connection Diagram (Bottom-Side View CPGA)
Chapter 17 Pin Designation Diagrams 301
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
17.1 Pin Designations by Functional Grouping
Pin Name Pin Number Pin Name Pin Number Pin Name Pin Number Pin Name Pin Number
Control Address Data Data
A20M# AK-08 A3 AL-35 D0 K-34 D52 E-03
ADS# AJ-05 A4 AM-34 D1 G-35 D53 G-05
ADSC# AM-02 A5 AK-32 D2 J-35 D54 E-01
AHOLD V-04 A6 AN-33 D3 G-33 D55 G-03
APCHK# AE-05 A7 AL-33 D4 F-36 D56 H-04
BE0# AL-09 A8 AM-32 D5 F-34 D57 J-03
BE1# AK-10 A9 AK-30 D6 E-35 D58 J-05
BE2# AL-11 A10 AN-31 D7 E-33 D59 K-04
BE3# AK-12 A11 AL-31 D8 D-34 D60 L-05
BE4# AL-13 A12 AL-29 D9 C-37 D61 L-03
BE5# AK-14 A13 AK-28 D10 C-35 D62 M-04
BE6# AL-15 A14 AL-27 D11 B-36 D63 N-03
BE7# AK-16 A15 AK-26 D12 D-32 Test
BF0 Y-33 A16 AL-25 D13 B-34 TCK M-34
BF1 X-34 A17 AK-24 D14 C-33 TDI N-35
BF2 W-35 A18 AL-23 D15 A-35 TDO N-33
BOFF# Z-04 A19 AK-22 D16 B-32 TMS P-34
BRDY# X-04 A20 AL-21 D17 C-31 TRST# Q-33
BRDYC# Y-03 A21 AF-34 D18 A-33 Parity
BREQ AJ-01 A22 AH-36 D19 D-28 AP AK-02
CACHE# U-03 A23 AE-33 D20 B-30 DP0 D-36
CLK AK-18 A24 AG-35 D21 C-29 DP1 D-30
D/C# AK-04 A25 AJ-35 D22 A-31 DP2 C-25
EADS# AM-04 A26 AH-34 D23 D-26 DP3 D-18
EWBE# W-03 A27 AG-33 D24 C-27 DP4 C-07
FERR# Q-05 A28 AK-36 D25 C-23 DP5 F-06
FLUSH# AN-07 A29 AK-34 D26 D-24 DP6 F-02
HIT# AK-06 A30 AM-36 D27 C-21 DP7 N-05
HITM# AL-05 A31 AJ-33 D28 D-22
HLDA AJ-03 D29 C-19
HOLD AB-04 D30 D-20
IGNNE# AA-35 D31 C-17
INIT AA-33 D32 C-15
INTR AD-34 D33 D-16
INV U-05 D34 C-13
KEN# W-05 D35 D-14
LOCK# AH-04 D36 C-11
M/IO# T-04 D37 D-12
NA# Y-05 D38 C-09
NMI AC-33 D39 D-10
PCD AG-05 D40 D-08
PCHK# AF-04 D41 A-05
PWT AL-03 D42 E-09
RESET AK-20 D43 B-04
SCYC AL-17 D44 D-06
SMI# AB-34 D45 C-05
SMIACT# AG-03 D46 E-07
STPCLK# V-34 D47 C-03
VCC2DET AL-01 D48 D-04
VCC2H/L# AN-05 D49 E-05
W/R# AM-06 D50 D-02
WB/WT# AA-05 D51 F-04
302 Pin Designation Diagrams Chapter 17
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Pin Numbers
No Connect (NC) VCC2 V
CC3 VSS V
SS
A-37 A-07 A-19 A-03 AJ-27
E-17 A-09 A-21 B-06 AJ-31
E-25 A-11 A-23 B-08 AJ-37
R-34 A-13 A-25 B-10 AL-37
S-33 A-15 A-27 B-12 AM-08
S-35 A-17 A-29 B-14 AM-10
W-33 B-02 E-21 B-16 AM-12
AJ-15 E-15 E-27 B-18 AM-14
AJ-23 G-01 E-37 B-20 AM-16
AL-19 J-01 G-37 B-22 AM-18
AN-35 L-01 J-37 B-24 AM-20
Internal No Connect (INC) N-01 L-33 B-26 AM-22
C-01 Q-01 L-37 B-28 AM-24
H-34 S-01 N-37 E-11 AM-26
Y-35 U-01 Q-37 E-13 AM-28
Z-34 W-01 S-37 E-19 AM-30
AC-35 Y-01 T-34 E-23 AN-37
AL-07 AA-01 U-33 E-29
AN-01 AC-01 U-37 E-31
AN-03 AE-01 W-37 H-02
Reserved (RSVD) AG-01 Y-37 H-36
J-33 AJ-11 AA-37 K-02
L-35 AN-09 AC-37 K-36
P-04 AN-11 AE-37 M-02
Q-03 AN-13 AG-37 M-36
Q-35 AN-15 AJ-19 P-02
R-04 AN-17 AJ-29 P-36
S-03 AN-19 AN-21 R-02
S-05 AN-23 R-36
AA-03 AN-25 T-02
AC-03 AN-27 T-36
AC-05 AN-29 U-35
AD-04 V-02
AE-03 V-36
AE-35 X-02
Key X-36
AH-32 Z-02
Z-36
AB-02
AB-36
AD-02
AD-36
AF-02
AF-36
AH-02
AJ-07
AJ-09
AJ-13
AJ-17
AJ-21
AJ-25
Chapter 18 Package Specifications 303
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
18 Package Specifications
18.1 321-Pin Staggered CPGA Package Specification
Figure 115. 321-Pin Staggered CPGA Package Specification
.060
.090
16-038-CP-5_AC
CGF321
EU160
4.27.99 lv
.050
1.795
1.805
1.940
1.960
Index Corner
(45¡ Chamfer)
.064 MAX
.100
.120
.130
.017
.020
1.795
1.805
1.940
1.960 A
C
.005 F
.030
.010 MMM
MC
CAB
321 x
.100
.050
.115
.143
.051
.060
1.768
1.776
1.221
1.295
.060
.090
Index Corner
(45¡ Chamfer)
B
1.221
1.295
1.768
1.776
1.940
1.960
1.221
1.295
1.768
1.776
1.940
1.960 A
B
Lid
SIDE VIEW OF CGF PGA
SIDE VIEW OF CPGA
ALL MEASUREMENTS ARE IN INCHES UNLESS OTHERWISE NOTED.
