主题：【文摘】一篇观点比较独特的分析文章。 -- Highway

Ars Technica Newsdesk

Intel to demonstrate 64-bit x86 "CT" chip in two weeks

Posted 02/07/2004 @ 1:22 PM, by Ken "Caesar" Fisher

Late January it was revealed that Intel did in fact have a 64-bit x86 solution in the works. While Intel's admission was understandably cryptic, it was clear that the CPU had long since been in development, even if its existence has largely been hush hush outside of the rumor mills. Now it is expected that Intel will demonstrate the technology at the Intel Developer Forum, February 17-19. The new 64-bit extensions, formerly codenamed Yamhill but now simply called CT, now give Intel two fronts in the 64-bit war: this new, x86 ISA, and of course Intel's own, older, Itanium IA-64.

Brookwood said Intel's x86-based, 64-bit chip is codenamed "Tejas." That desktop microprocessor is expected to be the follow-on to the newly-announced Prescott processor, which is basically a 90-nm version of the Pentium 4. The Pentium 4 processor line is based on 130-nm process technology. Intel's Prescott can also support 64-bit extensions, but it is unlikely the company will offer that chip in a 64-bit version, Brookwood. The Prescott, with 64-bit extensions, is not compatible with AMD's 64-bit devices, he said.

The decision must have been difficult to make for Intel. While Itanium 2 won't likely be threatened by x86 64-bit in Big Iron, everywhere else the Itanium has tried to penetrate could come under risk. A 64-bit x86 solution from Intel goes a long way towards publicly acknowledging that there is still life left in the x86 ISA, and by induction, in AMD's Opteron and Athlon64. To make matters worse, one has to consider the Xeon; a high-power Pentium 4 CT will certainly eat into the Xeon's share. Of course, that's unless the Xeon gets CT, as well (and Xeon may very well get CT before anything else, as was the case with HyperThreading).

Speculation mode! It must be noted that a demonstration of CT does not necessarily translate to CT making it market. Nor is it necessarily the case that even Tejas will come to market, if in fact it is the first fully 64-bit x86 CPU from the company. Intel may be flexing its muscle to gauge industry response, and judge accordingly. The company could take CT and roll it into a set of extensions that they tack-on to their existing product line, needing to only add 64-bit general purpose registers and instructions to the existing architecture, or they could pull off a more AMD-like redesign, and run with a fully revised CPU (perhaps this is Tejas). This is to say that there's at least two different possibilities for x86 64-bit at Intel, and only time will tell which one chases after the Opteron and PPC 970 trains that have already left the station. Left the station? Yes, Opteron is catching on.

Advanced Micro Devices Inc.'s newest microprocessor chip for server computers runs twice as many of the servers sold in the third quarter as a competing Intel Corp. semiconductor, researcher IDC said. Computer makers shipped 10,746 servers with Opteron, the chip from Sunnyvale, California-based Advanced Micro, Intel's biggest rival in processors, IDC analyst Mark Melenovsky said. Intel's Itanium powered 4,957 servers, which run networks and Web sites.

All aboard! Athlon64 growth, Opteron servers, Intel's CT, 3 GHz PPC 970... 2004 is gonna be great.