Originally Posted by Alatar
So with the DX11 efficiency talk I decided that I would do some small scale investigating.
Now since I do not have a 290 or a 290X I was forced to use my 280X. I OC'd the 280X and I downclocked my Titan so that they both offered about the same amount of GPU performance:
1100MHz 280X: http://www.3dmark.com/3dm/2734474
820MHz Titan: http://www.3dmark.com/3dm/2734934
Now since Star Swarm at stock settings is really inconsistent and doesn't offer similar results between different runs I used a custom scenario with a custom camera angle. This means that everything happens exactly the same between the runs and is viewed from the same angle as well. This means that Star Swarm works in the same way as any other normal bench. So running the custom scenario makes the results comparable between each other.
Here's the scenario if anyone else wants to test:
ScenarioCustom.csv 3k .csv file
Then I started testing both of the cards in the Star Swarm bench.
Then the Titan:
Results (avg fps) :
280X DX11: 31.69
280X Mantle: 42.52
Titan DX11: 39.3
Now remember that with these clocks the GPUs have approximately the same amount of GPU power (within a couple of per cent). And despite this:
Titan is 24% faster
than 280X in DX11
280X is 8% faster
than Titan when 280X is running mantle and Titan DX11.
Again, the cards are clocked so that GPU performance is comparable.
Even in windows 7 there's a very clear advantage in favor of Nvidia when it comes to DX11 CPU overhead.
Check your own test numbers.
280X DX11 avg unit count 3961
280X Mantle avg unit count 4436
Titan DX11 avg unit count 4057
It most certainly isn't a directly comparable scenario (280X under mantle had more stuff going on on average). Even with the same amount of units I doubt we could possibly have identical twin runs. Mantle is obviously nondeterministic (AI does whatever it sees fit between different runs).