Chipguy, that's interesting data. Looks like Opteron would never be suitable for high reliability systems. That's probably why you'll never see someone build a system greater than 4-way without clustering, so at least when a node goes down, others will continue running.
And to think that Dan3 is still obsessed over an old McKinley bug that Intel replaced with working parts. I wonder if AMD will replace any parts that fail in the real world due to deadlock or coherency issues? Probably none, since you can't prove whether the failure happened due to a CPU problem, or an issue with a Microsoft beta OS. <VBG>
Market Data 
Markets 
