Joe, I think one piece of the information (102 mm^2 dor 512KB and 114 mm^2 for 1MB L2) is wrong.
Both were definitely given from AMD to the press/public, but my first (conservative, as always) reaction was that, as the 102mm^2 is the most recent, maybe Opteron will turn out to be bigger than first thought. Doesn´t look that likely after thinking about that possibility. There was also the suggestion that removing the 2 HT links and some other minor optimizations would allow Winchester to be a 1MB L2 cache part @ 102mm^2. Considering what INTEL has in the pipeline against Winchester, that might turn out to be necessary.
I tend to just accept the given sizes for now as valid (including the 1MB/512kb L2), both look acceptable (good) from a manufacturing standpoint, so we´ll see.
Market Data 
Markets 
AI
