I mean, yeah, but if Proton is doing an absolutely flawless job, then it has 0 performance penalty compared to Windows. All the actual gains still do come from Linux having less overhead. So, both are true, that Proton is killing it and that the gains come from Linux.
fuckwit_mcbumcrumble@lemmy.dbzer0.com 2 days ago
DXVK (which also runs on windows) alone gives you a huge performance benefit. Playing world of warcraft on windows I’ll see about a 30% reduction in CPU usage and higher performance.
Proton doesn’t just get you to almost matching Windows’ performance. Proton easily outperforms windows even on higher end hardware where windows bloat isn’t a concern.
Fubarberry@sopuli.xyz 2 days ago
While proton enables that, that’s still just vulkan outperforming DirectX.
So technically proton isn’t improving performance here, it’s just allowing the game to run on better performing systems (like Linux and vulkan).
Cethin@lemmy.zip 2 days ago
It’s just Vulkan outperforming DirectX by translating DirectX to Vulcan. If you’re comparing the default experience with Windows and Linux, how can you say Proton isn’t technically improving performance? What would you call that if a performance increase is caused by running through Proton?
Aceticon@lemmy.dbzer0.com 1 day ago
As pointed out higher up this thread DXVK and Vulkan also work in Windows (without Proton) were they give performance improvements.
Further, it’s perfectly possibly to run Windows games via DXVK and Vulkan in Linux without Proton - just use plain Wine (of which Proton is a branch) instead - and you also get the performance improvements (certainly that’s my perception in my system since I tend to get my games from GoG instead of Steam when available and thus run them via Wine instead of Proton).
So that’s at least two situations were the performance improvements are present without Proton, hence you cannot logically claim they’re due to Proton, even indirectly.
Logically the place most likely to yield performance improvements is the full implemention of a rendering stack directly on top of the hardward which even has its own architecture - Vulkan - since there’s a lot more room to improve usage of hardware resources at that level, though things like pre-conversion and caching of Vulkan shaders from DirectX shaders, which are done at a higher level (Proton or DXVK), can also improve performance.
It’s possible that Proton itself is delivering some performance improvements (for example, via the trick of, pre-converting shaders from DirectX to Vulkan before game start, uploading the generated shaders to the Steam servers and then other users just download the converted shaders and do not require that step, which should speed up game start tough I have at least one game were it actually can slow down A LOT game start because the generated shaders are massive) versus solutions using DXKV + Vulkan without Proton, but that’s not really enough to sustain a claim that the performance improvements are mainly thanks to Proton in the face of also seing the performance improvements when Proton isn’t there.