1. Steam on Linux via Proton + Wayland (Niri)
2. Steam on Linux via Proton + X11 (Xfce)
3. Steam on Windows
4. Games on Linux launched via other means (it's possible I was missing out on certain flags/optimizations, but this is just about the average experience)
The biggest thing I noticed when switching to Linux was an improvement in framerate consistency, i.e. I'd have fewer situations where the framerate would drop momentarily. Games felt more solid and predictable.
The biggest thing I noticed when switching from X11/Xfce to Wayland/Niri was just an overall increase in framerate. I'd failed this jump many times over the years, so it was notable when I jumped and stayed there earlier this year.
It does feel like games take longer to launch on average, but this makes sense given the fact that it's launching via Proton/Wine.
With those admittedly limited examples though, I don't experience the same ranking in performance, but I attribute that to my non-gaming hardware vs. any problem with Linux or Proton/Wine. I play on a laptop with an Nvidia 3050 laptop GPU, and I get much better performance in Windows still. In Cities Skylines, for example, I'll get ~20 fps on Linux via Proton (but I do experience what you said, it's consistent no major spikes or drops) while on Windows I get between 45-60fps up until about 15k population or so.
Other games, despite working, remain unplayable to me due to performance. I can play Diablo 4 on windows no problem on medium settings, but even on low it's just too unresponsive on Linux.
Anyway, just my anecdotal experience. Those with dedicated gaming rigs will be more than fine with Linux, but those of us on underpowered hardware still seem better off with Windows, unfortunately.
It's better value for money for both the gamers and the devs if the devs just choose to engage with valve and get their game running perfectly under proton.
On the other hand, Linux (or more accurately, the Linux desktop ecosystem) doesn't support a lot of high-end PC gaming features well: HDR, Nvidia GPUs, VR, etc.
In that case it might not be anything the game devs or Steam can do anything about but something you'd have to fiddle with on your system.
That’s interesting and good to know. I’m running an 10th gen i9 with an RTX 3090, so I have plenty of headroom performance wise. I’ve been wondering about Linux gaming on lower end hardware for my younger brother’s sake, and hadn’t assumed it would be worse.
One thing to note: I’ve had all kinds of issues with power management impacting performance. If I let the computer sleep/standby, I’ll get 50% slower framerate until I reboot.
Given the fact that you’re on a laptop, I wonder if power management has contributed to the slowness.
1: the hardest part was finding a bar that supported i3status-rs; not a fan of GTK bars that eat up CPU. I settled on i3bar-river.
I wish more Wayland compositors took this option, seems like a cleaner method of keeping X compatibility and not allowing Xwayland to bring down the entire compositor.
The scrollable aspect just feels so natural and intuitive to me.
If you use ZFS (single nvme) then you can beat windows load times by a fairly large margin. My husband and I have identical hardware for our gaming computers (he uses Windows and I run Linux), it's not uncommon for my computer to load games 10 seconds faster than his.
Was it with any specific game? I just tried the GOG version of The Witcher 3 "Complete Edition" (which is the remastered one) with the Direct3D 12 renderer under both Xorg/Window Maker and Wayland/KDE using umu-run (essentially proton without Steam) and it had identical performance in both cases (i also tried to use Niri but it would launch in 60Hz mode and for some reason wouldn't allow the game to run at a higher framerate with vsync disabled regardless of any option i chose) in either low or high settings (which is basically what i expected since the window system shouldn't be a bottleneck unless something is either broken or you are running at something like 20000fps :-P).
> (i also tried to use Niri but it would launch in 60Hz mode and for some reason wouldn't allow the game to run at a higher framerate with vsync disabled regardless of any option i chose)
I had some issues early on related to refresh rate, and it turned out I didn't have an output defined for the correct display. The steps I took:
1. Run `niri msg outputs` to identify the Display ID and available modes. In my case: "DP-3" and "2560x1440@143.964"
2. Set up an output in niri's confid.kdl as follows:
output "DP-3" {
mode "2560x1440@143.964"
variable-refresh-rate
}Also anecdotal, but I feel like Steam games on Linux compile shaders on the CPU, and maybe not super optimized, compared to Windows where they either ship with precompiled shaders, or it might use the GPU?
Still, the very same games runs better on Wayland+Linux too for me, than on Windows, way less stutters in particular as you mention.
But I'm not sure if it's because of OS differences, or that it's so much easier to end up bloating a Windows install. I can't say I treat them the same, as one is mostly a work environment and the other one purely entertainment and creative usage.
One we can play AAA games I am literally ditching windows forever. Steamos is the best thing that has happened to gaming
We already have the technology now to do it better. A combination of only sending what info a client should have, and server-side checks. As soon as something like UT ships with that built in we can hopefully forget about this horrible hack we currently have to check for cheats.
The goal of anti-cheat isn't to stop the world's most advanced cheaters. Those are already unstoppable because they now use Direct Memory Access over the PCI-E bus, so the cheats don't even run on the same computer anymore. However since those cheaters are few and far in-between they can be handled through player reports.
