3090 FTW3 ultra plus 2 1080tis in SLI

[VisArt]

This is my workstation. I’ve been trying to optimize it for Davinci resolve and I can’t find information on how an SLI bridge plays into performance. I haven’t overclocked the GPUs yet, but I think this setup is ~ the equivalent of of a single 4080 system in this Benchmark.

I’m doing some A B testing but I don’t have a lot of consecutive time to test and really feel the difference between having the high bandwidth bridge attached to the 1080Tis vs not having it. Can anyone explain?

I originally put it on there because it looks cool but I took it out to run the benchmark without it. Even though SLI was never enabled, I got a notification saying to reattach the bridge for optimal performance.

This is a very dumb setup but I wanna see how good it can be.

[nnunn]

If you have the Studio Version of DaVinci Resolve, I think it will run any CUDA kernels separately on each device (ignoring the NVLINK bridge between those old 1080 Ti’s).

If you have a non-studio version it will use only one GPU, and if “GPU Selection mode” is set to “Auto”, it likely picks your 3090 (which would explain a PugetBench score similar to a 4080).

[VisArt]

This was the score with the studio version. It utilizes all the gpus but even with all of them being used I think it’s only better for vfx.

[GigaBusterEXE]

@VisArt

SLI bridge plays into performance.

it doesn’t only games