Kash,
i would hardly call this an immaculate launch.
I think you should come down to earth. We have some issues in Athlon platform that need improvement. You were the loudest critic of AMD in this area. Did you expect these issues to magically disappear on June 5th? AMD still depends on Via for chipsets, and there is not much AMD can to make Via devote more resources to Athlon projects.
The only thing AMD can do is to have an in-house team developing high quality chipsets, which is being addressed with 760 / 760MP. The fruits of which are on the way. There is no point bitching about it anymore. AMD has learned a lesson, and will not repeat the same mistake in the future. I expect AMD chipsets to be ready for Mustang and Sledgehammer.
Launch partners? It would be nice to have a dog and pony show that Intel orchastrates, but it is not essential.
This means that it is likely to be a chipset or MB problem why no higher speed grades were launched.
I don't think we have any evidence to draw this conclusion, but we know there is a problem with chipsets, not with the chipsets themselves, but with low quantity of them: theregister.co.uk
Board makers wonder if AMD, VIA can keep up
AMD probably has up to a million Thunderbirds and Durons ready to be sold, but Via is not ready. So what can the other Launch partners do? They have to wait. Compaq and HP need time for their machines to make it through the channel.
All this will mean that some parts than might have been sold in Q2 will instead have to wait until Q3 (I can't wait for Elmer's comments about the inventory)
Regarding L2 related issues, AMD took the safest route, a route with fewest risks and generous rewards. I think there would be a lot more room for error if AMD went out of it;s way in redesigning the core. That's the job for Mustang.
The job of Thundebird is to eliminate the performance penalty from off chip L2, lower cost, move to the Slot package, possibly cut the packaging time. I think it is a complete success. Would I like to see faster L2? Of course, but not if the cost would mean 1+ month delay. The cost of 1 month delay is about 500,000 units x $250 = $125 million. At least.
Joe |
