Originally posted by: RussianSensation
Even if GT300 is faster than any ATI card ever made, including 5870 X2, unless they have low- to mid-range cards under $300 that can compete or best ATI soon, then having 1 sole fastest card is meaningless. Right now GTX 295 is faster than 5870 but for $500 a piece, no one cares. What NV needs is not only a fast performing card, but one that can scale well to lower price levels, still be performance competitive and cost effective to manufacture relative to ATI!
Presumably NV knew well enough to factor all this into their post-GT200b plans, I think this much we can safely assume they are fully capable of doing.
Performance target-wise, I can't imagine an easier thing for NV to have assumed AMD would do with their Cypress-gen chip other than to just assume at a minimum it would be a 2xRV770 performance-level for the single-GPU product (and doubled-up again for the comparable X2 SKU).
So again I think that while we don't know what the performance target was for Fermi during the past years I don't see their internal competitive assessment efforts as being surprised by what AMD fielded this fall, which would imply that Fermi's designed performance target is probably not going to be a surprise (to the downside) either.
Does any one here really think NV designed Fermi in a vacuum to the considerations that it either needs to be harvestable for cut-down SKU's or that there was a parallel Fermi Jr chip developed alongside it that natively has fewer SP's just as Cypress vs. Juniper was done? These guys are pros, we are just armchair analysts here, piss-poor ones at that I might add.