Intel’s decision to take a heavily-cached Xeon and let it go slumming with gamers is a pretty brilliant tactical move against AMD.
It’s quite ironic, given AMD’s penchant for secrecy to the point of silliness, that Intel did the last-minute blindsiding.
It’s going to be very hard to seriously review an Athlon FX without comparing it to an Extreme Edition, and the likely general advantage a 2.2GHz FX would have over a “regular” 3.2GHz will vanish against this chip. No doubt the FX will still win some, but so will the EE. It ought to be enough to deny AMD a clean win.
But is it any more than that?
In one sense, yes, it’s more than that, if only because Intel said there would be a 90nm version of this chip (presumably when they have 90nm Xeons with 2Mb L3 cache to spare).
That’s an important point to note. This is a 130nm CPU. It’s overclocking potential is pretty limited.
Relative price is another point. The price of this chip is rumored to be a bit over $700. That may not look so bad compared to current 3.2 pricing, but by the time this CPU becomes available, 3.2s will cost just a bit over $400, and due to go to around $275 by February.
Will the price of the EE plunge likewise? If so, that would be really interesting.
If not, at least from the gamer perspective, this processor would have a very limited time on top in the overclocking world. A cheap overclocked Prescott next February probably will be able to match or beat the EE.
And that’s just what Intel is going to do next February.
Hammering Hammer Prices
What is much more important to AMD’s prospects is not what Intel does with Xeons tomorrow, but what they do with Prescott prices next February, just at the time when AMD will start seriously producing the things.
If you don’t overclock even a bit, the price of a 3.2GHz Prescott will be $278. Given that a current 3.2 at least holds its own against a 2.0GHz Opteron, it would be extremely reasonable to assume that a Prescott with double the cache and other enhancements would do at least 5% better than that, perhaps 10-15% better against a single-channel Athlon 64 with 1MB cache, and 15-20% more than an Athlon 64 with 256K cache.
If you do overclock, the price of a 2.8GHz Prescott will be $178. It’s probably safe to assume that an overclocked Prescott will hold the same type of advantages over overclocked FXs or 64s.
What’s important here is not who comes out on top and by how much. What’s important is the price point at which the competition takes place.
If anybody can get X level of performance from Intel for $278, how much can AMD charge for roughly the same level of performance with its FX, and how much can it charge for somewhat less with the big 64, and a lot less for the “little” 64?
If overclockers can get X level of performance from Intel for $178, how much can AMD charge for roughly the same level of performance with its FX, and how much can it charge for somewhat less with the 64?
This is going to put AMD at more than a bit of a manufacturing disadvantage, since both FXs and 64s with 1Mb caches are a lot bigger than Prescott will be.
If historic pricing patterns repeat themselves, a 2GHz FX in October should have a market value of about $180 next February, with a big 64 being less than that, a little one being a lot less than that.
What this will do to Opteron pricing also ought to be interesting. How much more will AMD be able to charge for a single or perhaps even dual Opteron CPU under these circumstances?
Will AMD recognize that? Or will they think that x86-64 is such a huge benefit that they can charge more than Intel solely due to that?
Our recent surveys have indicated that x86-64 is greatly valued by a relative handful of zealots who have an immediate use for it, and viewed rather indifferently by most of the rest.
There’s probably enough zealots to buy the initial token shipments of desktop Hammers at skyhigh prices the rest of the year, but then what?
Our fear is that AMD is going to be seduced by that handful of zealots, and stubbornly keep prices up until sheer lack of sales force them to do a 180.
Remember that both socket 478 Prescotts and the first generation of FX/64s will be superceded sooner rather than later, and a sudden AMD pricing reversal in April or May may not help matters much with second-generation Hammers due to arrive imminently.
There is one beam of sunlight in the gloom, though.
If this story is correct, Intel is still having problems with Prescott heat (at least they’re not changing the Prescott specs yet), and not even weird science is going to help future generations very much.
Intel processors will stay north of 100 watts, spec, even with it. This means more like 150 watts for overclockers.
This means a range of about 1-1.4 watts per square millimeter. That’s awfully hot, and will be hard to cool quietly and/or inexpensively.
If AMD can do better than that with its SOI sauce, and Intel starts catching a lot of hell for making processors that simulate it, or, even worse, pass away from the heat. That could open a lot of doors for AMD.
The main effect of the EE will be to ruin AMD’s coming-out party, and little more.
The real fight will be early next year, with cheap Prescotts going against AMD on performance AND price.
We don’t think much of either competitor due to the technical obsolescence issue, but it’s going to be hard for AMD to charge an arm and a leg for FX/64s starting next February.
One should plan accordingly. If AMD is going to have to bite the bullet in four months if they’re smart, six months if they’re not, there’s no point in most of you swallowing it whole before then.