Re: It's not unrealistic with a total capacity of close to 10,000 wafers per week...
It's more than reasonable for 10,000 200mm wafers, it's to be expected, in fact, it's low:
100mm (radius) * 100mm (radius) * 3.14159 (pi) * .8 (edge and cutting loss) * .8 (intel's famous high yields) * 10,000 (wafers per week) / 106mm2 (area of a coppermine or celeron II) = 1,896,809 die per week per Superfab. It really should be closer to 2 million per week than one.
And isn't Intel supposed to have two of those Superfabs and two additional .18 megafabs that build coppermines? And this was before the Israel fab dedicated to coppermines opened (it was supposed to be a flash fab, but when the crown jewels are in big trouble...).
So start with 10,000 wspw (wafer starts per week) and add some part (half?) of 3 other very large fabs that are producing coppermines. So there should be something like half of two 5,000 wspw fabs plus the one above plus 6,000? from Israel plus half of 10,000 wspw from the other superfab = about another 3 million die per week giving a rough total of 5 million .18 die per week at .8 yields.
That would result in 65 million chips per quarter at a .8 yield. It's no wonder that you and PB and pAUL kept telling use that AMD was going to get clobbered last fall, then right after New Years, then in the spring.
But instead we are seeing 30 million Intel chips per quarter with 60% (according to realworld technology) of those still expected to come from the old .25 fabs through the end of the year so there are only 12 million good chips coming from those 81 million die. I guess that .8 yield estimate was off a bit.
I'm sure that I'm off on some of these numbers, but the old story of yields being as good as in the old days doesn't seem to make much sense.
So just how poor are Intel's yields? I think that my .8 is high - is it closer to .23? What's the cost of production if 4 Celeron IIs have to be fabbed to get one good chip? What do they sell for after discounts?
Dan |
