Intel Cuts Prices On Top; AMD Doesn’t
As fully expected, Intel cut prices on PIV yesterday. This could make the 2.4GHz C a little bit cheaper, and gets the 3.0C under $300 and the 2.8C a bit over $200.
The 2.6C might get more of a look as an overclocking chip, especially if it ends up being priced about the same as the 2.4.
There’s nothing earthshattering in any of this.
AMD reduced prices on higher-end Athlons, but left prices alone on FX/Athlon64. This has come as something of a surprise/disappointment to many, and one even said that this means AMD can now
“deftly avoiding the price hammer strike, and doing what it wants.”
The comment is dubious when it comes to the FX, and downright silly when it comes to the Athlon 64.
In the case of the FX, sheer scarcity is bound to keep the price up. Based on this production estimate, only 55,000 socket 940 FXs are going to be made, ever, and only 15,000 are supposed to be made this quarter. Even if AMD manages to crank a few more out, this is still a very limited edition CPU.
Besides, what is Intel doing? It’s replacing a $637 CPU at its high end with a $925 CPU at its high end. That hardly puts any price pressure on the $733 FX-51.
It should be noted that AMD has cut prices significantly on single-CPU configured Opterons to price just a bit higher than equivalent PIVs.
Athlon 64? AMD has been trying for about a year now to get price parity with Intel on their high-end chips. For the Athlon64, AMD set the initial official price of the 3200+ to be exactly the same as Intel’s price for the 3.2C post-October 26.
So instead of cutting prices on the A64 right after Intel, they just “cut” the price to expected Intel levels long before Intel did.
You really can’t claim too much credit for standing up to someone when you fall to your knees before he even shows up.
The next Intel price cut for the 3.2C is scheduled for February 15, down to $278. Can AMD sell 3200+ A64s for around $400 for the next three months or so? Will they have to start discounting the chips as production rises as the old year ends and the new one begins?
In short, can AMD escape the historic discount it has had to offer compared to Intel when they’re trying to sell a sizable number of CPUs?
We shall see. I would bet that they can’t, and we’ll see another “advance” price cut on the A64 3200+ in early January down to the $250-275 level.
More Prescott Problems
This article basically translates a report from a Japanese website which says that Prescott is having big mobo-in-law problems.
It goes on to say that Intel has to redo packaging and come up with a C stepping hoping to come up with something that will work by February, with mass production only to come considerably thereafter.
Frankly, this sounds inexplicably bad. If (comparatively) dumb ignorant overclockers can get high-end Northwoods to run at 3.6GHz and 300MHz FSB speed with little fuss, and Intel can get Xeons to run at 3.4GHz and 3.6GHz, why can’t Intel get Prescott (which is essentially a die-shrunk Northwood) to run at 3.4GHz and a 200MHz bus?
This is beginning to sound like the 1.13 all over again, and this is a critical point most observers haven’t gotten yet.
A delay of a few months is really no big deal in the great scheme of things for Intel. It may be good for AMD in that they’ll be able to charge more for the relatively limited number of desktop Hammers it produces early next year (but not if those higher prices leave most on the sidelines).
However, if this is a 1.13 situation, if the process is basically broken past a certain speed, and Intel has to go back to the drawing board to serious revamp, we’re talking about Intel being on the sidelines for a long time, just like it was in the Thunderbird era.
At worst, Intel may have to forget about Prescott and Tejas, and force-feed the next major redesign, Nehalem. Should that be the case, the first Nehalems might as well get named Willamette IIs. They’ll probably be rushed, cut-down design implemented a process generation too early, just like Willamettes.
Should all these bad things happen, the next good processor from Intel might not come until 2006.
This would be much, much better for AMD, and if AMD can get 90nm SOI right more or less on time, they ought to do quite well.
Again, what is good for AMD isn’t necessarily good for overclockers. If AMD tries to charge skyhigh prices when they’re the only guys in town, their current customer base won’t buy, and it’s questionable how much and how fast Intel loyalists will shift over.
Remember that AMD had to slash prices even when they had a huge lead against the PIV with the later Thunderbirds, and we saw the spectacle of the world’s fastest desktop CPU at the time selling for about $110.
There’s some less-than-solid evidence that Intel plans to push its EE to higher speeds: 3.4GHz, even 3.6GHz.
I suppose that’s the contingency if Intel can’t get Prescott right fairly soon, and that would probably be a good stand-in for four-six months, but that’s it. Once AMD goes to 90nm, FXs/Opterons should be able to overshadow them.
Yammering About Yamhill
One analyst thinks that Intel will come out with Yamhill (their version of x86-64) and that Yamhill will somehow be incompatible with AMD64.
Please note that the analyst doesn’t know that, he’s just making an educated guess, and it’s not a bad one.
Up to now, the most educated guess has been that if Intel ever did let Yamhill out, it would do so only on the Xeon end to compete against Opterons.
However, if Intel is faced with a 1.13 situation and they have to go back to the drawing board, they have to do something to at least look competitive against AMD.
What’s one way to increase performance without increasing clock speed? x86-64.
Of course, there’s a big problem with that. If Intel adopts an x86-64 compatible with Hammer, that doesn’t help matters at all. AMD will say “thank you” and get the same kind (or greater) speed boost from all that x86-64 software Intel helped to promote.
The only way it makes sense for Intel to adapt x86-64 for the desktop (and frankly, this would be a flat-out admission by Intel that a 1.13 situation exists and they’re desperate) would be to come up with an x86-64 standard that is NOT AMD64 or even Hammer-compatible.
Is it the kind of thing Intel might do? Yes, but that doesn’t mean Intel will do it or perhaps even can get it done.
Just a few points to consider:
As we pointed out earlier, going with Yamhill on the desktop is a sure sign Intel is in deep trouble and can’t get out of it. It’s doubtful they’ve come to that conclusion quite yet.
Even Intel can’t bring on a flood of x86-64 software instantly, or even too quickly. It would probably be a year from an Intel decision before you’d have a reasonable amount of software ported/developed for it.
Just how incompatible to AMD64 could you make a version of x86-64, anyway? Even if you could, how willing is Microsoft to reinvent another 64-bit wheel, and how long would it take them? Make it seriously incompatible, and MS will take forever to do it (even assuming they’re agreeable to this, would you want to rewrite zillions of drivers?).
More likely than not, Microsoft would just take its Windows for AMD64, dink around with it a bit to make it “incompatible” (perhaps just as little as having the OS read the CPUID and reject anything that doesn’t have Intel inside it). If that happens on a Monday, by Tuesday, there will be a patch which will make the OS not “incompatible” any more.
Even at best, this is an extreme move, and one Intel won’t take unless they feel they have absolutely no other choice.