The Ashes of the Singularity DirectX 12 benchmark results has caused a great deal of animosity between AMD and NVIDIA. Previously, one of Oxide’s developers commented on the recent furore and suggested NVIDIA GPUs would struggle to utilize Async Compute Cores which should lead to greater gains using DirectX 12. Unsurprisingly, this viewpoint has been categorically shared by AMD’s Technical Marketing Lead, Robert Hallock:
“Oxide effectively summarized my thoughts on the matter. NVIDIA claims “full support” for DX12, but conveniently ignores that Maxwell is utterly incapable of performing asynchronous compute without heavy reliance on slow context switching.
GCN has supported async shading since its inception, and it did so because we hoped and expected that gaming would lean into these workloads heavily. Mantle, Vulkan and DX12 all do. The consoles do (with gusto). PC games are chock full of compute-driven effects.
If memory serves, GCN has higher FLOPS/mm2 than any other architecture, and GCN is once again showing its prowess when utilized with common-sense workloads that are appropriate for the design of the architecture.”
In basic terms, it appears that the Async Compute Cores rely on context switching which means threads are stored to allow for future executions. A good analogy is hyperthreading which works in a similar way. This process requires a large computational workload and it’s unsure if Maxwell’s core architecture will receive any benefit from Async Compute Cores. Supposedly, using this method will actually result in poorer raw performance.
WCCFTech received a fairly meagre response from NVIDIA surrounding this matter, and PR manager Brian Burke said:
“We’re glad to see DirectX 12 titles showing up. There are many titles with DirectX 12 coming before the end of the year and we are excited to see them.”
With DirectX 12 being the most revolutionary API ever devised, the war of words based upon contrasting technological approaches was bound to occur. It will be fascinating to see if the trend towards AMD gains occurs across many DirectX 12 titles. Arguably, NVIDIA feels this will become an isolated incident.
Thank you WCCFTech for providing us with this information.
Except A-Sync compute has been around since Fermi. https://www.reddit.com/r/pcmasterrace/comments/3j7646/with_nvidias_lies_blowing_up_in_their_faces/cumuubt
No DX12 support yet for Fermi.
“Yet”, being the operative word. WCCFTech are fanning the flames of another bullshit fire, and I would prefer to not see their lies on eTeknix. I come here for news, not bullshit. The title of this article is complete bullshit, a lie, a falsehood, NVIDIA is perfectly capable of A-Sync, they just haven’t implemented it yet.
Nvidia need to review there architecture be cause I had suspected they missed something in prep for dx12 maybe the software can utilise async but maybe there is a hardware incompatibility with dx12 and async running at the same time… I am a bit of an and fanboy but that is only because I like underdogs XD…. But a company overlooking something is in fact quiet known to happen.
Typical fanboi comment…have you read this ?
https://www.reddit.com/r/pcmasterrace/comments/3j131k/nv_gpus_do_not_support_dx12_asynchronous/
Typical fanboi comment…have you read this ?
https://www.reddit.com/r/pcmasterrace/comments/3j131k/nv_gpus_do_not_support_dx12_asynchronous/
Then why are you the only one writing BS?
Nvidia fakes hardware features in the driver once again? I can’t believe that.