FM_Jarnis: Async compute is about utilizing "idle" shader units. Slower the card, less idle ones you have.
Ultimately some AMD cards gain quite a bit (ie. they have a lot of shader units idling while rendering and they are very good at using them for the available paraller loads). Some AMD cards gain less or not at all (either less capable at paralleriziing, less idle shader units or no idle shader units at all - for example a HD 7970 is hard pressed to have any to "spare")
Figured as much, when I see a sharp drop in performance gains for less shader GCN, I know they are heavily focused on improving shader utilization with the Async approach taken and not a true multi-engine approach.
In Doom, RX 480 gains massively, almost as much as Fury X, which is indicative of a true multi-engine design as the RX 480 has so few shaders compared and a improved scheduler, pre-fetch, etc to improve it's shader utilization already.
When you get Rasterizers & DMAs to run in parallel with Shaders, you improve performance all round. Then add fillers to improve shader utilization on top and it's a huge gain for all GPUs that is capable of DX12/Vulkan Multi-Engine.
---------------------
The idea that a 7970 won't gain with AC because it has less shaders and less idle shaders is counter-productive to the point of a Multi-Engine design.
On consoles, they have even LESS shaders, ~half that of a RX 480 or 7970. Why do the devs who use AC on consoles report the biggest performance gains?!
Because it's NOT just about shader utilization. -_-
Example here:
https://www.computerbase.de/2016-07/doom-vulkan-benchmarks-amd-nvidia/
Look at the % gains the RX 480 gets with Doom's approach, compared to the 390 which has more shaders. About the same or better. This is because it's not an emphasis on just shader utilization, or filling in the gaps, overlapping rendering etc. It's about real parallel execution, multi-engine, getting the GPU's ROPs + DMAs + Shaders to work in parallel.