Yea or Nay?

An Intel exec is reported to have said it isn’t going to happen.

Others are saying that there’s initial silicon out there with x86-64 included.

A few points:

Paul Otellini may well have said sometime, somewhere, that x86-64 wasn’t in the cards.

That time was not in the speech he gave to the Securities Industry Association June 19. Here’s the transcript of that speech. The speech basically says Intel’s IA-64 servers are a lot more cost effective than multi-processor SPARC stations and the like.

Actually, the article doesn’t state that he said so in the speech, but rather in some sort of “report.

Looking a bit deeper into this, it appears the statements in question came during a press briefing, not the speech.

There is a line in the article which states, “Otellini said Intel is not planning chips that can handle both 32-bit and 64-bit code.” However, if you keep reading, it looks like he’s saying that Itanium chips won’t do both. Otherwise, the following comment doesn’t make much sense:

“By forgoing greater compatibility (i.e. running 32/64 bit code), Itanium can provide greater performance and ‘all the other things people care about,’ Otellini said.”

Additionally, the story ends with this line, ” . . . Otellini said Intel has reserved patents to extend its 32-bit Pentium and Xeon chips to handle 64-bit code.” That’s hardly the action of a company burning its bridges on x86-64.

Probably didn’t help that the author of the piece doesn’t seem to know that x86-64 and IA-64 are incompatible standards.

It’s Not Either/Or

x86-64 and IA-64 aren’t necessarily fire and water, either. Intel could do both, and they could coexist. So if Intel says they’re committed to Itanium, that doesn’t mean Yamhill can’t happen, nor vice versa.

The Itaniums are not Xeons with 64-bit channels. These are far more complicated, sophisticated CPUs, with hundreds of millions of transistors as compared to tens of millions for desktop/light end server-level CPUs like Xeons (or Sledgehammers).

Itaniums aren’t aimed at anything AMD is going to come up with; they’re aimed at SPARC and targeted towards those who’d find a $45,000 server dirt cheap compared to what they’re paying now.

AMD’s strategy with Sledgehammer is basically to say, “If you think 45K is cheap, how does 15-20K sound?” Maybe not quite as good as members of this audience might think. For sure, Sledgehammer will find buyers for whom the cost difference makes a difference, but the securities industry probably won’t be one of them.

Keep in mind that in this market, Intel is the underdog trying to move in. AMD will be the underunderdog.

The biggest problem Intel will likely have has if it adopts x86-64 won’t be so much competing against Intel as competing against itself, as in “Why should we buy Itaniums rather than Prescott Xeons?”

They Don’t Have To Turn It On

Hyperthreading is built-into current Xeons; it’s just not used yet.

As we and others have stated repeatedly in the past, from a silicon standpoint, x86-64 is not that big a deal; it requires little silicon real estate and relatively few CPU changes. That’s why AMD liked it so much.

If I had to bet on this today, I would bet that Intel will include x86-64 in Prescott silicon to cover itself. We’re probably talking about 3 sq. mm of silicon using a 0.09 micron process. Then the issue becomes “Will they turn it on or not?” and that will depend on what happens in the market.

Please also keep in mind that we’re talking about a processor that won’t be out for another year, OK?


Be the first to comment

Leave a Reply