Misha said:
Vapor-x looks nice, but I’m just wondering that in your graphs, somewhere is reference card faster for a few frames? But for example, this is really huge:
Those are just the minimum FPS, not the average, so they're not quite as important. The issue is that with games like GTA IV, no two runs are identical. This for the most part is a good thing, because it means the game isn't 100% scripted. It's a bad thing for benchmarking, though, because variances like these will occur. But in a case like this, I chalk it up to a non-issue.
I tend to only re-test if the average FPS isn't as I'd expect it, because for each time I re-run a test, it's minutes more out of my day, and given that I have 7 games to benchmark, at 3 resolutions each, at least 2 times over, that's a total of 42 hands-on benchmarking runs, per GPU I test. That's a minimum, because once in a while a game will throw a curveball and force me to start over.
I re-ran the tests though, and this time I got 38 / 67 and 39 / 66 (Min / Avg). See just how sporadic this game is with its results? This is the worst game of the bunch for this reason, but I still like to run it because it's such a total glutton. When games like Modern Warfare 2 get here, I'm going to consider replacing GTA IV with it.
Relayer said:
I would have been curious to see what kind of OC you could have gotten, and the corresponding performance, using 3rd party aps on the 5870 (like rivatuner possibly?).
The reason I haven't is simply because it's complicated. I have never been successful in overclocking with RivaTuner, and I don't really want to get into BIOS overclocking (if the 5000 series is even supported). The best solution would be for AMD to loosen it's ridiculous restrictions. I haven't touched NVIDIA's control panel overclocking for a while, but I'm going to jump to conclusions and assume it's nowhere near as restrictive as AMD's.
That aside, thanks for the nice comments guys... they're much appreciated
