Async compute won't be much of a factor through this current generation, as far as major and indie titles are concerned. They have to account for the 100s of millions of Keplers, Maxwells, and even APUs in the installed base. It'll matter for benchmarks, but that's more of an epeen thing. I'm still rocking SLI 980TI Lightnings @1500MHz, and they have no problem with keeping up with SLI 1080s, for half the price.
That may or may not be true, too hard to tell at the moment. And depends on what you mean by "current generation" Given the nature of the existing but more importantly upgraded gaming consoles, async will likely play a bigger role. The older GCN hardware was limited with dedicated hardware asynchronous computing, much more so than the latest GCN hardware. I suspect more emphasis will be put on using asynchronous hardware in the latest console GPU's, especially on the Neo paired with the anemic Jaguar based CPU. Also developers have had some time to get a better handle how to develop with Vulcan and DX12, so we should start seeing games created with this in mind rather than DX11 created games with crappy DX12 ports tacked on.
Doom may be one game but it's a shining example of what's capable and a glimpse at what's coming. The industry doesn't care about old Caymen, Kepler, Maxwell, and to a lesser degree GCN 1.0 based hardware, they'll design with the strengths of the consoles first and then the leftover optimzations go to PC ports. The more work required to optimise for PC ports the less likely it'll be implemented.
However Maxwell 2 based cards should fare well with some optimisation due to the quasi hardware based async implementation (Nvidia has the resources to optimise and they generally only optimise for the current architecture to keep the planned obsolescence moving) but I suspect anything older will suffer, including those 980Ti's and my 970
Last edited: