PB,
Do you have any WAG on defect densities for IBM's Cu process?
Mighty low. IBM is touting MTBFs of 30 years! on these machines, which are mainframes (ugly word these days, preferred name S390). Now, you know as well as I do that, with reliability = yield (or a strong derivative of it) IBM wouldn't be putting out MTBF claims in the 30 year range, on machines with 100s of millions of gates complexity, unless the defect density was way way under control and low.
BTW, MTBF of 30 years means most of these machines will never fail, because they'll be pushing up daisies (powered off, scrapped) long before that. If Intel wants to compete with IBM up there (THE most mission critical applications), this is the reliability standard they'll be hearing about. Of course, there's always the Sun Micro business to go after. Sun is nowhere near IBM wrt RAS.
Tony |
