First DirectCompute GPGPU Benchmark Released

[caveman-jim]

From NGOHQ comes a member-created GPGPU DirectCompute benchmark.

A forum member by the name of Pat has recently released a new benchmark tool for DirectCompute. This tool allows you to benchmark DirectX 11 latest general-purpose computing feature by calculating tons of FFT-like data and some memory transfers. DirectCompute is an application programming interface (API) that takes advantage of the massively parallel processing power of a modern graphics processing unit (GPU) to accelerate PC application performance. Be advised that DirectX 11 and the latest display drivers are required to run this benchmark.

Download this file in our downloads section.

Bright Side of News notes:

It is also worth to mention is that the benchmark is not limited for only DirectX 11-class hardware, so owners of nVidia hardware don't need to cry foul over nVidia's inability to produce hardware according to their own roadmaps and promises.

Get your copy from NGOHQ.COM.

[Pr3tty F1y]

Originally Posted by caveman-jim

It is also worth to mention is that the benchmark is not limited for only DirectX 11-class hardware, so owners of nVidia hardware don't need to cry foul over nVidia's inability to produce hardware according to their own roadmaps and promises.

Nothing like being harsh

[Dyre Straits]

FTR, this cannot be run on pre-DX 11 ATI hardware. At least not at this time.

I tried it on my HD4770 and got a "No DirectCompute Support" pop-up.

From the Readme.txt:

"Catalyst currently doesn't support DirectCompute on DX10/10.1 GPUs"

[caveman-jim]

Which drivers are you using? If you haven't tried them you could test the Stream 2.0 beta 4 drivers from AMD.

[Melchior]

Can't say I'm too impressed by the 5770's performance in this bench. It had my 2ghz x2 pegged at 100% solid on both cores for the cpu test. Assuming this scales across 4+ cores, I think any 3+ghz quad's CPU score could match my GPU score. The GPU test did not appear to be CPU limited, cpu utilization was less than 20% during the gpu portion.

[moshpit]

I'm not getting it. My score is WAY off. It's unrealistically high.



Other 5800 series cards only scored 11,000, which is only 1/3rd my score. Even if this benchmark scales 100% for a second GPU, something is WAY off here. Also taking into account the tiny difference between the possability of a 5850 and 5870 being used, I shouldn't be breaking 30K on the GPU test? CPU score looks exactly where it should, that GPU score just messes with my head though.

Edit: CPU scaling versus GPU scaling in this test is stupid.

[Melchior]

Seriously, your i7 has at least 4x the raw performance of my 3800x2, you should be scoring like 6000-8000 if it scaled like I thought it would.

[moshpit]

Originally Posted by Melchior

Seriously, your i7 has at least 4x the raw performance of my 3800x2, you should be scoring like 6000-8000 if it scaled like I thought it would.

That's why I said the CPU scaling of the test is borked. I've looked around. There's something wrong with the test. I'm fairly sure my scores are wildly inaccurate.

[DVS Darkside]

New version of benchmark 0.25 tested with different drivers score not compatible with 0.15 keep that in mind when comparing, we can clearly see driver version plays big role in DirectCompute performance. This benchmark doesn't take advantage of multiple GPU's sadly. As for 5870 its quite likely ATI/AMD don't have mature implementation yet , according to author of the tool his blog

I don't favour any vendor, the codepath is one and the same for ATI/NV any any other DirectCompute enabled device that could be istalled in the system. It looks like this:

benchmark(compute shader code) -> DirectX11 API (compiler) -> DirectX11 API (run) -> drivers -> GPU.