Originally Posted by CrazyElf
What Mahigan is saying is that historically, Nvidia has relied heavily on driver based optimizations. That has paid handsome dividends for DX11 performance. However the way they have designed their architecture - serial heavy, means that it will not do as well on DX12, where it more parallel intensive.
And yet even in this benchmark, which would appear to be well-suited to AMD's hardware, the Fury X is still neck and neck with the 980 Ti. The real issue the benchmark highlights, to me anyway, is not that AMD's DX12 performance is so good, but that their DX11 performance (particularly in this example) is so bad. I don't see how you can look at this one benchmark and draw the conclusion that AMD has a better architecture for the future, when their premier card is still tied with Nvidia's (likewise for the 980 and 390X).
Edit: And if anyone is curious, here are the Gen 3/Gen 2 results for the 3DMark API test on a 290X: (Gen 2 / Gen 3)
DX11 ST: 1,298,473 / 1,265,204
DX11 MT: 1,324,209 / 1,311,353
Mantle: 18,476,350 / 18,917,266
DX12: 17,774,164 / 20,542,867
So Mantle and DX12 are faster, but somehow DX11 is slower in Gen 3.