304 Package Specifications Chapter 18
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Chapter 19 Ordering Information 305
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
19 Ordering Information
AMD standard- and low-power products are available in several operating ranges. The
ordering part number (OPN) is formed by a combination of the elements below. See
Table 73 on page 306 for valid ordering part number combinations.
AAMD-K6-2E/
Package Type
Family/Core
A = 321-pin Ceramic Pin Grid Array (CPGA)
AMD-K6-2E Embedded Processor
Case Temperature
R= 0°C–70°C
Z= 0°C–85°C
400
Performance Rating
/400 = 400 MHz
/350 = 350 MHz
/333 = 333 MHz
/300 = 300 MHz
/266 = 266 MHz
/233 = 233 MHz
Operating Voltage
F = 2.1 V–2.3 V (Core) / 3.135 V–3.6 V (I/O)
M = 1.8 V–2.0 V (Core) / 3.135 V–3.6 V (I/O)
F R
306 Ordering Information Chapter 19
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Table 73. Valid Ordering Part Number Combinations1
Notes:
1. This table lists configurations planned to be supported in volume for this device. Consult the local AMD sales office to confirm
availability of specific valid combinations and to check on newly-released combinations.
Device
Type OPN Package Type Operating Voltage Case
Temperature Maximum CPU/Bus
Frequency
Low
Power
AMD-K6-2E/350AMZ 321-pin CPGA 1.8V–2.0V (Core)
3.135V–3.6V (I/O)
0°C–85°C 350 MHz/100 MHz
AMD-K6-2E/333AMZ 321-pin CPGA 1.8V–2.0V (Core)
3.135V–3.6V (I/O)
0°C–85°C 333 MHz/95 MHz2
2. Also supports 66-MHz bus operation.
AMD-K6-2E/300AMZ 321-pin CPGA 1.8V–2.0V (Core)
3.135V–3.6V (I/O)
0°C–85°C 300 MHz/100 MHz2
AMD-K6-2E/266AMZ 321-pin CPGA 1.8V–2.0V (Core)
3.135V–3.6V (I/O)
0°C–85°C 266 MHz/66 MHz
AMD-K6-2E/233AMZ 321-pin CPGA 1.8V–2.0V (Core)
3.135V–3.6V (I/O)
0°C–85°C 233 MHz/66 MHz
Standard
Power
AMD-K6-2E/400AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 400 MHz/100 MHz2
AMD-K6-2E/350AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 350 MHz/100 MHz
AMD-K6-2E/333AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 333 MHz/95 MHz2
AMD-K6-2E/300AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 300 MHz/100 MHz2
AMD-K6-2E/266AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 266 MHz/66 MHz
AMD-K6-2E/233AFR 321-pin CPGA 2.1V–2.3V (Core)
3.135V–3.6V (I/O)
0°C–70°C 233 MHz/66 MHz
Index 307
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Index
Numerics
0.25-Micron Process Technology . . . . . . . . . . . . . . . . . . . . . . . 4
100-MHz Bus
clock switching characteristics . . . . . . . . . . . . . . . . . . . . 268
input setup and hold timings. . . . . . . . . . . . . . . . . . . . . . 272
output delay timings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
321-Pin Staggered CPGA
package specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
3DNow!™ Technology. . . . . . . . . . . . . . . . . . . . . . . . . .3, 15–20
data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
execution unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19–20
INIT state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
instruction compatibility, floating-point and. . . . . . . . . 216
instructions . . . . . . . . . . . . . . . . . . . . . . 13, 56, 81, 197, 216
PREFETCH instruction . . . . . . . . . . . . . . . . . . . . . . . . . . 197
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
RESET state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .118, 179
software prefetching. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4-Kbyte Paging Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4-Mbyte Paging Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 50
66-MHz Bus
clock switching characteristics . . . . . . . . . . . . . . . . . . . . 268
input setup and hold timings. . . . . . . . . . . . . . . . . . . . . . 276
output delay timings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
A
A[31:3] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A20M# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
masking cache accesses with. . . . . . . . . . . . . . . . . . . . . . 204
Absolute Ratings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Accelerated Graphic Port (AGP) . . . . . . . . . . . . . . . . . . . . . 4–5
Address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
bus . . . . . . . . . . . . . . . . . . . .87–92, 101, 154, 158, 160, 199
generation sequence during bursts. . . . . . . . . . . . . . . . . 142
hold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
parity check. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
ADS# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
ADSC# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
AGP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–5
AHOLD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
-initiated inquire hit to modified line. . . . . . . . . . . 158–159
-initiated inquire hit to shared or exclusive line . . 156–157
-initiated inquire miss . . . . . . . . . . . . . . . . . . . . . . . 154–155
restriction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160–161
Airflow
consideration, layout and. . . . . . . . . . . . . . . . . . . . . . . . . 295
heatsink with fan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
path in a dual-fan system . . . . . . . . . . . . . . . . . . . . . . . . . 296
path in an ATX form-factor system. . . . . . . . . . . . . . . . . 297
Allocate, Write . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
AMD-K6™-2E Processor
3DNow!™ technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
absolute ratings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
block diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
bus cycles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
cache organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
clock control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
connection diagrams . . . . . . . . . . . . . . . . . . . . . . . . 299–300
DC characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
decode logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
electrical data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
floating-point unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
logic symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
low-power devices . . . . . . . . . . . . . . . . . . 255–256, 261, 306
microarchitecture overview . . . . . . . . . . . . . . . . . . . . . . . . 7
multimedia execution unit . . . . . . . . . . . . . . . . . . . . . . . 213
operating ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
ordering information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
package specification . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
pin connection requirements . . . . . . . . . . . . . . . . . . . . . 264
pin designations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
power-on initialization. . . . . . . . . . . . . . . . . . . . . . . . . . . 179
process technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
scheduler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
signal descriptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
signal switching characteristics . . . . . . . . . . . . . . . . . . . 267
Socket 7 platform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
software environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
standard-power devices. . . . . . . . . . . . . . 255–256, 260, 306
Super7 platform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
System Management Mode (SMM) . . . . . . . . . . . . . . . . 217
test and debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
thermal design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
write merge buffer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
AP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
APCHK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Application Segment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Asserted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
B
Backoff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
BE[7:0]# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
BF[2:0]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 179
BIST. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
BOFF# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94, 162–163
locked operation with . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
Boundary-Scan
bit definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
register (BSR). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
test access port (TAP) . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
