A great point that we all should not overlook. And, the e6760 will be customized, so I think the Wii U will be closer to a perceived 1tflops than we currently assume. This is why Nintendo has officially stated that the Wii U will compete graphically in the next gen war. Unless we want to think of Nintendo as a bunch of liars, we should keep that statement in the back of our minds when we speculate on the GPU's power.
I think taking compete too literally is not a good idea.
Thing is getting out there soon enough might make them the minimum denominator, and having a appropriate feature set and modern stuff like Tesselation units can certainly decrease the difference.
But it'll be there nonetheless.
A flop is a flop, isn't?
But the last rumour saying HD8000 GPUs on next gen don't fill those 1.2~1.5 tflops...
It is, but at the same time it can be bottlenecked or not, if it has lots of bottlenecks then you'll be wasting them on stupid places.
For instance, Xbox 1 only did 4 textures per pass, GC did 8 and they needed them all the time; so GC didn't have to re-render anything whereas they had to do the polygon trick and draw the whole scene on top of itself a second time; that would effectively halve the practical performance.
That's just an example though, there's lot's of ways of taking advantage of GFlops, and loosing them cheaply.
One is the difference between SM3.0 and SM4.0 or SM4.0 to SM5.0; since they allow so much more instructions to be ran per cycle; it doesn't matter if you have more GFlops, if you are taking advantage of it in one platform and not on the other then one is getting it easy and the other one is wasting lots of resources due to the fact that it needs more to pull less.
I really hope this is SM5.0.