Part 2: https://www.lttlabs.com/blog/2025/06/30/is-2025-the-year-of-the-linux-desktop-part-ii
LTT Fourm discussion as well https://linustechtips.com/topic/1616595-lttlabs-article-is-2025-the-year-of-the-linux-desktop-part-i-ii/
They approached this from a noob perspective and the benchmarks seemed pretty rough. The blog has an overall positive tone on linux which is nice even though it got murdered in performance.
I’d like to see a follow up with optimizations, get some of the linux community involved to help setup an optimized linux test bench to go toe to toe with their “golden image” windows 11 benchmark setup.
They benchmarked a few distros against each other and it was very samey which I expected, the real difference is between the drivers/kernel and desktop environment since most distros come very light in terms of installed software.
The fake shocked-face thumbnail trend is so tiresome.
Usually I’d agree, but in this case aren’t they mocking that trend by copying this meme?
Its a blog and thats just the blurred background because its a meme.
the benchmarks seemed pretty rough
Do they? I think they looked pretty good except for when raytracing was turned on.
The benchmarks go against the narrative that Windows and Linux are pretty much equal in performance. I’ve read regularly that Linux is “often” faster than Windows for gaming, especially from more recent Linux users.
5 years ago, 15% performance difference were the expected performance loss through DXVK and wine/proton, so these benchmarks would’ve been the expected result.
I’d argue that if you buy an Intel or Nvidia GPU when you want to run Linux, you are actively sabotaging yourself. Clearly those two don’t make good Linux drivers as the performance discrepancy is way larger compared to AMD/Radeon. For radeon the article says:
Ubuntu was 8% slower than Windows
(Leaving out raytracing bumps that to a 5.7% difference btw.)
Which seems ok to me. Especially considering that they mostly tested newer titles. If you check older titles, many of them perform on par or better on Linux than on Windows. For example notoriously World of Warcraft.
Quality can also be percieved by frame stability. 1% lows for example, which manifest in occasional hickups. While the average fps might still be equal, the game might feel less laggy if the framerate is more consistent. Which was not measured here at all. Other sites benchmarks (again for older titles) show less hickups iirc.
So the statement can be true depending on what you play and how you measure. It’s very difficult to make a study that is unbiased and reflects the real world experience of most people. Some will have great results and talk about them online. Others will have bad results and talk about them online.
Bonus edit:
On windows I remember it starting a background update while I was playing an online game and it started lagging, even though the fps stayed the same. And there was no way to prevent these background processes from running when I was playing. This would not be captured in such a benchmark as well, because they update everything before running the benchmarks. But it would frequently happen in real life use. For me this was the killer UX difference that made me fully move to Linux 9 years ago.
As an Intel/nVidia user when I switched (ie I already owned them). I can say that in real world gaming - ie not synthetic benchmarking, my vibes based analysis is that some games are definitely less fps, others I can’t notice, and a few i’m quite sure performed better.
If I were buying new I would probably go with AMD, but from my experience being team blue & green should be no impediment to moving.
Forget Linux vs Windows, the real question I have is why is Black Myth Wukong so poorly optimized that on a 4060 or 5060 it can’t reach 60 FPS at 1080p even on Windows? Freaking unacceptable.
You would think coming from the mobile world, Game Science would be used to low-spec hardware, e.g. phones, but this game can’t run well even on pretty new GPUs on PCs??? I had no idea this game had such abysmal performance. I wish people hadn’t bought it, so it could’ve flopped
I was hoping the Steam Deck would incentivize devs to have a lower spec target for their games, but so far I don’t think it had a huge impact…
I remember seeing a lot of benchmarkers stating that Linux offered better and consistent frame times compared to Windows which caused for a smoother gameplay. I wish to see that metric in their testing.
I was actually considering getting an Intel GPU since my current one broke, didn’t knew the sitation was still this bad.
The problem is they are using ubuntu 24.04 which runs on an older kernel by default. Im a mainline user so unsure which kernel ubuntu uses now but i think its 6.4 which lacks a lot of intel driver fixes and support for battlemage. I cant blame them for not knowing or specifying, the only person knowledgeable for this is wendel and his team (level1techs).
Outside of mainline, ubuntu has a 6.14 kernel build they support but you have to actively look for it in the ui.
It is however, a sign for me that ltt is not suited for this and should have waited and investigated more before posting. Or maybe asked external help to verify stuff idkIt might not be, there could be fixes out there that weren’t included. Idk I dont have one so I can’t say.
Lets talk linux gaming optimization. For them its probably best to take a distro like cachyOS out the box. The game should run inside gamescope with the desktop environment killed Feral gamemode enabled. Minimal services running in the background. I dont think they should go full HPC engineer (although that would be sick) but linux is all about taking full control of your software and hardware and pushing whats possible.