BR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Branch
execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
history table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . .1, 3, 11, 20–21
target cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
BRDY#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
BRDYC# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
BREQ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
BSR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Built-In Self-Test (BIST) . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
308 Index
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Burst
pipelined burst reads . . . . . . . . . . . . . . . . . . . . . . . . 142–143
reads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142–143
ready. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
ready copy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
writeback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
writeback due to cache-line replacement. . . . . . . . . . . . 145
Bus
address. . . . . . . . . . . . . . . . .89–92, 101, 154, 158, 160, 199
arbitration cycles, inquire and . . . . . . . . . . . . . . . . . . . . 148
backoff (BOFF#) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
order during misaligned I/O transfers . . . . . . . . . . . . 147
order during misaligned memory transfers . . . . . . . . 140
special . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
data. . . . 92, 95, 99–100, 116, 120, 136–138, 154, 160, 164
enables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
frequency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
hold request. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
lock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
request . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
states
address. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
data-NA# requested . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
idle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
pipeline address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
pipeline data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
state machine diagram. . . . . . . . . . . . . . . . . . . . . . . . . 135
transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
BYPASS Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Bypass Register. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
C
Cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
cacheable access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
coherency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
writeback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
writethrough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
disabling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
enable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
flushing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .104, 190
inhibit, L1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
instruction prefetch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
L1 . . . . . . . . . . . . . . . . . . . 42, 185, 192, 196, 199, 204, 227
-line fills. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
-line replacement . . . . . . . . . . . . . . . . . . . . . . . . . . .192, 201
masking cache accesses with A20M# . . . . . . . . . . . . . . . 204
MESI states in the data cache . . . . . . . . . . . . . . . . . . . . . 186
operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
organization. . . . . . . . . . . . . . . . . . . . . . . . . . . . 12, 185, 205
predecode bits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
prefetching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
related signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
sector organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
snooping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
write allocate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
write to cacheable page . . . . . . . . . . . . . . . . . . . . . . . . . . 193
writeback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 12, 204
writethrough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
CACHE#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .97, 189
Capture-DR state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Capture-IR state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Case Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285, 297
extended . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
measuring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288–289
Centralized Scheduler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
CLK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97, 269
switching characteristics. . . . . . . . . . . . . . . . . . . . . . . . . 267
100-MHz bus operation . . . . . . . . . . . . . . . . . . . . . . . . 268
66-MHz bus operation . . . . . . . . . . . . . . . . . . . . . . . . . 268
Clock Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97, 247
states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
halt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
stop clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173, 252
stop grant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173, 250
stop grant inquire . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
transitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
switching characteristics
100-MHz bus operation . . . . . . . . . . . . . . . . . . . . . . . . 268
66-MHz bus operation . . . . . . . . . . . . . . . . . . . . . . . . . 268
Coherency
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
writeback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
writethrough. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Component Placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Configuration
power-on initialization. . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Control Register 0 (CR0) . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Control Register 1 (CR1) . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Control Register 2 (CR2) . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Control Register 3 (CR3) . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Control Register 4 (CR4) . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Counter, Time Stamp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Customer Service. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii
Cycles
bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
hold and hold acknowledge . . . . . . . . . . . . . . . . . . . . . . 148
inquire. . . . 86–91, 101, 105–106, 122, 129, 144, 152, 154,
. . . 156–158, 160, 162, 166, 199, 202–204, 239, 247,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249–251
inquire and bus arbitration. . . . . . . . . . . . . . . . . . . . . . . 148
interrupt acknowledge . . . . . 87, 90, 92, 98, 114, 128, 132
locked . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
pipelined. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13, 88
pipelined write. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
shutdown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
special . . . . . . . . . . . . . . . . . . . . . . . .132, 170, 223, 249–250
writeback . .86, 88–89, 102, 105, 129, 144, 152, 156, 158,
. . . . . . . . . . . . 160, 162, 166, 188–189, 240, 248, 251
D
D/C#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
D[63:0]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
bus . . . . 92, 95, 99–100, 116, 120, 136–138, 154, 160, 164
cache, MESI states in the . . . . . . . . . . . . . . . . . . . . . . . . 186
parity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Data Types
3DNow!™ technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
floating-point register . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
integer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
MMX technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Index 309
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
Data/Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Data-NA# Requested . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
DC Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
in System Management Mode (SMM). . . . . . . . . . . . . . . 226
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .37, 241
DR3, DR2, DR1, and DR0 . . . . . . . . . . . . . . . . . . .39, 243
DR3–DR0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR5 and DR4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38, 242
