"If they are initial samples, then my prediction is almost guaranteed to come true."
Could be. Here is what Dirk was saying at the Analyst's conference (courtesy of JC).
"We will be sampling Clawhammer in the first half of next year and we will be shipping in the second half of next year."
Of course, since that was in 2001, next year is 2002.
On the subject of samples, here is what Hector had to say about SOI:
"When we go to 130 nm, and have been now for almost six months been running engineering samples with SOI"
Of course, he doesn't say engineering samples of what, it could be the SRAM that they did early in 2001. But then the times don't match up. Note: I am not by any stretch of the imagination claiming that these samples are Hammers. Given the timeline, they very well could have been Spitfires or Tbirds though, since those were mature and well understood. Remember that they did a K6-2 to qualify their copper process. And yes, I think Intel's habit of moving a mature product over as an actual product is a better use of resources.
And this is what Dirk is claiming about Tbred.
"We have fully functional Thoroughbred silicon in hands today. We are poised to provide engineering samples product this quarter and will commence shipment of this product in mobile systems in Q1 next year."
They seem to feel that they can go from engineering samples to production in 2 quarters or less. Now true, this is for what is basically a dumb shrink of an existing product.
Now, you ask, where am I going with this? As I have stated before, I still think that the supposed Hammer tapeout in Q4 of 2000 actually happened. But, it was aimed at Socket A (AMD initially made noises about ClawHammer being Socket A, but lower voltage) and suffered a fate not publicly stated. It could have been that the FSB bottlenecked it too badly, or there might have been some other problems. For what ever reason, they decided to drop it. But, they would have had a core that they could profile and check out. So this most recent tapeout would be for the core with any fixes and changes and the embedded memory controller. Ok, those things would still be a very big chunk and fraught with possible problem areas, but they wouldn't be a whole new design either. Since they have been running at least SRAMs and likely processors on 0.13 micron SOI since Q101, then they have at least some experience with SOI. So a really optimistic projection would be that they have samples sometime in early Q202 and they actually ship some production units on Dec. 31, 2002. That would give 3 quarters from sample to ship. Of course, if any major problems pop up, they are hosed, but minor things might be fixable in that timescale, given that AMD is likely to continue to use their cell-based design process. |
