Let's talk 64-bit

by Anand Lal Shimpi on February 3, 2004 12:09 PM EST
Back when AMD was hyping x86-64 and Hammer "later this year" I was constantly hounding Intel for their thoughts on the chip and AMD's strategy. It was the absolute most difficult job to try and get a straight answer from Intel, even from the individuals that had usually given me information on a more straight-forward basis. It got to the point where everytime I met with a new high level Intel employee, the 64-bit question was on my list of things to ask. I've asked Intel PR, I've asked Bill Siu and I've asked Pat Gelsinger.

The Intel PR response is usually not too useful, but both Bill and Pat gave me the same type of response: if and when it makes sense, 64-bit designs will be made available to the desktop.

I don't think anyone at Intel will argue that it will eventually make sense, so the "if" part is easily answered - what is unclear however is the second part of the statement: when.

Had it not been for AMD, then "when" would be at least 2 - 3 years from now. AMD has got the attention of Microsoft and I wouldn't expect MS to release an OS without plans for significant software support. If the x86-64 edition of Windows does take off, Intel will bring 64-bit extensions to the Pentium 4. However what must be understood is that Intel does not want this to happen right now and by demonstrating 64-bit extensions to x86 this early in the game, Intel would be doing nothing less than throwing their support behind AMD's move.

This brings me to the topic at hand; I've read the News.com reports that say Intel will demonstrate 64-bit extensions to the Pentium 4 and Xeon lines at IDF this month, and honestly it doesn't make sense to me. I could be very wrong about this but it just doesn't jive with the rest of Intel's behavior on this topic. I just can't see Intel saying that they don't agree with AMD's decision to move to 64-bit now and then turning around and supporting AMD's moves later this month.

Bottom line: the official Intel quotes that are floating around about 64-bit support on the desktop are nothing new. Where this demonstration of the technology at IDF business originated from is anyone's guess; I don't think it'll happen but I could be wrong.
Comments Locked

13 Comments

View All Comments

  • Optimus Prime - Thursday, February 5, 2004 - link

    BTW, would Intel have publicly denounced News.com's article about the Xeon CT had it been false, or do companies not have enough time to address rumors like that?
  • Optimus Prime - Thursday, February 5, 2004 - link

    Oh I understand that an "IA x86-64" is highly improbable...I was just saying *that* scenario would be the most-interesting to me. I wonder if the IA64 could be reverse-engineered (if that's the correct term) to include a seperate x86 die. Pure speculation, yes, but then, would anything *have* to be recompiled? If it's compiled for x86 or IA64, then it'd run in a 'dual core hybrid', right? Then the only performance hit is the CPU would have to account for 'which die to go to'?

    ...so now here I am making wild speculation after all!

    Derek, no offense intended, but isn't microsoft 'just patching and recompiling' for AMD64 right now? BTW, how many more years of "IF alternative-OS ##### takes off, THEN..." are your communities willing to commit to before they accept their current (and quite possibly permanent) market share? Do you think Linux will do what Mac has never done?

    [I'm not flaming, I'm asking a serious question]
  • Steve - Wednesday, February 4, 2004 - link

    IA-64 is a far superior ISA, which in the very long run I think will eventually replace x86 and its variants. But for the moment x86-64 will be the prevailing ISA for very simple reasons. WIth x86-64 you have such a massive application base that you can run and also the ability to run a newer more powerful 64-bit app base that doesn't need much re-working.

    I liken it to the automobile engine. In the past 100+ years engines have become faster, more powerful, and more efficient. But the underlying tech is still the same, its a gasoline powered internal combustion engine. In the same way that PC processors have become faster, more powerful, and more efficient. but still use x86 as the underlying technology.

    Although there may be a more efficient or better way to do things when its just not worth it the change just won't happen. There is just too much invested in the x86 ISA.

  • DerekWilson - Wednesday, February 4, 2004 - link

    Heh ...