DR5–DR4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38, 242, 244
DR7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37, 241, 244
Decoders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 9
Decoupling Recommendations . . . . . . . . . . . . . . . . . . . . . . 263
Derating, Power. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
Descriptors and Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Device Identification Register (DIR) . . . . . . . . . . . . . . . . . 234
Diagrams
key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
waveform definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
DIR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Disabling
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
DP[7:0] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
DR3–DR0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR5–DR4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
DR7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
LEN and RW definitions . . . . . . . . . . . . . . . . . . . . . . . . . 245
Driven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Dual Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
E
EADS# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
EAX Register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
EFER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40, 43, 182, 206
EFLAGS Register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Electrical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
absolute ratings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
DC characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
I/O buffer characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 264
operating ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
power and grounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
power derating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
power dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
Embedded Processor Features . . . . . . . . . . . . . . . . . . . . . . 1–2
EMMS Instruction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
EWBE#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .102, 205
EWBE# Control (EWBEC) . . . . . . . . . . . . . . . . . . . . . .205, 207
Exception. . . . . . . . 90–91, 100, 103, 116, 216, 226, 244–246
debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .225, 245
flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28–29
floating-point. . . . . . . . . . . . . . . . . . 103, 108, 213–214, 216
handler. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
machine-check. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
MMX technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
shutdown cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
System Management Mode (SMM). . . . . . . . . . . . . . . . . 226
Execution Latency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Execution Unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 18
3DNow!™ technology. . . . . . . . . . . . . . . . . . . . . . . . . . 19–20
floating-point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
multimedia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19–20, 215
register X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19–20
register Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19–20
Extended Feature Enable Register (EFER) .40, 43, 182, 206
External
address strobe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
write buffer empty. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
EXTEST Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
F
FEMMS Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
FERR#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103, 214, 216
Float Conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Float Delay Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Floated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Floating-Point
and MMX/3DNow!™ instruction compatibility . . . . . . 216
and multimedia execution units. . . . . . . . . . . . . . . . . . . 213
error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
execution unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
handling exceptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
instructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
register data types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
FLUSH# . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104, 179, 200, 228
FPU
control word register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
status word register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
tag word register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268, 280
multiplier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
operating. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 97, 179
G
Gate Descriptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52, 55
General-Purpose Registers . . . . . . . . . . . . . . . . . . . . . . . . . . 24
data types in. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
doubleword, word, and byte names . . . . . . . . . . . . . . . . . 25
Global EWBE# Disable (GEWBED) . . . . . . . . . . . . . . . . . . 206
H
Halt
restart slot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Hardware Prefetching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Heat Dissipation Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Heatsink
examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
sample measured data. . . . . . . . . . . . . . . . . . . . . . . . . . . 289
thermal resistance calculation . . . . . . . . . . . . . . . . . . . . 287
HIGHZ Instruction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Hit to
modified line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
modified line, AHOLD-initiated inquire. . . . . . . . . . . . 158
modified line, HOLD-initiated inquire . . . . . . . . . . . . . 152
shared or exclusive line, AHOLD-initiated inquire . . . 156
shared or exclusive line, HOLD-initiated inquire . . . . 150
HIT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
HITM#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
310 Index
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
HLDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
HOLD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
-initiated inquire hit to modified line. . . . . . . . . . . . . . . 152
-initiated inquire hit to shared or exclusive line . . . . . . 150
Hold
acknowledge . . . . . . . . . . . . . . . . . . . . . . . . . . .106, 148–150
timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .267, 282
I
I/O
buffer
AC and DC characteristics. . . . . . . . . . . . . . . . . . . . . . 265
characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
misaligned read and write . . . . . . . . . . . . . . . . . . . . . . . . 147
read and write . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
trap doubleword . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223–224
trap restart slot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
I/O Buffer Information Specification (IBIS). . . . . . . . . . . . 264
IBIS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
IDCODE Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Idle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
IEEE 1149.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
IEEE 754 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27, 213
IEEE 854 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
IGNNE# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108, 214, 216
Ignore Numeric Exception . . . . . . . . . . . . . . . . . . . . . . . . . . 108
INIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
-initiated transition from protected mode to real mode 176
state of processor after. . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
output signal state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
power-on configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . 179
processor state after INIT . . . . . . . . . . . . . . . . . . . . . . . . 183
processor state after RESET . . . . . . . . . . . . . . . . . . . . . . 180
register state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
RESET requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
signals sampled during RESET. . . . . . . . . . . . . . . . . . . . 179
Input
pin types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
setup and hold timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
setup and hold timings for 100-MHz bus operation. . . . 272
