>It depends on the year. Giving yamhill a full green light instantly kills any hopes for Itanium success - they won't make that particular move lightly, though the recent news about Yamhill may indicate that Madison isn't much of an improvement on McKinley.<
I would hardly call any of the Yamhill press news, it is barely more than rumor. And, much as you might like to think otherwise Yamhill will not kill off Itanium. The original report stated that Yamhill is a modification to P4, which means that physical addressing will be limited to the same as for P4. For Intel customers to address a large physical address space, they will still need Itanium.
>Once they decide to kill off Itanium, they need a lot more than just the yamhill chip.<
None of the rumors suggest that Intel has any plans of killing of Itanium. Itanium will, if it dies, do so without any push by Intel.
> They also need to get chipsets designed and tested, and motherboards designed and tested.<
The rumor is that Yamhill will use the same infrastructure as P4. Of course, this would be very atypical for Intel which has changed infrastructure with every new core, even the 486SX needed a new infrastructure.
> That's probably around 1H 2003 - when Clawhammer will be moving to .09 and 64mm2.
Where did this come from? The mobile Clawhammer will be introduced at 90nm in H2 2003, with a 90nm server chip following before the end of 2003. The desktop Clawhammer will not begin migrating to 90nm until H1 2004.
> At that size, Dresden should be good for close to 16 million Hammers per quarter, and by then AMD will have well established platforms to take those chips - including mobile platforms.<
Dresden might reach 16 million processors per quarter, or might not. But with the exception that this will occur in Q4 2004, not H1 2003, the statement of well established infrastructure by the time they fully transition Dresden to 90nm is accurate.
>2H 2003 and 2004, Intel would match, then exceed AMD's run rate, unless they expanded Dresden (which it was designed for).<
This could be true, it is possible that Intel could introduce Yamhill about mid-2003, and with Yamhill using the P4 infrastructure they could easily be shipping more Yamhill processors than AMD can Hammer processors by the end of the year.
>By 2005, the new joint venture FAB is scheduled to be able to ship 50,000,000 Hammers per quarter.<
The joint venture fab will not begin shipping until mid-2005, and at earliest would be ready to ship 50 million Clawhammer processors per quarter in Q1 2007. OTOH, AMD could continue shipping value segment 90nm Hammer processors from Fab30 through full ramp of the JV fab and outsourcing to UMC could ramp along with the JV fab. AMD could reach 50 million processors per quarter in Q3 of 2006 with the JV fab 50% ramped.
>The Itanium line, of course, would be all but dead at that point.<
Or Intel could have resolved all the issues with IA64 and be migrating it onto the corporate desktop as they service the consumer market with Yamhill.
>Intel could lose the ball game, here. It's no wonder they're scrambling to replace Itanic. But, as we've seen from P4, and Itanium, it takes Intel a long, long, time to bring a new platform to market.<
Intel could indeed lose the ball game here, even if IA64 successfully takes the corporate desktop as well as the workstation/server market. The key is whether AMD's 50 million units per quarter capacity drives Intel to selling IA64 at unsustainably low prices to get it onto the corporate desktop. AMD has made a profit before with no processor priced over $150, and can probably do so with a 40mm2 Clawhammer. With half of a single 300mm fab and Clawhammer at 40mm2, AMD might even be able to make a profit while driving the price of the fastest speed grade Clawhammer to under $100. If Intel loses their ability to demand a premium for their processors, and is reduced to a combined total of IA64 and x86 processors equal to the number of Clawhammer units AMD is then selling, can they make a profit while maintaining four 300mm fabs? |
