With the announcement that Intel is releasing a Xeon Processor with 64-bit extensions, perhaps the title of this column should be “Yamhill lives.” This is because Yamhill was a much rumored, but never confirmed Intel project to bridge the 32- and 64-bit world with a transition chip. Yamhill was always believed to be an Itanium (Intels from-scratch 64-bit CPU) killer, because it would be less expensive and have better 32-bit support. So Intel denied the existence of Yamhill—and most of us analysts believed it existed, but would never be released, because it would kill off Itanium. Given the billions invested in Itanium, wounding it with Yamhill seemed very unlikely.
But the market has changed. The emergence of a new Xeon supporting 64-bit extensions (likely to be followed in a year or so by a similarly styled Pentium) does not mean that Itanium has to die. The native 64-bit processor has matured over the past several years and now delivers better 64-bit performance than any transitional product could.
But for most buyers, Itanium is more of an academic than practical interest, as they simply dont buy this class of system—mostly high-end servers and workstations. These computers put massive loads on the CPU—and have already moved to 64-bits. Here the evolved Itanium architecture has an almost insurmountable advantage over any transition chip, and will survive handily—albeit at much lower volumes and higher costs than any transitional parts.
Itanium was supposed to be the transition part that moved Intels customers into 64-bit computing. But it was so late that the 32-bit performance wasnt fast enough to supplant the speediest Pentiums and Athlons. Thus it ended up as a 64-bit only platform. AMD spotted the opportunity, and developed the transitional product that Intel lacked. Thus AMD has enjoyed impressive design wins with companies like HP, IBM and Sun as a result of this first mover advantage.
AMDs success was also partially due to the creation of a group targeted at the enterprise and headed by ex-Gartner analyst Kevin Knox. Knox was able to open doors at large systems manufacturers that previously snubbed the company. Intel has applied considerable staff and money to the enterprise hardware market, but this is the first time, to my knowledge, that AMD stepped up to the challenge. It paid off very well for them.
As we move to a completely 64-bit desktop, well be able to step away from the old, and often unreliable and insecure applications we currently use. However, there will be some backwards compatibility problems to deal with.
In addition, though, a native 64-bit environment will lack a broad range of hardware drivers, which will make the migration more difficult 64-bit OSs need 64-bit drivers. Existing 32-bit drivers generally will not work, even in compatibility mode. And rewriting drivers takes time.
Hidden in Tuesdays announcement was the verification that Intel will be change its Pentium 4 in the second half of the year to enable all of the security capabilities in Microsofts Windows XP service patch (SP-2). Today only 64-bit processors canuse one critical feature that mitigates buffer overrun exploits—which have been a material part of Denial of Service (DOS) attacks.
AMD has a first mover advantage here, but the majority of the 64-bit opportunity is still to come—and Intel has positioned itself well for that future Intel could have pushed the problem off a year, but they would have continued to bleed share. Making the move now as opposed to later was a smart one. This battle will be fully joined in 2005 when all of the parts will be in place. The clock is ticking for AMD, which has but a year consolidate and improve its position. The easy battles are done—now the real war begins.