With all the fiasco of time spy async problem I went on to educate myself what the hell is async compute. After reading through Microsoft DX12 documents and watching some Youtube videos, I started to feel really good about the future of GCN based Fiji and RX480. DOOM Vulkan is just a taste how AMD's architecture truly shines. From the looks of it Maxwell or even Pascal are inferior construction comparing to Fiji or Polaris. Nvidia cards focus heavily on current gen graphics API and use brute force(Good old MHz) to gain the upper hand in battle against AMD.
Async compute will greatly boost AMD's design which usually features gigantic amount of stream processors. On the other hand both Maxwell and Pascal won't see much improvement by enabling Async since they are not designed that way.
There is a problem with educating oneself when there is a bias present (we all suffer from it - it's not as jab at you). The inherent problems of a compute heavy architecture are a far higher power draw than the simplified architecture of Pascal. While you say Nvidia brutes it with MHz, effectively AMD brutes it with hardware. Neither is right or wrong. Again, given the slow (but steady) uptake of DX12 it was a wise gamble by Nvidia to NOT go back to compute heavy design. Again, Given how low 1080 is on transistors and shaders, it still (due to ridiculous clock speeds) beats EVERYTHING.
Power efficient designs are favoured by the industry and OEM's - mobile especially. Pascal has that won.
Again, DX11 is still an option so even with DX12 giving AMD a good boost the gfx effects of DX11 aren't secondary. Everything GCN is good for AMD going forward - no doubt - no argument. Again, I should state I WANT Vega to be great - we all need it to be. I will buy a Vega card happily if it's better than big Pascal - I'll buy whatever is better basically, no colour bias.
This isn't as simple as AMD 'outsmarting' Nvidia. Nvidia left compute behind to go fast for DX11 (and some of DX12). If they have read things properly, they will probably make that call for Volta. Meanwhile, GP102/100 with twice as many cores as GP104 (I think I read that) will have blistering performance, Async or not.
Interesting times ahead with Vega versus Pascal proper.
PS - I really do want Vega to shine. I do. I refuse to buy a card like the 1080 because IMO it's stupidly priced.