I've heard around the Epic Games Forums that optimal performance in UT3 is gained by lesser-used SLI rendering modes. I tend to dip below 30 frames per second at times despite the power of my dual EVGA Geforce 7950GT's, so I thought I might test the theory.
If you're having crash issues, please read my UT3 Troubleshooting page for tips on debugging the beast. The best guide to tweaking every aspect of UT3 is by far is at tweakguides.com, but if you are wondering about actual benchmark results and SLI performance in particular, read on.
My first test was done with my fail-safe Forceware driver 169.13 in the default SLI performance mode, "Single GPU." Subsequent tests were done with nVidia Forceware 169.28 and the same battery of 6 tests were run in all available SLI modes: Single GPU, Split-Frame Rendering, Alternate Frame Rendering 1 and Alternate Frame Rendering 2. Review my raw data here
I know I don't have much sample data. But overall, there was not a significant difference between rendering modes. In the end I caved and switched to AFR 2 mode because it performed well enough and it was easy to set an application preference for ut3.exe as the standard for future driver comparison.
| Driver Version | SLI Mode | Avg FPS | % Above 30FPS |
|---|---|---|---|
| 169.28 | Force Alternate Frame Rendering 2 | 56.40 | 92.50 |
| 169.28 | Force Alternate Frame Rendering 1 | 61.61 | 91.42 |
| 169.28 | Split Frame Rendering | 56.74 | 91.15 |
| 169.28 | Single GPU | 58.04 | 91.58 |
| 169.13 | Single GPU | 52.33 | 89.33 |