setup and hold timings for 66-MHz bus operation. . . . . 276
Inquire
and bus arbitration cycles . . . . . . . . . . . . . . . . . . . . . . . . 148
cycle hit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
cycle hit to modified line . . . . . . . . . . . . . . . . . . . . . . . . . 105
cycles 86–91, 101, 105–106, 122, 129, 144, 148, 151–158,
. . . . 160, 162, 166, 199, 202–204, 239, 247, 249–251
miss, AHOLD-initiated . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Instruction
buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
decode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
fetch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
pointer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Instructions
3DNow!™ technology . . . . . . . . . . . . . . . . . . . . . . . . .81, 215
compatibility of floating-point, MMX technology, and
3DNow!™. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
EMMS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
FEMMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
floating-point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
integer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
INVD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
MMX technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . .78, 215
PREFETCH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16, 197
RSM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
supported by the AMD-K6™-2E processor . . . . . . . . . . . 56
test access port (TAP) . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
WBINVD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Integer
data registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
instructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Interrupt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
01h. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
03h. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
10h. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
acknowledge. . . . . . . . . . . . . . . . . . . 95, 110, 112, 116, 164
acknowledge cycle definition . . . . . . . . . . . . . . . . . . . . . 168
acknowledge cycles . . . . . . . . 87, 90, 92, 98, 114, 128, 132
clock grant state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
descriptor table register . . . . . . . . . . . . . . . . . . . . . . . . . . 47
flag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34, 110, 121
floating-point exceptions . . . . . . . . . . . . . . . . . . . . 213–214
gate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
INIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
INTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
IRQ13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
MMX instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
NMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114, 183
redirection bitmap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
request . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
service routine . . . . . . . . . . . . . . . . . . . . .110, 114, 214, 217
shutdown cycle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
STPCLK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
system management . . . . . . . . . . . . . . . . . . . . 121, 217, 219
System Management Mode (SMM) . . . . . . . . . . . . . . . . 226
type of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
INTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
INV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Invalidation Request . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
INVD Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
K
KEN#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111, 192
L
L1 Cache . . . . . . . . . . . . . . . . . . . . 42, 185, 196, 199, 204, 227
inhibit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Layout and Airflow Considerations . . . . . . . . . . . . . . . . . . 295
Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii
LOCK#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Locked
cycles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
operation with BOFF# intervention. . . . . . . . . . . . 166–167
operation, basic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Logic
branch-prediction. . . . . . . . . . . . . . . . . . . . . . . . 3, 11, 20–21
external support of floating-point exceptions . . . . . . . 213
symbol diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Low-Power Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
valid ordering part numbers. . . . . . . . . . . . . . . . . . . . . . 306
Index 311
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
M
M/IO# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Machine-Check Address Register (MCAR) . . . . . .40–41, 182
Machine-Check Exception . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Machine-Check Type Register (MCTR) . . . . . . . . .40–41, 182
Maskable Interrupt. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
MCAR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40–41, 182
MCTR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40–41, 182
Memory
management registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
or I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
read and write, misaligned single-transfer . . . . . . . . . . 140
read and write, single-transfer . . . . . . . . . . . . . . . . . . . . 138
reads and writes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
type range registers (MTRR). . . . . . . . . . . . . . . . . . .45, 207
MESI. . . . . . . . . . . . . . . . . . . . . . . . . 1, 148, 152, 198, 202, 204
bit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12, 186, 188
states in the data cache . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Microarchitecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 7
branch prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
centralized scheduler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
decoders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
enhanced RISC86 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
execution units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
instruction fetching and decode . . . . . . . . . . . . . . . . . . . . 13
instruction prefetch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
predecode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Misaligned
I/O read and write. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
I/O transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
single-transfer memory read and write . . . . . . . . . 140–141
MMX Technology . . . . . . . . . . . . . . . . . . . . . 15–16, 18–20, 23
3DNow!™ registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
INIT state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
instruction compatibility, floating-point and. . . . . . . . . 216
instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56, 78, 216
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
RESET state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .118, 179
Model-Specific Registers (MSR) . . . . . . . . . . . . . . . . . . . . . . 40
MSR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
MTRR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45, 207
Multimedia
and 3DNow!™ execution units . . . . . . . . . . . . . . . . . . . . 215
execution unit . . . . . . . . . . . . . . . . . . . . . . . . . . . .19–20, 215
functional unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
N
NA#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Negated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Next Address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
NMI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
No-Connect Pins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Non-Maskable Interrupt. . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Non-Pipelined Single-Transfer Memory Read/Write and
Write Delayed by EWBE# . . . . . . . . . . . . . . . . . . . . 139
O
Operating Ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
OPN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Ordering Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Ordering Part Number (OPN). . . . . . . . . . . . . . . . . . . . . . . 305
Output
delay timings for 100-MHz bus operation . . . . . . . . . . . 270
delay timings for 66-MHz bus operation . . . . . . . . . . . . 274
pin float conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
signal state after RESET. . . . . . . . . . . . . . . . . . . . . . . . . 180
signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
