AMD's upcoming Athlon 64 low-end variants, codenamed 'Paris' and 'Victoria', will not be offered as 64-bit processors but as 32-bit upgrades to the current Athlon XP line.
So claims Xbit Labs, having glanced at the chip maker's latest roadmaps.
Paris and Victoria emerged earlier this year, when they were revealed to be cut-down versions of the Athlon 64. At the time, it was assumed that was simply a matter of their smaller, 256KB L2 cache. Paris will be fabbed at 130nm, and is due to ship sometime during Q4 2003. Victoria will debut late Q2 or early Q3 2004.
However, it now appears that the chips will not operate in 64-bit mode. The Athlon 64 supports three modes of operation: 32-bit 'legacy' mode, dual 32/64-bit 'compatibility' mode and pure 64-bit mode. The last two require a 64-bit OS, and are essentially sub-modes of the chip's 64-bit 'long' mode.
If Xbit Labs' report is correct, Paris and Victoria will be limited to 'legacy' mode.
AMD has always touted Athlon 64's full 32-bit compatibility, partly to make it easy for users to migrate from the 32-bit domain to the 64-bit world, but mostly because it allows it to use one chip to target both domains, rather develop, market and manufacture two products, which is Intel's approach.
For now, AMD will continue to offer Athlon XP as a lower-cost alternative to Athlon 64 for OEMs and users who feel the latter has yet to prove itself. It might seem logical to just kill off the XP and drive punters toward the 64 - whether they use the 64-bit mode or not - but there would likely be some resistance to such a move. And some buyers will still assume - incorrectly - that the Athlon 64 is somehow not-quite-compatible with or unsuitable for 32-bit apps. Sure, the fanboys appreciate the difference, but AMD has and wants to target plenty of other buyers, many of whom don't.
So Athlon XP has a role for the immediate future. Yet it makes sense for AMD to kill off the XP and switch production to 64. The only way to reconcile those two forces, one marketing and one financial, is to brand some 64s and XPs. Hence, presumably, Paris and Victoria. Such a plan also allows AMD to offer better XPs than today's generation of the chip without having to commit resources to extending the current XP architecture.
It also allows AMD to eliminate Socket A without killing off XP - Paris and Victoria are said to be Socket 754 parts.
At some point, the market will accept that Athlon 64 is fully 32-bit compatible, and is faster in that mode than 'pure' 32-bit chips like Athlon XP, and AMD can rebrand the last few XP as a Duron, and focus its attentions on 64-bit CPUs. But if the Xbit Labs report is accurate, it doesn't look like that's going to happen until well into 2004, and possibly further off.
The downside with this tactic is that it ultimately reduces the number of shipping 64-bit systems in the field - by allowing buyers who might choose an Athlon 64 to stick with XP - and thus make it harder for software developers to justify porting their 32-bit apps to AMD64. But presumably AMD feels that those whose apps will see a real benefit from the 64-bit domain will make the move in any case - just look at Tim Sweeney's eagerness to do so, even when there are no Athlon 64 machines on sale yet - and the rest won't, no matter what the mix of 32-bit and 64-bit CPUs it offers. What software support it does lose by pursuing such a strategy is far out weighted by the sales it loses from customers who refuse to be forced down the Athlon 64 path.
AMD's move should pan out something like this:
|Q3 2003||Q1 2004|
|Athlon 64 FX|
|Athlon XP||Athlon XP (Paris)|
|64-bit Opteron core/Socket 940|
|64-bit ClawHammer core/Socket 754|
|32-bit Athlon XP core/Socket A|