Replies to post #1786 on Intel Corp (INTC)

[Bonefish]

RE:"He argued that for the first half of next year, the 512 KB "Barton" Athlon XP will represent the high-end desktop CPU, and that AMD will concentrate their efforts on producing Opterons for the lucrative server market"

Surely you know why this could be true.
AMD shot their wad with SOI and Low K only to find out yields are bad. With yields bad they have no choice but to push Opteron ahead of Clawhammer because they can make up for the poor yields with high ASP server chips.
Maybe later they can get yields high enough to offer some desktop Opterons.
It all blew up in their face in Setember.

Intels only competition now is themselves. Whether they ar spending too much. Paying too much. Giving out too many stock options etc.
Bone

[fingolfen]

Krewell dropped a bombshell when he predicted that the 256 KB Clawhammer is dead. He argued that for the first half of next year, the 512 KB "Barton" Athlon XP will represent the high-end desktop CPU, and that AMD will concentrate their efforts on producing Opterons for the lucrative server market.

Interesting rumor... of course, Pual Hales on the Inq is predicting a clawhammer Athlon in Q1: http://www.theinquirer.net/?article=5839

Given the recent CC, I think the roadmap the article Paul found online is out of date...

This will allow AMD to substantially bump up the Average Selling Price (ASPs) of its processors. The Clawhammer will essentially be an Opteron with fewer HyperTransport links. I discussed AMD's difficult situation with him, and we agreed that if Clawhammer comes out with 1MB L2 cache it will indisputably regain the performance crown.

At what frequency? When compared to what? Northwood or Prescott? A 1.6GHz K8 even with 1MB of cache would probably have trouble keeping up with a 3.06 GHz HT P4-N'wood. A 2.0 to 2.2GHz K8 would probably bury a Northwood, but will probably run into trouble when it goes up against a Prescott.

This way, AMD will only be producing one core, which will reduce manufacturing headaches. However, it will then be imperative for AMD to transition smoothly to 90nm, in order to have substantially smaller die sizes than Intel. If AMD stumbles on the 90nm transition, then they will have to either cut the L2 cache to 256 or 512 KB and match (but not exceed) Intel's 90nm performance, or come out with a full 1MB Clawhammer and accept the huge die size penalty.

This part I'll agree with completely. Of course, given the problems AMD has encountered at 0.13 micron and 0.13 micron SOI, I'm not hopeful that they'll have a smooth 90nm transition. They were a year behind Intel moving to 0.13 micron, I honestly don't see how they can "make up" that whole year in one fell swoop, given they only have one development fab. I predict Intel will beat AMD to 90nm by at least 6 months.