I 100% disagree. Nvidia started with a massive die, shrinked things down a bit. AMD started with a small die, and had room to make it bigger, possibly up to 400mm. GF110 was just enabling everything, raising clocks, and shaving away some things, partially not used, partially CUDA related. This got more performance due to enabling everything, aswell as the higher clocks. 10-15% higher. Less energy came from what they cut down and such. AMD seems like they are trying to get about 20% more die size than the 5870 with the 6970.
Well, didn't you already explain why? nVidia hadn't enabled all functional units yet.
AMD had, so they had to add them to the die, growing the size.
Cayman is really a fairly new arch.
Says AMD... they don't say WHAT is new though, which gives me the impression that there's not too much to talk about.