valid delay timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
P
Package
321-pin staggered CPGA . . . . . . . . . . . . . . . . . . . . . . . . . 303
Socket 7 platform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Super7™ platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
thermal specification. . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Packed Decimal Data Register. . . . . . . . . . . . . . . . . . . . . . . 30
Page
cache disable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
directory entry (PDE) . . . . . . . . . . . . . . . . . . . . . 50–51, 188
flush/invalidate register (PFIR) . . . . .40, 46, 182, 200–201
table entry (PTE) . . . . . . . . . . . . . . . . . . . . . . . . . 50, 52, 188
writethrough. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Paging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Parity. . . . . . . . . . . . . . . . . . . . . . . . . 83, 90, 92, 100, 116, 138
bit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90, 100, 116
check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90–91, 100, 116
error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .91, 116, 154, 231
flags. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Part Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
PCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115, 188–189, 196
PCHK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
PFIR . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40, 46, 182, 200–201
Pins
connection requirements . . . . . . . . . . . . . . . . . . . . . . . . 264
designation diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
designations by functional grouping . . . . . . . . . . . . . . . 301
float conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Pipeline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136–137, 142
address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 10, 12, 19, 21
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
register X and Y. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Pipelined. . . . . . . . . . . . . . . . 19, 114, 137, 142–143, 160, 185
burst reads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
cycles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13, 88, 99
design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Pointer, Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Power
and grounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
consumption and thermal resistance. . . . . . . . . . . . . . . 287
derating based on lower CPU frequencies . . . . . . . . . . 260
dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
low-power devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
standard-power devices . . . . . . . . . . . . . . . . . . . . . . . 260
Power-on Configuration and Initialization . . . . . . . . . . . . 179
Precision Real Data Registers . . . . . . . . . . . . . . . . . . . . . . . 30
Predecode Bits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13, 187
PREFETCH Instruction . . . . . . . . . . . . . . . . . . . . . . . . 16, 197
312 Index
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Prefetching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
PREFETCH instruction . . . . . . . . . . . . . . . . . . . . . . . 16, 197
software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Processor
heat dissipation path . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
state observability register (PSOR) . . . . . . . . . 40, 45, 182
PSOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40, 45, 182
PWT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .117, 188
R
Read and Write
basic I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
misaligned I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Reads, Burst Reads and Pipelined Burst . . . . . . . . . . . . . . 142
Register X and Y
functional units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
pipelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10, 23, 44, 216
3DNow!™ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23, 31
boundary-scan (BSR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
bypass (BR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
control 0 (CR0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
control 1 (CR1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
control 2 (CR2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
control 3 (CR3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
control 4 (CR4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
data types, floating-point . . . . . . . . . . . . . . . . . . . . . . . . . . 30
debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .37, 241
descriptors and gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
device identification (DIR) . . . . . . . . . . . . . . . . . . . . . . . 234
DR3–DR0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR5–DR4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
DR6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
DR7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
EAX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
EFLAGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
extended feature enable (EFER) . . . . . . . . . . . . . . . . . . . 43
floating-point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
general-purpose. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
IR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
machine-check address (MCAR) . . . . . . . . . . . . . . . . . . . . 41
machine-check type (MCTR) . . . . . . . . . . . . . . . . . . . . . . . 41
memory management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
memory type range (MTRR) . . . . . . . . . . . . . . . . . . . . . . . 45
MMX technology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23, 31
model-specific (MSR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
packed decimal data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
PFIR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
precision real data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
PSOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
reset state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
segment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
SYSCALL/SYSRET target address (STAR) . . . . . . . . . . . 44
System Management Mode (SMM) initial state . . . . . . 219
test (TR12). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
test access port (TAP). . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
TR12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
UC/WC cacheability control (UWCCR) . . . . . . . . . . . . . . 45
write handling control (WHCR) . . . . . . . . . . . . . . . . . . . . 44
X and Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18–19
Regulator, Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Reserved . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
RESET. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118, 180
and configuration signals for 100-MHz bus operation . 278
and configuration signals for 66-MHz bus operation . . 279
and test signal timing . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
signals sampled during . . . . . . . . . . . . . . . . . . . . . . . . . . 179
state of processor after . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Reset and Configuration Timing . . . . . . . . . . . . . . . . . . . . 283
Return Address Stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
RISC86® Microarchitecture . . . . . . . . . . . . . . . . . . . . . . . . . . 8
RSM Instruction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222, 225
RSVD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
S
SAMPLE/PRELOAD Instruction . . . . . . . . . . . . . . . . . . . . 236
Sampled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Scheduler/Instruction Control Unit . . . . . . . . . . . . . . . . 10, 17
SCYC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Sector, Write to a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193, 197
Segment
descriptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26, 52–54
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
task state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Shift-DR state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Shift-IR state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Shutdown Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Signals
A[31:3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A20M#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86, 218
