AMD 64-Bit K8 Platform Preview

Published on: October 17, 2000
Last Updated: October 17, 2000

AMD 64-Bit K8 Platform Preview

Published on: October 17, 2000
Last Updated: October 17, 2000

Are you seeking faster performance, more accuracy, higher precision, and support for large amounts of memory? Advanced Micro Devices plans to deliver these aspects, plus many other exciting new features with the upcoming 64-bit computer platform.

64-bit technology will soon reign in a new era of PC technology, and the AMD K8, better known as Hammer, may be positioned to take the crown.

Based on AMD’s x86-64 concept, the K8 will be a natural evolution of the original x86-32 K7 core design.

The K8 is a derivative of the K7, as each CPU core features many overlapping features. The K8 development process was similar to the K6-2.

The original K6 proved to be an effective CPU design, but AMD saw its possible potential, and decided to redevelop the core technology and expand the feature set for even greater performance and efficiency.

The same philosophy was used in building the K8 architecture. The original K7 has great potential, so AMD simply choose to enhance its capabilities with a wide array of new options and features.

Who Benefits From 64-Bit?

High performance servers, database management systems, CAD workstations, and even common desktops will benefit from a 64-bit design.

The K8 will allow extremely large memory amounts and offer newly optimized processing capabilities.

The K8 also expands the amount of available CPU registers, thus offering faster computational performance than current CPU designs do.

Processor intensive software functions, such as graphics transform and lighting, circuit simulation, and scientific processing will directly benefit from the x86-64 architecture.

Combined with native support for 16- and 32-bit software, the AMD K8 will prove to be a very competitive platform.

K8 Specifications
AMD x86-64 bit
Legacy Support x86 Mode: 16-bit and 32-bit Operating System Support
Compatibility Support Long Mode: Allows 64-bit Operating System to Execute 16- and 32-bit Software
64-bit Long Mode: Support for Advanced 64-bit Operations, Via x86-64 Extensions
64-bit Virtual Address Space
x86-64 Register Extensions: – Eight New General Purpose Registers – General Purpose Registers Widened to 64-bits – Eight New 128-bit Streaming SIMD Extension Registers
64-bit Instruction Pointer
Instruction Pointer Data Addressing Mode
Flat Addressing Model: Space For Single Code, Data, and Stack Information

Operating Modes

The x86-64 platform will offer two distinct modes of operation: long and legacy modes.

Long mode provides support for all 64-bit operations, and allows support for existing 16/32-bit application execution through a long mode compatibility option.

To operate in either form of long mode, the system must be utilizing a 64-bit compliant operating system.

By offering two levels of long mode support for all 16-, 32-, and 64-bit x86 applications, AMD is trying to position itself as the leader in performance, support, and compatibility.

Legacy mode provides backward compatible 16/32-bit support for more traditional operating environments, such as Windows98 and IBM OS/2.

Legacy mode supports traditional x86 real, virtual, and protected mode software. AMD is planning for full compatibility by offering complete support for current and future software while maintaining true cross platform 16, 32-, and 64-bit operation support.

Extended Register Extensions

The K8 will feature a 64-bit decode and execution subsystem, fully streamlined for maximum performance.

The addition of several new CPU-level registers will provide faster execution and improved efficiency.

A register is a fast memory location within the CPU used to create and store the results of CPU operations and other calculations.

The standard 32-bit x86 platform includes eight general-purpose registers (GPRs), but AMD has expanded this to sixteen with the K8.

Also added are eight 128-bit XMM/SSE registers, which provide more room for SIMD calculations, which are most often used in multimedia, graphics, and game programs.

All CPU registers allow 64-bit operations to be executed natively. By offering more register space, the K8 will be able to more efficiently process data, and move more information per clock cycle.

Some readers might ask, “Why not add more registers for even more performance?” AMD has conducted thorough studies comparing increased performance to the increased die size and cost.

The information they gathered indicates that sixteen GPRs prove to be the most efficient design while still minimizing consumer costs.

Adding an excessive number of registers does not lead to an exponential performance gain, and would only serve to increase the high production costs of an already advanced CPU design.

