AMD Aims To Take Over x86 Leadership

Pages: 1 2 3

AMD – Not Just An Imitator

In most industries the market leader would refer to the other companies participating in that market as competitors – but not Intel. It is so arrogant and self-focused that it publicly proclaims the x86 universe consists of itself and several “imitators”. Advanced Micro Devices (AMD) is one competitor that has long ago decided imitation was the road to oblivion and it needed to take its destiny into its own hands. As a result two years ago AMD made two surprising announcements.

The first announcement was that its future K7 processor would use the mechanical form factor of Intel’s slot 1 connector and cartridge. Instead of using the P6 bus (which it agreed never to do as part of a wider legal settlement with Intel) AMD announced it had licensed the bus technology of the next generation Alpha high end RISC processor, the 21264 or EV6. This was a very bold gamble. Though Alpha technology is widely viewed as leading edge, chipsets developed for the EV6 would likely be far too expensive for the PC market. AMD was basically committing to building an entire infrastructure of chipsets and motherboard reference designs for the K7 with or without initial support from third part chipset vendors.

The second surprise was AMD’s 3DNow extensions to the x86 instruction set architecture (ISA). Up to that point the x86 instruction set was basically considered the exclusive property of Intel. Any move to change or add to x86 would be tantamount to proclaiming incompatibility with Intel devices. Fortunately, the 3DNow extension were welcomed by a small but important niche of x86 customers, game developers and game players. More importantly in the long run, Microsoft blessed the 3D now extensions with support for them in the DirectX graphics application programming interface (API). It would be about a year before Intel could answer the 3DNow extensions delivered to customers in AMD’s K6-2 MPU with its own streaming SIMD extensions or SSE, first incorporated in the “Katmai” Pentium III.

To Boldly Go Where No x86 Has Gone Before

A few weeks ago AMD presented a few details about its next generation K8 processor code-named “SledgeHammer”. The most surprising disclosure was that it would extend the x86 ISA to 64 bits (x86-64). Intel long ago decided against extending x86 to 64 bits and instead, in conjunction with Hewlett Packard, developed a completely new, RISC-like 64-bit ISA, called IA-64. Although IA-64 has an x86 compatibility mode it will very likely operate at a much lower performance level than native 64-bit code. Also, with IA-64 it is very difficult and inefficient to mix x86 and IA-64 code within a single application thread. It certainly seems that Intel intends to wean its customers off the x86 platform in favour of IA-64 starting first at the high end (servers and workstations). This contrasts sharply with the AMD’s evolutionary approach which can support 32 and 64 bit x86 code nearly equally and apparently allow efficient mixing of code at the same level as can be done with 16 and 32 bit x86 code since the Intel 386.

If AMD succeeds in eliciting support for x86-64 from Microsoft then it will be quite obvious that it has taken over leadership of the x86 architecture. This is likely for several reasons. The biggest reason Microsoft would want to strongly support x86-64 is Compaq’s recent public announcement that it would not support development of a 64 bit version of Windows as a product offering for Alpha. Microsoft has great ambitions for Windows challenging the traditional proprietary versions of Unix for the role of running the largest and most prestigious computer systems, the high end, multiple processor Enterprise class server. But these systems are almost exclusively based on 64 bit RISC processors.

With support for both IA-64 and Alpha, Microsoft seemed to have the two most powerful 64-bit platforms in its stable. The loss of Alpha makes 64 bit Windows an operating system exclusively targeted for IA-64. This leaves Microsoft’s ambitions increasingly vulnerable to Intel, which has plans of its own. Intel is in the driver’s seat for IA-64 because it can point customers to Monterey, Solaris, HP-UX, AIX and Linux as alternatives to Microsoft’s 64 bit OS offering. As testimony has shown in the DOJ antitrust trial against Microsoft, there is very little love lost between these two ruthless and highly competitive giants.

This is where AMD’s x86-64 comes in. I wouldn’t be surprised at all if knowledge of AMD’s plan to extend the x86 ISA to 64 bits was a major factor in Compaq’s decision to pull Alpha out of the Windows arena and focus on TRU64 and Linux. In a two way competition for 64 bit Window platforms between IA-64 and Alpha, Compaq likely thought it could stay a ahead on native performance and match or beat IA-64’s x86 compatibility mode with its FX!32 binary recompilation tool, while offering smaller and cheaper processors and chipsets. In a three-way match, Alpha would likely be squeezed out on price and x86 performance by AMD’s x86-64 based processors.

Pages:   1 2 3  Next »

Be the first to discuss this article!