ADS# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
ADSC#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
AHOLD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89, 249
AP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
APCHK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
BE[7:0]# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
BF[2:0]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 252
BOFF# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94, 162, 249
BRDY#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95, 224, 249–250
BRDYC# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
BREQ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
CACHE# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97, 189
cache-related . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
CLK . . . . . . . . . . . . . . . . . . . . . . 97, 247, 250, 252, 267–268
D/C#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
D[63:0]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
DP[7:0] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
EADS#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101, 251
EWBE# . . . . . . . . . . . . . . . . . . . . . . . . . . . 102, 205, 249–250
FERR#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103, 216
FLUSH# . . . . . . . . . 104, 132, 179, 200, 226, 228, 249–251
HIT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105, 251
HITM#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105, 251
HLDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
HOLD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107, 249
IGNNE#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108, 216
INIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109, 218, 249–250
INTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110, 249–250
INV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
KEN#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
LOCK#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Index 313
22529B/0—January 2000 AMD-K6™-2E Processor Data Sheet
Preliminary Information
logic symbol. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
M/IO# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
NA#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
NMI. . . . . . . . . . . . . . . . . . . . . . . . . . 114, 218, 226, 249–250
output. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
PCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
PCHK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
pin connection requirements. . . . . . . . . . . . . . . . . . . . . . 264
pin designations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
PWT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
RESET . . . . . . . . . . . . . . . . . . . . . . . 118, 249–250, 252, 267
reset state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
RSVD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
sampled during RESET . . . . . . . . . . . . . . . . . . . . . . . . . . 179
SCYC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
SMI# . . . . . . . . . . . . . . . . . . . . . . . . . 121, 217, 224, 249–250
SMIACT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .122, 217
STPCLK# . . . . . . . . . . . . . . . . . . . . . . . . . 123, 247, 250–251
switching characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 267
TCK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .124, 267
TDI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
TDO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
test access port (TAP). . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
TMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
TRST# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
VCC2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
VCC2DET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
VCC2H/L# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
VCC3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
W/R#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
WB/WT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
SIMD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Single Instruction Multiple Data (SIMD) Operations. . . . . 11
Single-Transfer Memory Read and Write. . . . . . . . . . . . . . 138
SMI# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
SMIACT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Snoop . . . . . . . . . . . . . . . . . . . . . . 122, 129, 144, 200, 202–203
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
internal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Software Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
application segment types . . . . . . . . . . . . . . . . . . . . . . . . . 53
descriptors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
exceptions (summary) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
instructions supported . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
interrupts (summary) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
model-specific registers (MSR) . . . . . . . . . . . . . . . . . . . . . 41
paging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Software Prefetching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Special Bus Cycles . . .95, 102, 104, 123, 170–173, 190, 223,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249–250
definition (table). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
encodings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
summary (table) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Speculative EWBE# Disable (SEWBED) . . . . . . . . . . . . . . 206
Split Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Standard-Power Devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
valid ordering part numbers . . . . . . . . . . . . . . . . . . . . . . 306
State
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
machine diagram, bus. . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
processor
after INIT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
after RESET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Stop
clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
clock state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173, 252
grant inquire state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
grant state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173, 250
STPCLK# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Super7™ Platform
enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
initiative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Switching Characteristics
CLK switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
diagram key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
float delay timing diagram . . . . . . . . . . . . . . . . . . . . . . . 282
input setup and hold timing diagram . . . . . . . . . . . . . . 282
input setup and hold timings for 100-MHz bus. . . . . . . 272
input setup and hold timings for 66-MHz bus. . . . . . . . 276
output delay timings for 100-MHz bus. . . . . . . . . . . . . . 270
output delay timings for 66-MHz bus. . . . . . . . . . . . . . . 274
output valid delay timing diagram. . . . . . . . . . . . . . . . . 281
RESET and configuration signals for 100-MHz bus . . . 278
RESET and configuration signals for 66-MHz bus . . . . 279
RESET and configuration timing diagram . . . . . . . . . . 283
TCK timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
test signal timing . . . . . . . . . . . . . . . . . . . . . . . . . . . 280, 284
TRST# timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
valid delay, float, setup, and hold timings. . . . . . . . . . . 269
SYSCALL Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . 44, 73
SYSCALL/SYSRET Target Address Register (STAR) .40, 44,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
SYSRET Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44, 73
System
design
airflow management . . . . . . . . . . . . . . . . . . . . . . . . . . 296
component placement . . . . . . . . . . . . . . . . . . . . . . . . . 263
decoupling recommendations. . . . . . . . . . . . . . . . . . . 263
I/O buffer characteristics . . . . . . . . . . . . . . . . . . . . . . 264
pin connection requirements . . . . . . . . . . . . . . . . . . . 264
power connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
thermal considerations . . . . . . . . . . . . . . . . . . . . . . . . 285
segment and gate types. . . . . . . . . . . . . . . . . . . . . . . . . . . 54
segment descriptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
System Management Mode (SMM)