Single Instruction Multiple Data Support

Recent efforts by both AMD and Intel have positioned SIMD as a successor to the original x87 floating point unit design.

SIMD allows multiple floating-point operations to be combined and executed in only one CPU clock cycle.

The addition of eight XMM registers will speed SIMD performance to extreme levels. These new registers will provide 128-bit computation with double precision accuracy in both vector and scalar calculation modes.

Examples of FPU intensive software that could benefit include CAD, 3D modeling, vector analysis, and virtual reality simulation.

Two different base SIMD instruction sets are currently available, Intel’s SSE and AMD’s 3D-Now.

Each standard offers similar technology, but they are not cross compatible. Many software vendors are optimizing code for SIMD support, but the two competing standards are making this a slow and tedious task.

The K8 will resolve this issue in a creative manner. AMD has recently licensed the Intel original SSE and updated SSE-2 instruction sets, and will include support for both on the Hammer platform.

The K8 will also be able to execute 3D-Now, thus offering a wide range of compatibility.

Memory Addressing

Current x86-32:     4,294,967,296 Bytes (~4 Gigabytes)

New x86-64:     4,503,599,627,370,496 Bytes (~4,500 Terabytes)

The 64-bit platform allows for more than fast processing. Since the CPU can process 64-bit operations, it can also address 64-bits of memory address space.

Systems such as network and database servers require vast amounts of memory to effectively process the large amounts of archived information.

The current four gigabyte restrictions are limiting the efficiency of these large enterprise solutions.

The new x86-64 design allows for nearly 4.5 Petabytes, thus offering more available memory than any current or near future application will ever need.

AMD will be offering true 64-bit memory addressing within the K8 product line, but will limit the early production models to a 40-bit (48-bit virtual) memory allocation scheme.

Even with 40 bits, the memory limitation is about one terabyte. As much as memory technology is advancing and software demands are increasing, it will still be quite a few years until even the common desktop user will be measuring memory needs in the hundreds of gigabytes.

Lightning Data Transfer Bus

A very exciting feature offered by the K8 architecture is the Lightning Data Transfer (LDT) bus. AMD is attempting to offer a universal bus interconnect standard with LDT.

The LDT bus provides a high bandwidth connection between the system’s Northbridge controller to other bridge chips and controllers.

LDT allows 2- to 32-bit wide connections to be established between devices in a two stream unidirectional configuration.

LDT supports Microsoft’s Plug and Play initiative, so ease of compatibility is maintained.

Several uses for the LDT design are possible, with symmetric multiprocessing being the main driving force.

The LDT bus could provide a high-speed connection for multiple chipsets, each with multiple CPUs connected.

A multiple chipset architecture allows system designers to implement a massively parallel computing system within one structure.

Over 40 major hardware vendors have already endorsed the LDT interface and plan to support the standard when publicly released.

The K8, combined with the LDT bus, could offer the market a viable solution for x86 supercomputing.

LDT Goals
LDT I/O
LDT Multiprocessing
LDT Features

Operating System Support: x86-64 Linux

AMD recently entered a mutual development agreement with SuSE, Inc., a leader within the Linux OS community.

AMD’s plan to offer open source standards for the K8 has been welcomed within the Linux demographics.

SuSE has direct access to AMD’s latest technical data, and plans to have a complete 64-bit operating system available with the K8’s market release.

Currently dubbed x86-64 Linux, it doesn’t quite roll off the tongue, but having a readily available OS is a must when introducing a new system architectural design.

Along with underlying OS support, SuSE will also be offering 64-bit development tools for the K8 architecture.

Currently, a pre-alpha system-level simulator is being readied for shipment. The simulator will allow software programmers to develop 64-bit capable code while utilizing today’s 32-bit platforms.

AMD hopes many will support the initiative, as this would allow x86-64 compatible software to be available even before the K8’s official debut.

Also being developed is a compatible version of gcc, the well-known open source compiler.

Programming languages currently supported by gcc include C, C++, Objective C, Fortran, and even Java.

To further supplement gcc, SuSE will also offer a 64-bit ported version of Binutils. Binutils is a collection of binary development utilities that can increase the efficiency of software development.

