Quote from: captainobvious on March 03, 2026, 13:31:13So even with all the bloat, telemetry and AI surveillance Windows is still faster when test is biased towards CPU @ 1080p.

Well, not really, because for 1. I think it's fair to say most people would call that equal performance in that circumstance as it's within the margin of error, and for 2. It's not a one-to-one comparison, as Linux is using multiple translation layers (meaning it's essentially "porting the game in real-time") at once to run the game.
It's not uncommon for the gains in performance due to Linux being a more lighter/optimised OS to outweigh what is lost in the translation. Not only that, but sometimes the translation is done innovatively enough that it bypasses entire steps Windows has to take, so it's ironically "closer to the metal" than DX12 despite the translation... But that's only in some cases, and in other multiple steps have to be taken for just one in Windows (a lot of work is still waiting to be done, and figured out).

So depending on what libraries, functions, etc a game arbitrarily uses, it may luckily happen to use the translation that have already been heavily optimised (e.g. Cyberpunk on Linux with AMD runs ~20% better for some people), or might use the ones that need much more work (e.g. Forza Horizon 5 runs ~20% on Windows for some people). It's more nuanced than: "game ran better in this case, so that means X".  On top of that, Capcom games are closed sourced, we don't know exactly what may or may not be causing different behaviours.

A fairer comparison would be if VKD3D (i.e. DX12 to Vulkan translation) was used on Windows as well and then compared; still not completely fair but perhaps better.

Still though, with all that in mind, the fact that under Linux a brand-new game that was not designed for it and with an opaque backend, runs effectively equivalent or better, using multiple different community FOSS projects in unison to achieve it, I'd argue is itself a pretty damning indictment for Windows (and for closed source software in general); that it's even approached given that all of the factors are in its favour, let alone beaten, is actually very indicative of all of that bad stuff you listed.

-
Sorry the comment was long, but hey, it's really for other people who may not know this stuff and be confused by what I replied to.

I know of at least two extremely well documented reviews that demonstrated Fedora Workstation outperforming CachyOS.  It's quite possible that the outcome would have even more strongly favored Linux if Fedora Workstation had represented Linux.

For example, see ukGages' YouTube video, "Linux Gaming Showdown: CachyOS vs Fedora" dated Feb 25, 2026.

So even with all the bloat, telemetry and AI surveillance Windows is still faster when test is biased towards CPU @ 1080p.

So... You guys forgot to say what GPU was being benchmarked.

Please edit the article.