The graphics division was one of the only things that was going ok for AMD but looks like they got hit with a big blow. Wonder how Charlie will spin this one.
Btw, could the graphics division (ATi) jump out of the sinking ship that is known as AMD?
Whether or not ATI is a stand-alone, might not matter much. With HD4000-6000 series they offered price/performance and that didn't work. With HD7000 series, they now have price/performance, game bundles, overclocking and single-GPU performance crown and are still losing. If you buy an AMD card, you can do whatever you want, BIOS flash, overclock, push it hard as hell and that's still not selling to enthusiasts who seem to care more about saving 20-30W of power over saving $ and getting free games.
Even when AMD had the market to itself for 6 months with HD5000 series, it hardly hurt NV. Clearly, the issue is deeper. It's probably explained by brand value, features and drivers, as well as the negative association with the AMD brand as a whole. NV has PhysX and "bullet-proof" driver myth still persists. Remember when A64/X2 beat Intel's CPUs easily? Yet Intel still had the majority of market share. Sometimes consumers will simply buy a product because they feel safe buying a popular brand.
Think about it, AMD is losing this generation badly despite having a faster product line that costs less most of the time, and if you buy in retail, you can even get free games. So if you were a CEO of AMD, how do you propose to take away market share from NV, especially if you know GK110 will probably beat HD8000 easily? If AMD couldn't even take away market share this round, they are hopeless for 2013.
True, and let's not forget smaller size. AMD cards are still as huge as ever. The stock GTX 6xx cards are tiny by comparison, important for smaller OEM cases and HTPCs.
In Q2, AMD gained market share in the discrete space and HD7000 series cards were just as huge back then. What explains such a drastic market share erosion in Q3? GTX650/650Ti/660/660Ti all launched.
Huge DX11 *and* 40nm lead. It wasn't until GTX 460 that NVDA had something to fight back with in terms of market share, because GTX480/470 were high-end cards that few people buy and thus not as good for gaining back big chunks of market share with. But by then a lot of people had bought ATI cards, blunting some of the impact of GTX 460.
Reverse the situation. If ATI/AMD was 6-8 months late with HD5000 vs. Fermi and sub-$300 28nm GCN roll-out, they would have been crushed. NV does it, gets away with it and actually gains market share (they gained it with Fermi too). HD5850 was $269 and yet $199-229 GTX460 that launched 8-9 months later sold better. So people waited 8-9 months to upgrade to an NV card. We saw the exact same scenario this time as people waited 6-8 months for low-end sub-$300 Keplers. On our forum, the guys who buy AMD cards don't wait 6-8 months to buy AMD if NV has a superior product available. In other words, if GTX660Ti launched at $349 in February, people wouldn't be waiting 6 months to get an HD7950 for $299 for 6 months. That's the difference.