Applications offered include an open source linker, a portable assembler, and binary file manipulation software.

Operating System Support: WinNT-64

Microsoft will be supporting the x86-64 architecture with their upcoming WinNT-64 operating system.

Porting NT to a 64-bit operating system shouldn’t prove a major challenge for MS, as their WinNT 4.0 OS already offers support for Compaq’s DEC Alpha platform.

Source information indicates an Itanium (Intel’s 64-bit architecture) version of NT will be available by year’s end, with a K8 compatible port following shortly after.

A 64-bit Windows platform will prove useful, especially for AMD’s deployment strategy. Many companies and end users will be seeking to upgrade their hardware to the 64-bit design, but may still rely on older 32-bit applications and utilities.

Since x86-64 offers native backward compatibility with existing 16/32-bit code, the K8 should be able to execute even the oldest Win9x and NT based programs.

As it will take many software vendors several months to make the evolutionary step to 64-bit, K8 owners will be prepared to make the transition smoothly, as current software can be utilized until the new 64-bit versions become available.

As with x86-64 Linux, expect Microsoft to offer their highly regarded Visual Studio programming interface optimized for the K8.

With support for several coding languages, such as Visual C++ and Visual Basic, a 64-bit version of Visual Studio will allow programmers to easily port existing 32-bit code to x86-64.

AMD again hopes to capture market support, as with Linux, by having a wide variety of software available for the K8 release, plus making porting existing code as effortless as possible.

The open source architecture of x86-64 will usher in a plethora of powerful, but widely available, software for consumer use.

Design/Marketing Strategy

Unlike the more traditional design methodology, AMD did not initially set a specific timetable for the K8 development and release.

They are seeking to offer the greatest range of compatibility with the best performance possible. Until these rigorous conditions were met, AMD did not plan to introduce a 64-bit platform.

Being an evolutionary extension of the original K7 core, the K8 is already building on a proven design.

Over two years of continuous research have led to the development of the x86-64 instruction set, which is at the heart of the K8 CPU core.

AMD once again has produced an award winning design without sacrificing the options needed for the future’s computing demands.

Officially, AMD is planning for a K8 CPU release in late 2001. Insiders seem to contradict this position, as many sources claim AMD could possibly start K8 sampling before year’s end.

Either way, AMD is rapidly closing in on Intel’s infamous Itanium. Intel has been plagued by core problems and poor fabrication yield rates with its 64-bit design.

If AMD could produce the K8 for widespread marketing before Itanium officially debuts, it could cause massive problems for giant Intel.

AMD is seeking complete open source standardization for the x86-64 architecture. By allowing their software and hardware partners full access to the K8 operating schematics, AMD is hoping for massive support of their newly emerging 64-bit platform.

All specifications will be open for viewing, ranging from software implementation to hardware design and reference.

Anyone seeking to support the x86-64 standard will be granted full use of any data AMD currently has available.

Having a substantial level of both hardware and software support, AMD’s K8 is positioned to offer serious market competition.

Sledgehammer: Server Class Computing

The AMD Slegdehammer will be the first incarnation of the powerful new K8 CPU core design. It will feature extreme performance and scalability.

AMD hopes to target server configurations first, then expand into other computing arenas.

Expect performance to be staggering, as the Sledgehammer will be built using a .13 micron copper interconnect technology.

The small die size and use of copper will allow this CPU to achieve 2,000+ megahertz speed levels.

The Sledgehammer will also be a highly scalable CPU, as the design allows multiple cores to be integrated into a single chip.

By offering single chip scalability, and then expanding this functionality with symmetric multiprocessing (via the LDT bus), the K8 Sledgehammer could effectively scale to super computer status.

AMD hopes to capture the high end market with the Sledgehammer, and offer new competition to high end market companies, such as Intel, IBM, Sun, and DEC Alpha.

Clawhammer: Desktop/Mobile Solution

Sometime after the initial release of the enterprise class Sledgehammer, AMD will release a more traditional market CPU, codenamed Clawhammer.

Clawhammer will feature the same x86-64 technology as its sibling, but will be scaled down for desktops and notebooks.

