PC Watch Japan has a picture of an Intel roadmap through 2004 (it’s towards the bottom of the page).
It shows that Prescott will come out in 2.8GHz and 3GHz versions in the first quarter, 2004. These will be the “OEM” CPUs; the next generation 1.6A and 1.8As.
This means Intel fans could have something new to do then.
If, Maybe, But . . .
Prescott is going to have a pretty short useful life: a year. It’s not going to scale very much. Actually, it’s not going to scale much at all, from 3.4GHz to just 4.0GHz in a year’s time.
Then Intel will move to Tejas.
This is not exactly a robust ramping schedule. In fact, it’s the smallest ramp percentage-wise after a die shrink in recent Intel history, just 15%.
The initial ramping schedule of a processor usually gives you a pretty good idea of what you can expect from it overclocking-wise. This schedule seems to indicate: not much more compared with what you can do today.
If 3.3GHz is pretty much a norm today, maybe we’re looking at 3.8-4 GHz (and maybe not even that much if the heat problem continues unabated) for the initial 2.8-3GHz Prescotts.
Intel’s roadmap makes one greatly suspect that Intel is trying to stretch out current, relatively inexpensive manufacturing processes to the max before they have to bite the bullet and start using weird science for Tejas.
And, like the PIII 1.13GHz processor, it may prove to be a bit too much of a stretch.
Keep something in mind. It wasn’t like Intel ever planned on having 5GHz Prescotts. The range was always limited, and that was before the Prescott heat problems emerged.
In other words, they weren’t expecting much, and at the moment are getting rather less than that.
When the Prescott wattage problem emerged, a lot of people said, “those Intel bast—-,” like it was a deliberate move, an Intel scheme.
The reality is really quite the opposite. The last thing in the world Intel wanted to say to its business customers was “Oops, we screwed up.”
While “runaway heat” may be too strong a term to use, Intel currently has what looks to be a major problem on its hands. If most AMDroids couldn’t stand Delta 7Ks, just how do you think Intel folks are going to feel?
Right now, the Prescott heat problem is about at the early stages AMD was with demonstrating 800MHz Opterons. It’s hardly conclusive proof yet, it may well be fixable, but it’s definitely a matter for concern.
Maybe Intel can fix this. Maybe they can’t for a while. Maybe they can’t at all, in which case Prescott will be even more stunted and Intel will be more or less stalled in 2004 until they can get Tejas (which presumably addresses these problems one way or the other) out.
They’ve already postponed their 90nm notebook processor, Dothan. Will Prescott be next?
Should that happen, no matter what Intel says, that will be the likely culprit.
Is God An AMD Fan?
If this situation can’t be resolved, we may have something similiar to another Pentium 1.13 situation on our hands. While it’s hardly going to bankrupt Intel, it could easily keep somebody else from doing so: AMD.
Recall that Intel’s inability to push the PIII to 1.13GHz and beyond gave AMD a chance and let it make some money. If that hadn’t happened, AMD’s financial position would be dire.
History could repeat itself.
If this problem isn’t easily fixable, some people at Intel are going to end up scratching their heads. Every time they seem to have AMD on the ropes, something happens and they get away.
Otto von Bismarck once said, “”The Lord God has special providence for fools, drunkards and the United States of America.” Intel execs may take out the “USA” and slip in “AMD.”
Meanwhile, the AMD execs are probably muttering under their breaths, “Burn, baby, burn.”
While this doesn’t make the initial generation of Athlon64s any better for overclockers, in the overall computer market, Intel heat problems and resultant publicity would inevitably help Athlon64 sales (it probably would also help Opteron server sales, you don’t want furnaces in those blade servers, you know).
If AMD has a relatively easy 90nm transition, and IBM’s help makes SOI more rampable, we could even have a repeat of the TBird ascendancy. It probably wouldn’t last terribly long, but it would be enough to establish the product and keep AMD going for a few more years.
Intel’s PIII problems essentially shifted about 5% of the x86 CPU market from Intel to AMD. The last eighteen months or so, AMD more-or-less gave that 5% back.
That 5% may not sound like much, especially not to would-be Intel conquerers, but it’s the difference between AMD being somewhat profitable and being faced with perpetual losses.
So even a modest shift could have a big impact. It would be like a giant giving a couple pints of blood to a very ill child. The giant may be left a little queasy afterwards, but it could be life-saving for the child.