SI
SI
discoversearch

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.
Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?


To: Elmer who wrote (41394)11/12/1998 11:57:00 PM
From: Scumbria  Read Replies (1) | Respond to of 1571843
 
a larger die will yield lower percentages with the same defect density as a smaller die.

Imagine a wafer with 30 randomly spaced defects on it. If the die size was huge, (1 die per wafer), the yield would be 0% (no good dies per wafer.) On the other hand, if the die size was small (1000 dies per wafer), we would expect to have 970 (1000 - 30) good die or 97% yield. 200 die per wafer would yield 85% (170 good die.)

Anticipating the possibility of some inane retorts, I will reply in advance with some of my own.

1. Yes it is possible to have multiple defects in one die, thereby increasing the yields.
2. Of course it is ridiculous to have one die per wafer.
3. I made up the 30 defects per wafer number.
4. Defects are not the only reason to reject a die.
5. This message has been formatted to fit your screen.

Scumbria



To: Elmer who wrote (41394)11/13/1998 12:17:00 AM
From: Bill Jackson  Read Replies (1) | Respond to of 1571843
 
Elmer, What you and Scumbria say seems quite valid. I expect the gradual elimination of assorted defects is the real art in fine uning a line. AMD/Intel might be expected to approach a similar end point as they find that there is a law of diminishing returns operating and it might be pointless to try to eliminate some defects as the cost exceeds the extra profits.
Fine tune the production dance.

Bill