NVIDIA 378.09 Linux Driver Benchmark Tests On A GeForce GTX 1080

Last week marked the debut of the NVIDIA 378.09 Linux driver beta. While the release notes didn't mention any widespread performance improvements, an individual or two at least in the forums seemed to think it did and have already been inquiring why I wasn't yet using this new (beta) driver in my Linux benchmarks. Anyhow, here are some 375 vs. 378 Linux driver tests.

As I had anticipated, there aren't any big performance gains to note when switching from the NVIDIA 375.26 to the NVIDIA 378.09 driver... I've run OpenGL, Vulkan, and OpenCL tests on a few systems to find basically the same result of the performance being unchanged.

When testing with a GeForce GTX 1080, the changes were basically flat.

the biggest change I saw was...

A slight improvement for SHOC's FFT SP test.



But with many other benchmark runs, there was no real change in performance.
Really unless affected by a particular bug, you're probably only likely to see better performance with the NVIDIA 378 Linux driver if using a game/application that already supports the ARB_parallel_shader_compile extension. The NVIDIA 378 driver also enables the OpenGL threaded optimizations by default, but most Steam games and such were already shipping with this environment variable set by default. So while it's fun to upgrade to the NVIDIA 378 Linux driver for the several new features, don't expect overall better performance.