birdieWhy would a hardware vendor use a feature which makes its products perform slower than without it? Can we stop with AMD/Asynchronous Compute fanboyism? D3D12 is not about Async Compute - it's just one of its features. And not the most crucial one. In fact D3D applications may run just fine when Async Compute requests are executed synchroniously. I vividly remember how everyone hated NVIDIA for using Gameworks which still used standard D3D11 features yet made a better use of NVIDIA hardware. Now we have the same situation with D3D12/AMD and everyone has suddenly forgotten this recent vendor specific "debacle" and praises AMD for basically becoming NVIDIA of the past. Ew!

I'm really glad 3DMark "noticed" Vulkan. Being them I'd even make it a primary benchmark, but then I understand they don't want to be enemies to Microsoft.



Pascal does supportAsync Compute. End of story.

If you had taken the time to read my original comment or further in my 2nd comment you would have realized that I never said Pascal couldn't do Async. The link you provided is to another one of your posts on the forum and I have no idea what your intention was with it. Excuse me I guess for my poor choice of words, I forget that I'll have fanboys nipping at my buttox at the slightest non decisive wording on anything Nvidia related.

Gameworks does not make better use of Nvidia hardware, it simply applies insane amount of tessellation (far past the point of visual impact) to criple AMD's cards. AMD makes technology like TressFX that works flawlessly on both AMD and Nvidia so the green team doesn't really have an excuse.

XzibitThe 1060 isn't showing any gains. Does The Division even make use of async compute in the DX12 patch ?

If a card utilises it's hardware 100% effectively, or close to it, in DX11, it cannot perform better using another API. We know Nvidia have got DX11 coding gown to a very fine art and squeeze everything out of the hardware. Conversely, we accept AMD takes a lot longer to get the most out of their hardware.
Under Async favouring conditions, the API can better make use of the AMD hardware and therefore it's far easier to code for.
It's easy to see how this all works. Nvidia cannot make a card that runs near 100% as designed at release, perform significantly faster under any new condition.
AMD, who might only run at 90% of design parameters get a much greater uplift with more refined drivers and/or DX12 and Vulkan.
Looking at the far smaller gains in DX12 the 480 got compared to DX11 against the 1060 (in that recent revisited review) backs up that hypothesis to a degree.

the54thvoidIf a card utilises it's hardware 100% effectively, or close to it, in DX11, it cannot perform better using another API. We know Nvidia have got DX11 coding gown to a very fine art and squeeze everything out of the hardware. Conversely, we accept AMD takes a lot longer to get the most out of their hardware.
Under Async favouring conditions, the API can better make use of the AMD hardware and therefore it's far easier to code for.
It's easy to see how this all works. Nvidia cannot make a card that runs near 100% as designed at release, perform significantly faster under any new condition.
AMD, who might only run at 90% of design parameters get a much greater uplift with more refined drivers and/or DX12 and Vulkan.
Looking at the far smaller gains in DX12 the 480 got compared to DX11 against the 1060 (in that recent revisited review) backs up that hypothesis to a degree.

I'm was curious to know how he came to such a conclusion while pointing to those numbers. Pascal is in both 1080 and 1060 one shows gains the other doesn't. My other interest was does he or anyone know to what extent its being utilized in the patch ?

What we could be seeing are other improvements of DX12 that aren't async compute related.

If one looks at the source for the comparison.

Pascal which is both in the 1080 & 1060
The only time DX12 is faster on a Pascal card is under 1080p with the 1080 only. At 1440p and 4k the Pascal cards show a negative result compared to DX11.

ComputeBaseNvidia graphics cards benefit only from the CPU limit of DirectX 12 in The Division. In the GPU limit, the low-level API or the driver is slower than under DirectX 11. AMD's graphics cards benefit both in the CPU and GPU- Limit of DirectX 12. It is striking that the Polaris GPU seems best to deal with DirectX 12. Here, probably more modern architecture than with Fiji is an advantage.

I'm curious why he came to the conclusion he did with such an example.

XzibitI'm was curious to know how he came to such a conclusion while pointing to those numbers. Pascal is in both 1080 and 1060 one shows gains the other doesn't. My other interest was does he or anyone know to what extent its being utilized in the patch ?

What we could be seeing are other improvements of DX12 that aren't async compute related.

If one looks at the source for the comparison.

Pascal which is both in the 1080 & 1060
The only time DX12 is faster on a Pascal card is under 1080p with the 1080 only. At 1440p and 4k the Pascal cards show a negative result compared to DX11.



I'm curious why he came to the conclusion he did with such an example.

I think a lot of Pascal's different response to DX12 or Vulkan is the core counts. it's a very compute lean design and the smaller core cards like 1060 (1050 as well) must suffer as compute (via Async and equivalent) is handled less by driver and more by inherent design. That, IMO, is why AMD is so much better going forward on these mid range to low range cards.
At the other end, looking at 1080 and especially Titan X, the core count is pretty high so can actually deal with the compute side better. In my simplified view, Pascal's metal API prowess is bottlenecked by low core count. A higher core count Pascal pulls away from the field.

My response hasn't dealt with the conclusion of Computebase but I think it's hard for anyone to give a concrete answer without Nvidia themselves explaining it, which we know they won't do.