I have used the HD7970 in my review that is why i have higher fps than the 6950.
No, that is not certain. I'm pretty sure you will have significantly lower avg fps than 80 with the 7970, too. And honestly, I don't believe your benchmark with 3.2 and 4 GHz. No way that you could reproduce the first run so accurately over a full 325 seconds with 16 units, camera movements and the AI.
A more reliable method is the following (at least works on Intel CPUs):
- Run the battle, set the camera to an interesting scene, press P for pause, make a screenshot with Afterburner OSD enabled. Alt tab out of the game.
- Go to Windows energy options and lower the maximum processor state from 100% to a lower value. This decreases the frequency (no reboot required, you can do this on-the-fly
- Go back into the game, check the CPU clock via the OSD and make another screenshot
This gives you completely repeatable results between different clock speeds. It's not a complete benchmark run, but that is impossible to reproduce anyway.
Edit: also i have run the same bench of my review with the HD6950 and i get almost the same performance as the battle i recorded.
See above.
Consider the following.
Perhaps the GTX580 needs more CPU performance than the HD6950 or the HD7970.
The game is faster with the AMD cards, dont forget it is a Gaming Evolved title.
Now we're getting somewhere. That would be interesting to investigate.
From the data i have gathered with the HD7970 in those 9 games the conclusion is that most of the Games are GPU limited at 1080p. Now, perhaps if i use a GTX680 the results may be different but the bottom line is that those games when played with the HD7970 are GPU bound.
Substitute "games" with "timedemos" and you are right.
Edit 2: I will bench the same battle with the FX and the core i5 2500K with the HD6950 in the weekend and ill let you know of the results.
How will you do that? You cannot reproduce the battle accurately enough. IF you commanded all the units the same way and IF you moved the camera the same way and IF you zoomed the same way and IF the AI reacted the same way, THEN ok. Alot of if's...
Edit:
I ran the integrated benchmark "game settings" with and without SLI.
Without SLI: 25.8fps
With SLI: 50.5fps
100% GPU limit, GPU load on both GPUs always 99%. The highest load on any CPU thread during this benchmark was around 60%. During actual gameplay, this is quite different. The CPU load is much higher, often close to 100% even with 4GHz, as you can see in my screenshots. The benchmark suggest you might run the game just fine while in fact you may not (in the thick of battle 20-30fps in my screenshots with the exact same settings).
I don't know what else I should tell you. It is completely logical. Just analyze what the benchmark actually shows you and you will realize how different this is from actual gameplay:
Few units in the field of view. Often closeups of only few soldiers. No action. No physics, AI, pathfinding.