DirectCompute benchmark

[gareth170]

could it be the driver??

 

[hyperst]

anyone care to post both tests with their gtx*** for both tests?


S- (GPU) DX11 SINGLETHREADED flag set / (CPU) 1 thread
M- (GPU) default dx11 device settings / (CPU) 12 threads


it looks to me ati series are running single threaded on both tests

 

[tmaxxthomson]

only getting 33k with my 5870

 

[ViRuS2k]

this program doesnt seem to be useing crossfire.

 

[Rroff]

For GPGPU multi GPU useage I think is down to the drivers not application level.

 

[Bloomfield]

anyone care to post both tests with their gtx*** for both tests?


S- (GPU) DX11 SINGLETHREADED flag set / (CPU) 1 thread
M- (GPU) default dx11 device settings / (CPU) 12 threads


it looks to me ati series are running single threaded on both tests

We can't, as it's DX11 Singlethreaded .

 

[Rroff]

nm

 

[hyperst]

We can't, as it's DX11 Singlethreaded .

so you can only run the multi threaded dx11 but not single, seems weird.

directcompute benchmark = fail

 

[Rroff]

For your information, I tried to talk to AMD about this benchmark, but AMD doesn't support consumer projects like this one, or projects like ATI Tray Tools [despite the fact they use it themselves]. It appears they only support multi-billion developers like Adobe.

 

[Rroff]

so you can only run the multi threaded dx11 but not single, seems weird.

directcompute benchmark = fail

We can run it - but currently on DX10 hardware it doesn't quite work as intended...

 

[Broken Hope]

We can't, as it's DX11 Singlethreaded .

Sure you can, the DX11 API supports DX10 cards just fine. People really need to stop thinking that because a card is DX10 that it doesn't support DX11, the only features in DX11 that DX10 cards don't support are Tessellation and Direct Compute 5.0.

 

[hyperst]

We can run it - but currently on DX10 hardware it doesn't quite work as intended...

get your scores up then as it seems that is the test that will be comparable

 

[Rroff]

Not really - as its not an option that would be used for real workloads anyhow.

 

[hyperst]

Not really - as its not an option that would be used for real workloads anyhow.

well your previous tests would not be an option for real dx11 workloads either, surely as you card is dx10 and not dx11

 

[Rroff]

Obviously if a game is using DX11/DC5 then the 200 series can't compare... if your just chucking data at DX to compute then this is what would happen.

 

[Dinkas]

For your information, I tried to talk to AMD about this benchmark, but AMD doesn't support consumer projects like this one, or projects like ATI Tray Tools [despite the fact they use it themselves]. It appears they only support multi-billion developers like Adobe.

So..?

 

[hyperst]

We can't, as it's DX11 Singlethreaded .

blatant lie, obvious no one wants to post their single threade scores as your getting beat by ati so badly

 

[Bloomfield]

the only features in DX11 that DX10 cards don't support are Tessellation and Direct Compute 5.0.

Well since it's a DirectCompute benchmark i'd assume it uses directcompute 5 when running dx11 mode, unless it goes to 4 when there's a dx10 card being used.

When i run it on DX11 singlethreaded it just stops responding.

 

[Rroff]

Just so typical of ATI.

People moan about the nVidia Way It's Meant To Be Played platform but they do atleast get behind developers and give them the support they are asking for - usually even smaller developers.

Unless your one of the small number of developers deep in ATI's pockets they just don't wanna know.

 

[Rroff]

blatant lie, obvious no one wants to post their single threade scores as your getting beat by ati so badly

Currently the single threaded option on DX10 hardware does not work as intended for DC4 producing results that can't be compared or just crashes - hard locked my machine on most attempts.