Everything I've read about the 90nm process indicates that shrinking the die size lowers the overall amount of heat, but concentrates that smaller amount of heat into a smaller area. So I would assume from this that the heat of the chip decreases by a smaller percentage than the percentage that the chip shrinks. So if the heat decreases overall by 15% by going to the 90nm process, and the die shrink from 130nm to 90nm is roughly 30%, then you have 15% less heat over a 30% smaller surface area. That is why manufacturers had so much trouble, from what I understand. But, that 15% decrease in heat means that is 15% further you can push an overclock before reaching the peak that the chip can deliver. (These figures are all theoretical, in case you couldn't tell)So why go with the 90nm Athlon 3200+ S939?
From Anand's review of the 90nm athlons, I assume that all of the 90nm athlons can overclock to roughly FX-55 speed. That's 2.6 GHz. Almost every review done on this site, if not every review, has said time and time again that the performance increase gained by moving to 1MB cache from 512KB is negligible. The 3000+ in 90 nm is just under $200, the 3200+ is about $220, and the new 3500+ is expected to be somewhere just above $300. The way I see it, the 3200+ is not too much more than the 3000+, and I won't have to push it as hard to get 2.6 GHz. The FX-55 ought to start out selling at close to $900, so saving almost $700 sounds good to me.Why not go for the FX55? What is the biggest difference here? Can you OC the 3200+ on air?