-
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
- Thread starter ShakenNstirred
- Start date
Soldato
- Joined
- 21 Oct 2009
- Posts
- 2,742
Yep, tried it about 3 times and just does this
.That's why i assumed it was trying to use DirectCompute 5.0.
at last some honesty S17448 thats what i expected about a quarter the performance
thankyou
Yep, tried it about 3 times and just does this
That's why i assumed it was trying to use DirectCompute 5.0.
Its because SLI is enabled I think - D3D11_CREATE_DEVICE_SINGLETHREADED probably breaks SLI (tho it shouldn't).
If you drop SLI it should run fine but then you will get a lot slower performance anyhow.
well you cant compare any score then unless you have a dx11 card, it must be artificial with a gtx2** as its a dx10 card
Yeah you can... with normal standard calls as it does without that option its essentially telling DX - heres this data crunch it as fast as you can - it doesn't discriminate as to which path to use.
Now if we loaded it up with stuff that made use of double precision we might see a different story
Now if we loaded it up with stuff that made use of double precision we might see a different story
Soldato
- Joined
- 21 Oct 2009
- Posts
- 2,742
I'd also assume that using a single core on an ATI card and an nVidia card isn't comparable? as they have different amounts at different speeds etc.?
Edit -
Meh, i cba disabling SLI.
Last edited:
Yeah you can... with normal standard calls as it does without that option its essentially telling DX - heres this data crunch it as fast as you can - it doesn't discriminate as to which path to use.
Now if we loaded it up with stuff that made use of double precision we might see a different story
so i cant compare the single threaded score , but you can compare the score that gives you the best results
Last edited:
that would then be comparable, no card has an advantage
Soldato
- Joined
- 21 Oct 2009
- Posts
- 2,742
Hmm, if what i said was actually right (i don't really know much about GPGPU tbh), then i really don't see how it would be comparable at all, if one company uses lots of low performance cores, whilst the other uses fewer high performing ones that will in no way be comparable.
What matter is in your nominal GPGPU or game environment when something is handed off indiscrimantly for processing what results in the fastest performance.
Single threading realitically isn't going to be used outside of examples and test apps - to keep things simple when your trying to learn the code or experiment with features.
Single threading realitically isn't going to be used outside of examples and test apps - to keep things simple when your trying to learn the code or experiment with features.
But your (unaccredited) quote isn't from the author of this benchmark. You're sneakily trying to make ATI look bad for whatever reason. If I went to Nvidia with this benchmark and asked for help, would I get it?
Thinking about it, I'd probably come away with a pot of coins, a "Huang'll Fix it" badge, and clear instructions on how to make this benchmark favour Nvidia
But your (unaccredited) quote isn't from the author of this benchmark. You're sneakily trying to make ATI look bad for whatever reason. If I went to Nvidia with this benchmark and asked for help, would I get it?
Thinking about it, I'd probably come away with a pot of coins, a "Huang'll Fix it" badge, and clear instructions on how to make this benchmark favour Nvidia
this benchmark dosnt favour nv in single threaded ,,,, but everyone with nv are now saying that test is not comparable in single lol
its also obvious that this benchmark is using single threaded on ati in both tests
Last edited:
Associate
- Joined
- 23 Mar 2007
- Posts
- 957
- Location
- Scottish Borders
But your (unaccredited) quote isn't from the author of this benchmark. You're sneakily trying to make ATI look bad for whatever reason. If I went to Nvidia with this benchmark and asked for help, would I get it?
Thinking about it, I'd probably come away with a pot of coins, a "Huang'll Fix it" badge, and clear instructions on how to make this benchmark favour Nvidia
haha xD
How is it obvious? just because the scores are similiar doesn't mean anything for deffinate.
Whats going on under the bonnet is different for single threading on a purely DX11 pipeline and DX11 software and DX10 hardware - without knowing more details you can't directly compare... even if the ATI cards are only running in single thread for the main test you still can't compare without knowing more about whats going on underneath.
Unfortunatly I don't have instrumented drivers installed here so I can't see what the ATI card is doing on the main test but if someone does have they can prolly give us an answer. (You'll need perfstudio)
Unfortunatly I don't have instrumented drivers installed here so I can't see what the ATI card is doing on the main test but if someone does have they can prolly give us an answer. (You'll need perfstudio)
Last edited:
well care to explain the very small margine of difference between my scores for multi and single