Originally Posted by MoorishBrutha
You do understand that Nvidia lied to their 980Ti customers about being fully-supported of DX12 while knowing they didn't have one of the main features of DX12- Async Compute
Ashes of Singularity wasn't a fluke, it was a reason why AMD smoked Nvidia on those benchmarks.
You mean "smoked" using updated drivers vs no drivers? You do remember what happened when nvidia brought updated drivers right? There was a nice 180 in terms of performance.
Btw, AMD are doing the same thing with transference calculations.
No one rising up to the pitchforks about that for some reason.
The 980 TI will fully support all DX12 API. That is what they promised. I'm not sure why lie about it if it ends up being true.
quote name="MoorishBrutha" url="/t/1590939/wccf-hitman-to-feature-best-implementation-of-dx12-async-compute-yet-says-amd#post_24883183"]
I think some Nvidia shill said so but alot, not just one, benchmarks saw bad performance from Nvidia due to the lack
of Async Compute Engines in their GPUs. It's confirmed via White papers that none of Nvidia's cards (Maxwell, Kepler, Fermi) have Async Compute Engines built inside of them.
The queue engine is in software, but the wrap scheduler are in hardware. So they have "partial" of overall compute engine in hardware.
Since AMD's "lead" in async was gone, regardless of nvidia doing it in software, currently it doesn't seems to be an issue.
Unless an early alpha version of the game shows up, use optimised AMD drivers but not working nvidia drivers, AMD will call "we are the winners!" and a couple of months later an updated driver comes from nvidia which again remove AMD's lead.
AMD are very early to declare things based on alpha or beta versions in order to establish a "win" regardless of the end results on release date.