• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

DirectCompute benchmark

My results..

HD5870/1GHz vs i7 920/4GHz

compute.jpg
 
calculating tons of FFT-like data and some memory transfers
Click to expand...

Don't think they are doing anything different - same workload regardless of the API version... in which case... scores would be comparable.
 
Says DX11 for the API version (check) and DX10 or higher for hardware (check)

If the scores are comparable the good old 200 series are stomping the 5800 at their own game...
 
These tests won't be using tessellation...

Infact if these numbers are real and comparable... you could probably move DX11 tessellation into a software emulator running on directcompute and get better performance for tessellation with the 295GTX and 200 series multi card setups than the 5870... bit of a hack but won't be the first time someones done something like that... obviously tho that would be just one feature from the DX11 set - trying to do that with the whole works wouldn't work.
 
Last edited:
Rroff said:
Says DX11 for the API version (check) and DX10 or higher for hardware (check)

If the scores are comparable the good old 200 series are stomping the 5800 at their own game...
Click to expand...
yours says D3D_FEATURE_LEVEL_10_0, my says D3D_FEATURE_LEVEL_11_0.

so what i said is right. a dx11 card is using more or all of its feature.
 
Last edited:
Rroff said:
Don't think they are doing anything different - same workload regardless of the API version... in which case... scores would be comparable.
Click to expand...

No idea, but 'something' is clearly up, look at the scores posted not in comparison to the geforce cards, but in comparison to the radeon 5700. They're no higher. I suspect the benchmark is naively written and doesn't really take advantage of vectorisation properly, hence the low scores and lack of performance scaling on the Radeon cards. That, or something is really, really wrong with AMD's directcompute implementation. I don't think that's the case though, as according to Anandtech, the 5800 series run the Nvidia wave demo much faster than the GTX200 series... I suspect it's just a bit of a crap benchmark, in all honesty.

http://www.anandtech.com/video/showdoc.aspx?i=3643&p=8
 
Yeah but according to the blurb... the test is computing the same sample data - not using different tests data for different API levels...

Its just using the GPU in a similiar fashion to folding for the purpose of a benchmark then doing some bandwidth tests.
 
You must log in or register to reply here.
Back
Top Bottom