Little is known about the design, other than it will be marketed as a consumer CPU.

If Clawhammer features the same core as the Sledgehammer, expect powerful 64-bit desktop computing to arrive in the near future.

Competition/Comparision

AMD K-7AMD K8Intel P3Intel P4Intel
Itanium
General
Purpose
Registers
81688128
Floating
Point
Registers
8888128(4)
SIMD
Registers
81688128(4)
SIMD
Support
3dnow!3dnow,
SSE,
SSE2
SSESSE,
SSE2
SSE,
SSE2
x86-32
Support
NativeNativeNativeNativeTranslation
Market
Role
Desktop,
Mobile,
Server
Initially
Enterprise
Server
Desktop,
Mobile,
Server
Desktop
Server
Enterprise
Server
SMP
Support
SMP avail.
Q4 ’00
Full SMP,
Multiple
Cores/Chip
P3: 2-way,
Xeon: 8-way
Full SMP(1)Full SMP,
up to n-way(2)
Memory
Support
SDRAM,
DDR RAM
DDR RAMSDRAM,
RDRAM
SDRAM,
DDR RAM,
RDRAM
RDRAM
Bus
Support
PCI-33,
AGP 2/4x,
ISA
PCI-33/66,
PCI-X(3),
AGP 2/4/8x,
LDT, ISA
PCI-33,
AGP 2/4/8x,
ISA
PCI-33/66,
AGP 2/4/8x,
ISA
PCI-33/66,
PCI-X(3),
AGP 2/4/8x,
ISA
AvailabilityCurrentBy Q4 ’01CurrentQ4 ’00By Q2 ’01
(1)Not publicly announced, until now.
(2)n-way: Maximum SMP CPU support limited only by chipset.
(3)PCI-X: Newly developed expanded PCI format, more info to follow at later date.
(4)Itanium SIMD and FPU registers reside within main General Purpose Registers, per Intel IA-64 specifications.

Originally dubbed Merced, Intel’s new 64-bit Itanium will be in direct competition with AMD’s K8 design.

The Merced was Intel’s first 64-bit design to be produced, but poor yield rates and low clock speeds resulted in scrapping the project and moving to a new core design, the Itanium.

Production problems have not improved, however, as Intel is having a difficult time producing a design with a decent yield rate.

Assuming these fabrication problems can be resolved, expect to see the Itanium surface on the market in early 2001.

The Itanium is 64-bit native and will not be compatible with AMD’s competing x86-64 specification. Instead, Intel decided to implement a completely new architecture, known as IA-64.

IA-64 supports a wide range of high performance options, such as 128 general-purpose registers, 128 floating point registers, advanced branch prediction, and execution speculation.

Another feature is a three level cache architecture, featuring two high speed on-die caches, with a massive off-die cache of up to four megabytes.

The Itanium will also be able to execute current x86-32 code, but not in a native mode. It must first decode and restructure the 32-bit x86 instructions into data the core-processing unit can understand.

To perform the data manipulation needed, the Itanium must utilize several clock cycles, thus lowering the performance level.

It is speculated that the IA-64 platform will be slower than current x86-32 configurations when executing 32-bit code.

AMD could hold a serious advantage in this realm, as the K8 can process 32-bit instructions natively.

Click here for more information about the Itanium architecture.

Conclusion/Comments

By building the K8 around the concept of high performance and excellent compatibility, Advanced Micro Devices plans to leverage its x86-64 platform as the next computer technology evolution.

By offering an advanced core design with support for current and future software, the K8 could be the platform to usher in a new era of computing.

The next few months will be very interesting, especially with the recent problems Intel has encountered in the competing Itanium architecture.

Expect AMD to offer a very competitive product with great potential of becoming the market leader.

Stay on top of the latest technology trends — delivered directly to your inbox, free!

Subscription Form Posts

Don't worry, we don't spam

Written by Bobby

Bobby Lawson is a seasoned technology writer with over a decade of experience in the industry. He has written extensively on topics such as cybersecurity, cloud computing, and data analytics. His articles have been featured in several prominent publications, and he is known for his ability to distill complex technical concepts into easily digestible content.