Asynchronous Compute was developed w/ Cooperation from AMD, they supported it since HD7000 Series,

So HD7000, HD8000 (Mobile), Rx-200, Rx-300 and Rx-400 Series.

nVidia on the other hand didnt care much for it and didnt really work on it until Pascal Architecture.

nVidia's original excuse was their scheduler was already that efficient it didnt need Async to fill processing gaps. etc.


Which kind of makes Sense,
Using the same Engines
AMD on DX11, gets spanked by nVidia on DX11
AMD on DX12 w/ AC Off, Has a Performance boost to bring withing reach of nVidia.
AMD on DX12 w/ AC On, Has this Drastic Performance increase, overtakes nVidia, who pretty much either had no performance gain or a regression.

AMD's DX11 Drivers had a bottleneck, that's slowly been fixed over the last year or so since RTG Started.