I cannot wait for GamersNexus to agree on a testing framework for Linux and then see how many games will run actually better on Linux than on Windows, either native or through Wine/Proton.
Games which run on Vulkan / OpenGL don’t have any GPU translation overhead, and some run straight-up better via Proton than they do on Windows. Doom 2016 does for me, for instance.
Of course, that game is so well optimised it’s the difference between 140 fps and 200+ fps, which is not terribly obvious, but even so.
Doom ran at 100+ fps at 4k on my 1070ti with graphics maxed out. It’s hard to tell what optimization allows it, but the game runs way better than anything else that looks at least as good.
It’s not one big optimization, it’s a product of Id actually having some of the best UE developers on the planet being able to tweak the engine to run like a beast. Each level is crafted from the ground up to allow for some sweeping optimizations revolving around actor loading and culling, and the game uses proper light baking to allow raytracing to handle marginal calculations instead of explicit path tracing every shadow. It’s a lot of little things that all take impressive amounts of skill and management to pull off effectively, a lot of this stuff is implemented poorly in other games and it shows.
I can’t wait to see their content on the fediverse. They made a video about getting away from big tech but don’t mirror their stuff here. I think it’s a damn shame.
You would think so, but Windows 11 is so bloated out so badly that I actually got much better performance on Linux for most games I play. I’ve only found 2 games so far that run better on Windows than via Proton on the same hardware.
Windows 11 24h2 update completely screwed all game performance for me so badly I had switch.
Sometimes you even get fewer bugs on Proton.
One of the best examples was the release of FF7 Remake - it had really bad stutter on release on Windows… but not on Proton.
People even used DXVK (which is part of Proton) on Windows in an attempt to fix it.
I know it happens, but it’s rare. Other example is Nier:Automata after valve created super fast shader compiler for AMD cards - game is so unoptimized that saving CPU cycles on shades not only compensates for overhead but also exceeds windows performance.
From the video it looks like they’ll test on Bazzite, it’s probably more stable than Windows for that, just choose a snapshot and it’ll always be the same
Really the only factors in software are kernel and compatibility layer. Everything else is not a huge factor in Linux; this is mostly akin to saying “we need to test games with every different windows app running in the background”.
Of course for individual machines there will be external factors that users themselves need to consider (like don’t be doing Blender renders in the background lol) but there should be a huge difference between distros.
Perhaps custom desktop managers should be tested along with KDE and GNOME, but I’m honestly not sure much even those factor in.
I cannot wait for GamersNexus to agree on a testing framework for Linux and then see how many games will run actually better on Linux than on Windows, either native or through Wine/Proton.
Didn’t they just announce this? Or are they still deciding on the “how” and not the “if”?
Games which run on Vulkan / OpenGL don’t have any GPU translation overhead, and some run straight-up better via Proton than they do on Windows. Doom 2016 does for me, for instance.
Of course, that game is so well optimised it’s the difference between 140 fps and 200+ fps, which is not terribly obvious, but even so.
Doom ran at 100+ fps at 4k on my 1070ti with graphics maxed out. It’s hard to tell what optimization allows it, but the game runs way better than anything else that looks at least as good.
It’s not one big optimization, it’s a product of Id actually having some of the best UE developers on the planet being able to tweak the engine to run like a beast. Each level is crafted from the ground up to allow for some sweeping optimizations revolving around actor loading and culling, and the game uses proper light baking to allow raytracing to handle marginal calculations instead of explicit path tracing every shadow. It’s a lot of little things that all take impressive amounts of skill and management to pull off effectively, a lot of this stuff is implemented poorly in other games and it shows.
I can’t wait to see their content on the fediverse. They made a video about getting away from big tech but don’t mirror their stuff here. I think it’s a damn shame.
They could even host their own instances.
That would be nice!
Not many because there’s still translation overhead - unless you have very good CPU, the results will be slightly worse.
You would think so, but Windows 11 is so bloated out so badly that I actually got much better performance on Linux for most games I play. I’ve only found 2 games so far that run better on Windows than via Proton on the same hardware.
Windows 11 24h2 update completely screwed all game performance for me so badly I had switch.
It actually happens more than you’d expect.
Sometimes you even get fewer bugs on Proton.
One of the best examples was the release of FF7 Remake - it had really bad stutter on release on Windows… but not on Proton.
People even used DXVK (which is part of Proton) on Windows in an attempt to fix it.
I know it happens, but it’s rare. Other example is Nier:Automata after valve created super fast shader compiler for AMD cards - game is so unoptimized that saving CPU cycles on shades not only compensates for overhead but also exceeds windows performance.
Again it’s rare and relates to poorly coded games
Im not sure how they even would make a testing framework. Its not like windows, where you have the os as standard and then just swap parts to see.
Its so fragmented the amount of combinations is mind-boggling. I guess they choose the 3 most popular and just run a limited series of hardware tests?
From the video it looks like they’ll test on Bazzite, it’s probably more stable than Windows for that, just choose a snapshot and it’ll always be the same
Really the only factors in software are kernel and compatibility layer. Everything else is not a huge factor in Linux; this is mostly akin to saying “we need to test games with every different windows app running in the background”.
Of course for individual machines there will be external factors that users themselves need to consider (like don’t be doing Blender renders in the background lol) but there should be a huge difference between distros.
Perhaps custom desktop managers should be tested along with KDE and GNOME, but I’m honestly not sure much even those factor in.
Sharing a common kernel is probably why support is so vast, and then people are using the same Vulkan tools or Proton etc.