Discussion in 'Videocards - AMD Radeon' started by Srsbsns, Aug 2, 2015.
Stock 7970Ghz Crossfire Mantle is destroying DX12. How does yours compare?
Looks like DX12 test doesn't use Crossfire at all?
lol yeah, i'm curious to see from where the problem is, AMD's drivers or M$'s API code.
Any free legal download for that benchmark? Not included in 3DMARK demo xD
No, API test is only available in paid version.
Buy or torrent.
What is strange in this benmchmark is the mantle result.
It seems to be double of what it should be.
Can you run the test forcing single GPU mode (disable crossfire)?
Why is having double score in crossfire strange? He has cpu grunt to run 24M dc. And crossfire works in mantle.
Look this single GPU API test:
Now look this CFX API test:
Mantle and DX12 are the same results more or less in single or CFX GPU mode.
something wrong with crossfire mantle is destroying DX12
Clearly Mantle supports 4 GPU's better than DX12 currently. You best report that to Microsoft/AMD.
The last update for the api test was for build 10130, the test is still not optimized for not only the 10240 windows release build but also release drivers.
If they start changing things all previous tests will be void. It's perfectly fine for 10240.
Biggest difference I see with the high mantle scores and sammarbellas scores is 4 cores vs 6 and 8 cores
I think this is not the main difference.
Obviously my 3770 locked at 4.1 Ghz will performs worst than a 6 or 8 cores Intel CPU at similar clocks.
The main difference is there is.....no difference between single and CFX results for Mantle and DX12 in my tests.
Maybe latest drivers added CFX support for mantle only.
I did the tests with 1023.7 and nobody has CFX scaling for mantle or DX12 then.
3Dmark is the same version btw.
The only change is drivers version AFAIK.
This is getting stranger by moments...
Now your single GPU results for DX11 API are similar your CFX DX11 API results!
Ran mine at 4K....everything looked good
The resolution makes no difference to the test.
I was under the impression that the API tests use your CPU not your GPU the only thing the GPU is used for is just to show you what's going on and has no bearing on the finale API score