Sniper Elite 3 and Thief
The games on this page (along with Civ Beyond Earth) provide a good example of Mantle vs. DX11 with these assorted processors. AMD's Mantle API offers enhanced performance with less system overhead, seemingly ideal for gaming with a less powerful CPU.
Sniper Elite 3
As you can see the R9 290X results using Mantle easily trump those with DX11 with this GPU, and even though the GTX 980 provided better average FPS numbers the 290X/Mantle combination provided higher minimum frame rates for more consistent play. This effect, as with Civilization Beyond Earth, was less significant with lower-end GPUs. I must also note that the result with the R7 260X + i3-4130 is unusually high, and may be the result of some other factor as this GPU/CPU combination was done last. It was repeatable, but I'm pretty sure the i3-4130 wouldn't consistently outperform the other CPUs like this.
Here are the same results ranked by overall average FPS (again, take the 260X + i3-4130 result with a grain of salt):
Now we look at the same results at 1440p. (Note that these results were ammended due to an error in the data presented in the original chart. There was a spike with some of the FX 9590 and FX 8350 results from misplaced 1080p data in the spreadsheet. This has been corrected.)
At 2560×1440 it's the GTX 980 on top again, with all three Intel processors, followed with mostly Mantle results to lead the AMD GPUs. There was not much variance between CPUs with tests at 1440p, and this is particularly evident with a look at the results down at the bottom with the GTX 750 Ti. We had a "Ti" (sorry) with a 38 FPS average give or take 0.3 with all CPUs, perfectly illustrating that a card like the 750 Ti isn't bound by the CPU at 1440p.
Another game with full Mantle support, we'll first look at the results by GPU yet again.
No surprises here as the Mantle R9 290X results are all higher than their DX11 counterparts, and the 290X bests the GTX 980 because of this. Oddly enough this is another example of a lesser Intel i5-4440 providing equal or better results than the mighty i7-4790K. Turns out this modest Core i5 processor, commonly found in commodity OEM desktop systems, is a pretty awesome gaming CPU.
One again the lower-end GPUs didn't differ much with CPU changes, indicating the bottleneck isn't with the CPU there. This will likely be even more noticeable with the 1440p results.
At the higher resolution the GTX 980 takes over the crown, with the Intel Core i7 and i5 CPUs offering better numbers. The i5-4440 again has oddly high results. I had to re-check everything, but there really wasn't an easy explanation for this. There was no overclocking or other tweaks done to any hardware, so I just have to shrug. Maybe the 4440 is secretly the best gaming CPU from Intel? (Conspiracy? Or possible random hardware weirdness. Who knows.)
Finally we'll look at these results again in order of average FPS.
Read on for the conclusion to this quad-core CPU gaming experiment!