The goal is to stop the mediocre cheater who simply downloaded a known cheat from a cheating forum. If you don't stop those you'll get such a large wave of cheaters that you can't keep up with banning them quickly enough.
For example: in competitive shooters (where cheaters are most prevalent) you can't have things appearing out of thin air. The client needs to know about things ahead of time to play sounds and to give other environmental hints.
I think that traditional kernel-level anticheat is going away. But the reason is more that when CrowdStrike caused mass outage, Microsoft stated that they want to provide standard interfaces for security sensors, and forbid kernel-level access otherwise (and anticheat can be considered a kind of security sensor too).
If these interfaces become standardized then Valve/Linux could in principle implement them too.
Any anti-malware software ends up ultimately being a cat and mouse game, but that doesn't mean we stop updating our signature updates.
Once you get to match making, global ranks, etc it's just getting too sweaty and ruined by cheating/low trust/etc.
You can see this in existing games with current games with community servers. GTA V's modded FiveM and CS2 Face-IT include more anti-cheats, not less.
I want good balanced matches with players of my similar skill level via matchmaking.
They say they don't support Linux because it's too complicated to be worth the ROI. Really, it's that they don't want to boost a platform where Steam is far and away the default store.
ETA: EAC still supports Linux gaming today, but the rumors remain that Epic could remove that at their whim.
Just avoid it.
¹: Rainbow Six: Siege and Apex Legends, respectively.
The two most popular ACs by far are Easy anti cheat and Battle eye which have natively supported Linux for years, but it is entirely up to the game devs to enable it.
About 40% of all games with AC are working areweanticheatyet.com
Then, there are things like head tracking which are either another dedicated peripheral which may or may not get drivers, or a set of apps which feed from a webcam and output the signal to a standard driver that games know to check for.
Finally, most 3rd party add-ons have custom installers, and I'm guessing most of them won't have a working Linux version. So, while I'm sure it's possible to run, say, a vanilla X-Plane on a non-Windows installation with no peripherals/apps/add-ons, I just see a mountain of work to get a normal, heavily custom installation working.
I know that this isn’t an easy solution/doesn’t go against your argument, because it isn’t download-and-run simple, but discord’s version can be modified with no consequences in a build_info.json file. I used to do it manually, back when they updated it every once-in-a-while, but due to their current tendency to push updates every few days or so, I’ve made a few-line bash script to fetch the latest version (thank you httptap) and patch the file for me. For screen sharing, I use whatever current discord client on GitHub supports it for Wayland, which usually has the added benefit of not limiting quality and framerate options.
But yes, you do have a point, it’s not just ‘as simple’ as it is under Windows - when Windows works properly.
This missing piece is sort of a fun "whatever happened to VAC and why hasn't it kept up with the times?"
It seems like Linux would be a good excuse to reinvest in VAC and make it a bigger competitor to the current favorites like Easy Anti-Cheat (EAC).
Not a big issue if you're just using kb/mouse/controller but you can get into the weeds with VR, flight sticks, wheels, etc.
Does Valve run a SteamOS CI/CD farm? I could see a Rust based template and library for calling into this set of APIs that you could upload your well structured project and it would build and test for all platforms. Rust would just be the skeleton, your game logic could be in anything Rust could link to.
Test your game to make sure it works on the Steam Deck and avoid features that don't work on Proton, but you still have to primarily target Windows.
I never came up with a good explanation for that.
Disk cache, to be precise.
Not out of box - games require mild tweaking but nothing wildly challenging. Add parameter to launch command line etc. The proton database & comments on there usually explain what tweaks the game needs
Don't think I'll switch back
But even then, assuming that is true, if they're pretty much the same would people care about maybe some fog looks a little different but you get an extra 15-20fps in a game? I think a lot of people would still prefer the boost in frames.
Proton supplies a DLL that implements the Win32 API using Linux syscalls. Windows supplies a DLL that implements that Win32 API using Windows syscalls that you're not really supposed to use directly.
so 'translation layer' is not that unfair.
For so long it was one of those "and now you have two problems" technologies and now it looks like it's the slow blade that could actually kill Windows.
I'm a Mac guy now mainly because of my job and I like UNIX-y stuff now, but of course, gaming is even more lacking than Linux.
We're so close. Once AAA releases and GPU drivers get there, it's over the cliff, and I could see that being in the next five years.
Check whatever you want to play at https://www.protondb.com/ - chances are, if it isn't intentionally borked with anticheat, it runs just fine. Looking at the top 300 games by Steam player count, 17 don't work, and probably 5 of those are utilities (like Crosshair X and Lossless Scaling).
E.g. the difference between the Lenovo and Asus Win11 drivers is sometimes bigger than the difference of the faster Windows driver to Linux.
It's also not all that surprising though, there's a lot of very smart people working on Proton while the general quality level in the Windows ecosystem is slowly but steadily declining.