    Intel wouldn't make their own 64bit extensions unless they further extended the AMD extensions. Microsoft is already writing an OS to support x86-64, and there is no way they're gonna try to release and support two flavors of x86-64. Just not gonna happen. Intel has to support what AMD has done at this point.

    IA64 is a better architecture than x86. The problem is migration. As long as there are pure x86 architectures out there, IA64 isn't going to be able to emulate at fast enough speeds to get people who need to run "legacy" software at acceptible speeds.

    Of course, there is a way around that ... If Linux and OSS really take off and kick some Microsoft butt, we won't be locked into the need for emulated software: we'd just need a bit of patching and recompiling ;-)
  • Optimus Prime - Wednesday, February 4, 2004 - link

    To me, what would surprise me is not the possibility of Intel making a 32/64 but if they made a 32/64 using AMD64 architecture. I would be less-surprised to see "IA x86-64" than "Prescott w/AMD64 extensions" by far. I just can't imagine that Goliath would not only play David's games, but also by his terms??? What would be most-exciting to me (but potentially the most-catastrophic) is if Intel made their own x86-64 up...after all, I thought it is Anand's, and others', opinion that IA64 is superior pure-64bit compared to the 64bit extensions in AMD64...I'd think Intel would have a superior alternative architecturally speaking. Isn't AMD64 AMD's very first 'architecture' of their own???

    Whatever happened to those rumors of a 64bit co-CPU sandwiched atop the 32bit CPU?

    No offense TrogdorJW, but 20-40mil trans either *are* or *are not* "unaccounted for". You say it, first, as if you *know* they're "unaccounted for", then you question yourself later, which to me implies you're not "sure" what you know. Now there's nothing wrong with knowing or not knowing but I don't come to Anandtech for rumor mills based on personal speculation. How do you even calculate 20-40 million unaccounted for transistors?
  • pumpkinierre - Tuesday, February 3, 2004 - link

    The cat is out of the bag. All the bad rumours about prescott have been right. Prescott was a semi paperlaunch with no 3.4E and cpus not generally available til March. Maybe Microsoft is on a roll with 64bit and intel dont want to be left behind. The walls are shaking- exciting times ahead.
  • TrogdorJW - Tuesday, February 3, 2004 - link

    I posted some stuff on this in the Prescott article, and I'll mention it here. Others have pointed this out as well. There seem to be roughly 20 to 40 million transistors in the Prescott that are unaccounted for. The only possibilities that I can think of are 64-bit extensions (most likely) or dual-core CPU (less likely).

    I think that Prescott has a very high probability of containing some 64-bit extensions. Based off of Intel's recent statements ammounting to "we'll have 64-bit CPUs when there's a market for them," I would imagine Intel is going to wait for the official release of Windows XP-64 and then release a patch that will unlock the 64-bit capabilities of Prescott.

    Any thoughts on this? I mean, those 125 million transistors went *somewhere*, and increased caches, a few new instructions, some work in Hyper Threading, and the change to 31 pipeline stages can't possibly have used all of those. Or could they?
  • Steve - Tuesday, February 3, 2004 - link

    I don't doubt that a demonstration of the minimal x86-64 capability in Prescott will occur behind the scenes at IDF. But as far as a public presentation or statement of any sort solidifying Intel's backing of x86-64 won't likely happen until Tejas. With Tejas I believe Intel will enable a fully x86-64 compatible capability in the Tejas version of Xeon. Then later on when Tejas hits 1200MHz FSB we will see full support in Pentium, a technology release very similar to HT.
  • Anonymous - Tuesday, February 3, 2004 - link

    Anand be like Trump, "You're fired!"
  • Anand Lal Shimpi - Tuesday, February 3, 2004 - link

    Derek

    You don't have to agree with me, I won't fire you ;)

    I'm pretty sure the first story was just reading too much into a quote, the second story I'm not so sure about. You'll find out at IDF :)

Log in

Don't have an account? Sign up now