Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: Windsock who wrote (150954)	12/4/2001 12:15:47 PM
From: wanna_bmw	Read Replies (2) \| Respond to of 186894

Windsock, I'm sure there is partial truth to what both you and Paul is saying. However, as a whole, it's all rather dubious. For example, I don't believe Austin manufactures 100% CPUs. Therefore, 5,000 wspw is meaningless in these calculations. Second, AMD is using their Dresden fab to out-do Intel's manufacturing capabilities (not in volumes, but performance). Therefore, many of Dresden's wafer starts are probably not going to production worthy CPUs. Additionally, AMD is using aggressive gate dimensions and high speed bins to give the perception that they still lead in performance, even though they can't make enough parts to satisfy demand (which, relatively, is small compared to Intel). Therefore, yields on these CPUs is probably low enough to drag down the average. I am willing to believe that Athlon XP yields far worse than Duron or Thunderbird. Remember when AMD tried getting higher bins on their K6-III? The yields were supposedly quite abysmal. Now, AMD is fighting for control of the market until Hammer arrives, and they know that going up against Northwood will be more difficult than going up against Willamette. That's why they are trying to establish model numbers now, so that their lack of frequency later on isn't so noticeable. Athlon XP will no doubt yield better on a fully .13u process, but AMD seems to be having trouble getting this started. I don't doubt the rumors that Intel is also having problems with .13u, but at least they have it started, and are much closer to being at full production. The only thing I think we should take away from this, rather than spreading FUD about AMD's yields, or plant efficiency, or anything else, is that they are falling far short of their maximum capacity in order to compete more aggressively with Intel. If they were conservative and tried to max their output, they would never have launched Palomino above 1.5GHz, but would have waited for Thoroughbred. We've always thought that getting market share is the most important thing to AMD. Perhaps we were wrong. Maybe the most important thing is maintaining a facade of leadership, while they have to sacrifice production to make that possible. Obviously, they won't be gaining too much market share, despite Intel's difficulty meeting demand, since most of their output is being thrown away due to aggressive requirements. wbmw

To: Windsock who wrote (150954)	12/4/2001 12:32:55 PM
From: Elmer	Respond to of 186894

All of the statements in Paul's calculations are absolutely solid facts with the exception that the assumption of 10,000 wafer starts per week. Even at 7000 WSPW their numbers are terrible. My calculations show they should have at least 2X the capacity they now produce. Standard yields on a new process are 70% or better initially and go up from there to 80% or even 90% on a mature process. Winsock, this point needs clarification. Quoting yields in percentages is meaningless because of differing die sizes. 80% yield would be incredibly good for a very large die and incredibly bad for a very small one. The defect density for the process plus die size gives predicted yield. Seeing as die size is a constant for Athlon the only way to reduce yield is to increase the defect density (or general misprocessing). The problem we face is accounting for such disasterously low yields. I find it hard to believe that defect density alone could do it. It is far more likely that they are simply losing lots of wafers because of misprocessing in trying to achieve their 70nm channels. Nobody could have these terrible yields without some real extenuating circumstances. Another possibility is they have decent yields but must scrap half their production due to poor binsplits. Bottom line is that for whatever reason less that half the expected output is coming out the door. EP