However, you seem to be confused with math a little
bit. Since the wafer area remains the same, it does not
matter how many dies are per field, and the
throughput of a stepper in terms of wafer per time
remains the same. Certainly you have few Willi
chips out of this, but I do not think that fewer
samples per wafer would be so critical for
"process qualification" unless the yield is
in single digits.
- Ali
What you say is very true.... assuming that intel qualifies their process based on number of wafers out, and not on number of good die out. Do they? Even if the actual time of each wafer through the stepper for CuMine and Willamette is the same, and the yields were the same, the Willamette wafers would have fewer total die per wafer.
I guess an argument could be made that a process would be qualified, when you reach a certain number of "Good Die Per Wafer", but how would you know that the particular run wasn't an abberation? I'm not aware of what exactly intel calls "qualified", but if it were me, I would have to have, a certain number of total Good Die Out, before I would consider a process qualified, not just a certain percentage of die per wafer, or a certain number of wafers.
You could be right, maybe the yields aren't all that good, but I have a feeling that the truth is otherwise.
If I were starting up a new Fab, on a new Process, I'd want to start with a product that I already had allot of manufacturing data on, not on a new product. It's allot safer that way. But that's just me.
SemiconEng |