base address . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
debugging in. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
default register values. . . . . . . . . . . . . . . . . . . . . . . . . . . 217
exceptions in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
halt restart slot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
I/O trap doubleword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
I/O trap restart slot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
initial register state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
interrupt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
interrupt active . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
interrupts in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
operating mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
revision identifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
state-save area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
T
TAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Task State Segment (TSS). . . . . . . . . . . . . . . . . . . . . . . . . . . 48
TCK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124, 280, 284
TDI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
TDO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Technical Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii
314 Index
AMD-K6™-2E Processor Data Sheet 22529B/0—January 2000
Preliminary Information
Technical Support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
ambient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .285, 288
case-to-ambient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
extended rating for low-power devices. . . . . . . . . . . . . . 285
Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
boundary-scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
built-in self-test (BIST). . . . . . . . . . . . . . . . . . . . . . . . . . . 227
clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
data input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
data output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
L1 cache inhibit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
-logic-reset state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
mode select . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
pins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
register 12 (TR12). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
reset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
switching characteristics of signals. . . . . . . . . . . . . . . . . 280
test access port (TAP). . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
three-state test mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Test Access Port (TAP)
instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
BYPASS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
EXTEST. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
HIGHZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
IDCODE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
SAMPLE/PRELOAD. . . . . . . . . . . . . . . . . . . . . . . . . . . 236
registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
boundary-scan (BSR) . . . . . . . . . . . . . . . . . . . . . . . . . . 231
bypass (BR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
device identification (DIR) . . . . . . . . . . . . . . . . . . . . . 234
instruction register (IR). . . . . . . . . . . . . . . . . . . . . . . . 231
signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
states
capture-DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
capture-IR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
shift-DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
shift-IR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
state machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236–237
test-logic-reset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
update-DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
update-IR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Test Signal
timing at 25 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
timing diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
Thermal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .286, 295–296
design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
extended temperature rating. . . . . . . . . . . . . . . . . . . . . . 285
heat dissipation path . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
heatsink sample data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
layout and airflow consideration. . . . . . . . . . . . . . . . . . . 295
measuring case temperature . . . . . . . . . . . . . . . . . . . . . . 288
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Third-Party Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Three-State Test Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Time Stamp Counter (TSC) . . . . . . . . . . . . . . . . . . . . . . 42, 250
Timing Diagrams. . . . . . . . . . . . . . . . . . . . . . . . . .133, 139–177
waveform definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
TLB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
TMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
TR12 . . . . . . . . . . . . . . . . . . . . . . . . . 40, 42, 182, 188, 195, 239
Transition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
from protected mode to real mode . . . . . . . . . . . . . . . . . 176
Translation Lookaside Buffer (TLB) . . . . . . . . . . . . . . . . . . 185
TRST# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125, 280, 284
TSC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40, 42, 182, 249–250
TSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48, 54–55, 244
Typical and Maximum Power Dissipation . . . . . . . . . 258–259
low-power devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
standard-power devices. . . . . . . . . . . . . . . . . . . . . . . . . . 260
U
UC/WC Cacheability Control Register (UWCCR)40, 45, 180,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182, 191, 208
Uncacheable Memory . . . . . . . . . . . . . . . . . . . . . . 45, 206–207
UWCCR. . . . . . . . . . . . . . . . . . . . . . 40, 45, 180, 182, 191, 208
V
Valid
delay, float, setup, and hold timings . . . . . . . . . . . . . . . 269
masks and range sizes . . . . . . . . . . . . . . . . . . . . . . . . . . 210
ordering part number combinations . . . . . . . . . . . . . . . 306
VCC2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
VCC2DET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
VCC2H/L#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
VCC3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Voltage . . . . . . . . . . . . . . . . . . . . 126–127, 134, 254, 256, 264
dual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
W
W/R# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Waveform Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
WB/WT# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
WBINVD Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
WC/UC Memory Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
WCDE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
WHCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40, 44, 182, 196
Write
handling control register (WHCR). . . . . . . 40, 44, 182, 196
to a cacheable page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
to a sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193, 197
Write Allocate . . . . . . . . . . . . . . . . . . . . . . . 187, 192–193, 196
enable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
enable limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
limit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
logic mechanisms and conditions. . . . . . . . . . . . . . . . . . 196
Write Merge Buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
EWBE# Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
memory type range registers (MTRR). . . . . . . . . . . . . . 207
Write/Read . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Writeback . . . 97, 99–100, 111, 117, 122, 129, 132, 144–145,
. . . . . . . . . . . . . . . . . . . . . . . . 170, 185, 191, 198, 204
burst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 12
cycles. 86, 88–89, 102, 105, 129, 144, 152, 156, 158, 160,
. . . . . . . . . . . . . . . . 162, 166, 188–189, 240, 248, 251
or writethrough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Write-Combining Memory. . . . . . . . . . . . . . . . . . . 45, 206–207
Writethrough and Writeback Coherency States. . . . . . . . 204
