idk man... if you think back... a couple of years ago most people wouldnt even know about the mfc process at all, they wouldnt know what 150nm and 130nm mean and what yields are etc... the more time passes, the more general knowledge about the hardware we all love to play with appears in the forums and spreads... i still have hope that one day we get those numbers... i mean after some years its really not sensitive data anymore, and in case somebody still remembers and is nice enough to share it... it might happen!
like what does nvidia care if initial g70 yields would appear online now?
it would have no relevance, or well, maybe 10% relevance to GF100 and even less to future products... its only weird people like us who find that stuff interesting
and theres a lot we can guess form sales of cards when they reach their end of life...
at that point margins grow slimmer and slimmer so the actual retail price is close to the mfc cost of the card.
and that gives us a good idea of the yields at that time... for example, thats how you can figure out that gt200 never had great yields, even now that 55nm is mature...
260s only sold for 175$ for a short time, which sounds like they barely made profits at that point, and the prices went up to 200-250$ now... and 260s are the worst gt200 chips with 1 or possibly even 2 defects! and if they cant sell for less than 200$, its a big hint that 55nm gt200 yields arent great even now that the process is mature... beyond mature even... eew :S
G92 on the other hand has been an awesome yield chip at least on 55nm, they are still selling G92 cards in large numbers for 100$ and even less which points towards very very good yields...
that explains why nvidia wanted to shrink G92 even to 40nm, im sure a lot of people at nvidia love g92
and on atis side, rv770 is still selling and for very low prices, pointing towards great yields there as well...
same as rv670, which sold well into the rv770 cycle for very low costs...








Bookmarks