AMD/ATI missed a window of opportunity to "flood to zone" with aggressively priced dx11 parts and significantly increase market share. Was it really because Nvidia reserved all the excess 40nm production? Talks about a desperate, rear-guard type of action, yet...it sort of worked didn't it...in terms of being the most effective Nvidia could do during such a vunerable period?
Yet it was AMD that decided to outsource and share fab space with Nvidia. Seems like a huge management blunder.
1) They were already sharing fab space with NV since forever, because of the nature of the product and of fabs.
1b) GF/AMDs fabs can't make GPUs because the GPUs aren't designed for the process.
2) The only way to flood the market with aggressive pricing would be at the cost of profit. AMD launched aggressively priced parts in many ways, and then RAISED prices because they obviously felt they could get away with it.
2b) AMD and NV have both been supply constrained as the TSMC 40nm process ramps up, so they only way they could gain marketshare as a % would be to produce only smaller die products and forgo the high end (due to the supply constraints).
3) NV have been producing 40nm chips since long before the GTX480. The GTX480 was the first DX11 40nm chip, and their largest 40nm chip, but the volume market had been supplied with 40nm GPUs from NV for a long time in the form of lower end DX10 chips.
3b) NV basically switched their production from the low end chips to the higher end DX11 parts when they were finally getting ready for release, so they didn't have the supply reserved, they were actually just switching over their already booked supply.
4) NV ships more units than ATI anyway so it's not surprising they would have a greater preference at TSMC since they are a larger customer, ATI also has the opportunity to move to GF going forwards (when GF have an appropriate manufacturing process), so it would harm TSMC in the long run to alienate NV, while with ATI it's not (IMO) such a big problem, and they were only really maintaining the status quo anyway.
5) The increase in 40nm production (with new NV 40nm cards with larger dies) comes also as TSMC are getting more wafers produced and with better yields, so NV could indeed probably book that part ahead of time in anticipation of future products.
Both companies talked about supply constrains easing as the year went on and TSMC managed to ramp up their 40nm production.