DiRT 3 (DirectX 11)
A continuation of the Colin McRae series, but without his name, DiRT 3 is one of the top racing games in the world and offers stunning imagery along with support for features of DirectX 11.
Our settings for DiRT 3
DiRT 3 was a game that only showed a couple of instances of runted frames in our HD 7970 testing so I was curious to see how a step down in performance levels would affect our results. Already we see that the Radeon HD 7950 CrossFire results aren't quite matching up in our FRAPS and Frame Rating testing.
Our plot of frame times shows us where those drops in observed frame rate occur on the CrossFire setup – the three big orange blobs towards the middle of the test run. The blue line that is the GTX 660 Ti SLI is definitely more consistent than the HD 7950s in CrossFire but it has some larger swings of frame time early in the benchmark. Both single cards perform quite well, however.
First, on single card performance, the HD 7950 is just barely keeping ahead of the GTX 660 Ti. The SLI results though jump from about 90 FPS to 155 FPS average which is much better than we see with CrossFire thanks to the scaling issues found above.
Keeping an eye on the scale on the left for frame time variance, even the HD 7950s in CrossFire aren't too bad, but they are the worst of this group.
The frame time issues are so bad in DiRT 3 that CrossFire HD 7950s are actually showing up as observed frame rates UNDER that of a single card!
Our frame time plot shows the large swings of the Radeon CrossFire results once again, but notice that the SLI results are actually tighter at 2560×1440; the added GPU bottleneck of the higher resolution results in a better result for NVIDIA and a worse result for AMD.
Our minimum FPS graph shows the significant scaling of the GTX 660 Ti cards in SLI and the lack thereof for CrossFire.
Only the Radeon HD 7950s have a big problem with frame time variance, but the difference is dramatic.
Observed frame rates on the HD 7950s in CrossFire are much lower again with some runts and dropped frames but the SLI results are consistent in both FRAPS and Frame Rating capture testing.
While nice to look at from a design perspective, the orange line that represents dual HD 7950 cards shows the "0 ms" dropped frames. The GTX 660 Ti cards in SLI do have some spikes and hitches but it seems pretty reasonable in real-world game play. Both single cards do quite well with tighter series' of results.
With SLI scaling from 38 average FPS to about 65 FPS, the GTX 660 Ti cards prove to be more than capable of running DiRT 3 at 5760×1080. The minimum FPS of the GTX 660 Ti SLI configuration does drop down a bit at the 90th percentile or so but it it consistently stays ahead of any other of our four options.
Because we are seeing drops rather than runts, the variance of the frames that remain for the Radeon HD 7950s in CrossFire isn't that bad – but don't think you can just look at this graph and get the whole story. The GTX 660 Ti SLI setup definitely has more variance than the single cards though is able to increase the average frame rate 71%.