What about https://gfxbench.com/benchmark.jsp (GFXBench) anyone tried it? also there's https://github.com/vkmark/vkmark (VkMark).

Last edited by Koopacabras on 11 September 2020 at 9:12 pm UTC

GFXBench run only opengl on Linux. But, we can may be use DXVK. I didn't try vkmark because there is no screenshots

Just use games. It's better.

Devil May Cry 5 has an intro that uses in-game graphics. To access it, just don't press anything in the "PRESS ANY BUTTON" screen for about 30 seconds.

Another real world Linux Vulkan benchmark is Wreckfest. Finish 1 race then let the replay loop infinitely.

Not really a benchmark proper, but I use The Witcher 3 in Wine+dxvk on max settings. It saturates the GPU 100%.

Then measure performance using DXVK_HUD and VK_LAYER_MESA_overlay.

Last edited by Shmerl on 16 September 2020 at 5:51 pm UTC

Oh right, in-game benchmarks are a thing. Shadow of the Tomb Raider has a built-in benchmark mode if you want something native. As well as some other Linux+Vulkan games like Rise of the Tomb Raider, The Talos Principle and Dirt Rally, but those won't tax your GPU as much.

In my experience Rise of the Tomb Raider is better benchmark than Shadow Of the TR, with Shadow of the TR there is not much difference accross gpus with my previous gpu (570) I was getting 52 fps on medium, and with this 5600xt I have more or less the same score, the thing that changes if I crank the settings to high the 5600xt that's when I see some difference. But Rise of the Tomb Raider the difference is more easily noticeable on all type of settings.

On my side, I prefer use a dedicated app to benchmark instead of in game or benchmark in game. Because an app is developed to this purpose, optimized, with a low cpu consumption and also it's start faster than a game

Quoting: VinouchOn my side, I prefer use a dedicated app to benchmark instead of in game or benchmark in game.

hashcat does a fine bit of 'straight numbers' GPU benchmarking i would think.. And it doesn't even rely on having any GUI stuff running..

$ hashcat --benchmark
or
$ hashcat --benchmark > /tmp/example_gpu_benching.txt
to store the output for reviewing and to "diff" against when overclocking..

( example output: https://hastebin.com/osunebaxey )


With hashcat in combination with sensors , and a bit of grep, awk etc. , it'd probably be a relative breeze to script a speed vs temp. GPU benchmarking script thingy...



Last edited by Duck Hunt-Pr0 on 19 September 2020 at 7:35 pm UTC

Quoting: Duck Hunt-Pr0

Quoting: VinouchOn my side, I prefer use a dedicated app to benchmark instead of in game or benchmark in game.

hashcat does a fine bit of 'straight numbers' GPU benchmarking i think.. It doesn't even rely on any GUI running..

$ hashcat --benchmark
or
$ hashcat --benchmark > /tmp/example_gpu_benching.txt
to store the output for reviewing or to "diff" against when overclocking..

example output: https://hastebin.com/osunebaxey

These compute workloads aren't even remotely similar to how games or 3D graphics benchmarks make use of your GPU, so the results aren't very relevant. And the OP is specifically asking for a Vulkan benchmark.

Quoting: tuubiThese compute workloads aren't even remotely similar to how games or 3D graphics benchmarks make use of your GPU,

Oh..My bad .. I just happened to see the "[...]in order to adjust my overclocking" bit, and was thinking faster is better.. for 3D graphics in games as well as 'compute workloads'

Might i suggest checking rendering times (of an identical scene) using Blender? Or is that too perhaps something unrelated to 3D graphics performance?

Last edited by Duck Hunt-Pr0 on 19 September 2020 at 8:08 pm UTC