Originally Posted by SirPauly
Maybe, but also maybe at the expense of the sweet spot and mobile, where nVidia is aiming for this time before the big guns are ready. Spend engineering resources on a robust sweet spot offering and then sweeten it with a seamless high end offering when the 28nm matures. In other words, don't make the mistakes of Fermi.
What happened with Fermi was they placed resources on the high-end at the expense of mobile and performance and down sectors and AMD did over-take nVidia in over-all discrete market share.
Absolutely agree on all points, as the volumes sold on midrange and lower end cards far outpace the high end sales to a huge degree, even though the profit margins are much lower per card....It's what pays the bills basically, mainly OEM contracts, even if it doesn't get the press that the high end cards do, wich is mostly for publicity and halo effect.
Though there is the quadro and tesla markets and i wonder how their requirements will shape the overall architecture of the chip, while still not hurting it in terms of straight gaming performance, as even the HD7970 now has built in C++ hooks, ECC memory controlers, larger caches and 1 terraflop of dual precision floating point math, and 3.7 terraflops of single precision math.....Basically capabilities that aren't really used at all as a pure gaming card, but does cost transistor budget and makes the chip larger and use more power too...