I also wouldn't be all that surprised if running a D3D11 or D3D12 game on a Proton-layer on Windows would be faster than running that same game without Proton. Sometimes Proton might have workarounds for 'API abuse' problems of specific games which the native D3D implementation or driver doesn't have.
Windows 8 tried to divorce more of the compatibility layers (start them up only as needed) of "The Old Desktop" to a lot of flak from the development community (and some user confusion), but if you were paying attention and used almost exclusively Windows 8 "Store" apps at the time you could get some serious memory usage wins.
Windows 8.1 walked so much of that back and made the Desktop/Explorer and all of its compatibility layers boot first again, but there was a small period where Windows 8 shined.
Presumably this sort of stuff is what the new Windows Xbox efforts are doing again, but the "boots to a full screen experience without 'a Desktop'" expectations of games and game-focused hardware makes it easier to boot the Desktop only if needed in a way that makes sense to game players that didn't make sense to general Windows users with a long tail of ancient applications that they didn't or couldn't "just" upgrade to new ones.
As for the performance, its a 15W handheld trying to play games that 600W PCs and 300W consoles struggled with just a few years ago.
Anybody know if Steam and games in general refuse to install in Windows LTSC? Its basically the stripped down ultimate lean version of windows. Boots insanely fast - no tracking bullshit - no windows store or candy crush. Battery life hugely improved. No big updates - security only - and for a longer supported time.
I know Adobe has forced their installers now to refuse to outright install on LTSC (for no real reason) which is annoying as hell. First they stopped it installing on Windows Server.....
Hopefully we do not see the same thing with graphics drivers and Steam and games because right now its the ultimate gaming OS (especially if you are running it as a second OS while daily driving Linux or MacOS)
1. There is no standard debloated windows 11 to compare against since Microsoft adds more bloat each month.
2. Users aren't going to be running a debloated windows 11 anyway
> We then installed Windows 11 on the handheld, downloaded updated drivers from Lenovo's support site, and re-ran the benchmarks on the same games downloaded through Steam for Windows.
The foreground application gets a priority boost, it's threads are going to take a larger slice than any background thread, and with the number of cores available to the OS, they will likely run just fine concurrently.
This as some evidence of bloat equaling poor performance isn't justified. It's Ars' using 6 month old display drivers.
I'm not claiming newer drivers will solve the performance problems entirely or at all, but for the purposes of accurate testing, they should have updated them rather than stick with what the ODM is providing, which are always out of date.
But then I remember that it's Nutella at the helm over there and he'll gladly give up ground to focus more on hype and share price.
What a waste.
Besides, MSFT is almost $500, so they're doing something right.
Care to expand on this? Or did you just want to make a random statement?
> Besides, MSFT is almost $500, so they're doing something right
This very reductive way of thinking is exactly why everything is getting enshitified to the max.
While I agree with the basic sentiment, racism really doesn't deserve a place here, please don't do this.
Grow up and stop finding racism in everything.
Today, though, Ars testing on the Lenovo Legion Go S finds recent games generally run at higher frame rates on SteamOS 3.7 than on Windows 11
That's not just a buried lede, this title is straight up wrong (or at least, not backed up by data)With SteamOS coming to arbitrary hardware, that is a very bold claim to make. And not one that ars has data to back up, apparently.
It's also an embarrassment of an article because they were gifted the steam version of the handheld, then compared that performance against them installing windows... on the steam version of the handheld. Why not buy the version with Windows by default?
Personally I'm nearly certain that SteamOS would give better apples to apples performance than windows, but we shouldn't give an article that shits on both the scientific method and journalistic integrity the light of day
Correction: it’s been a month and they had both devices:
https://youtube.com/watch?v=CJXp3UYj50Q
And it’s $130 cheaper.
Borderlands is not cel-shaded. People keep making that mistake for quite a while though.
My observation is that windows is slow at everything. I think because of Defender, but I'm not certain. If you set up a Linux VM on a Windows machine, most tasks run much faster than an identical task on the host OS. It's insane.
I run my C++ compiler in a linux VM because running it on windows is, no exaggeration, twice as slow.
Microsoft really needs to release a gaming version of Windows without the bloat. I only use Windows to launch Steam these days.
They should focus on optimizing the current OS to more optimally run games rather than segmenting the OS.
I made about 100-200 save files in a game, so opening a screen with list of saves took about 10 seconds.
But when I used the same save folder in the same game installed under wine in Linux, its loading screen took half the time. Even though NTFS is not native for Linux. I have no idea why. Windows was without antivirus software.
If folks can figure out how to run Gamepass on Linux before then, I'll bounce, but I understand it's pretty tightly coupled to the Windows OS.
Maybe 60% of games work and it's such a headache trying to get it working, if it can be fixed at all.
Modern games however tend to work really well.
I've recently found about Dreamm[1] by AAron Giles (a well known emulator developer) which is basically a very lightweight os-indipendent reimplementation of some windows and directx calls specifically for some Lucasarts games written during that time period, It would be nice to see a similar project expanding in such direction without having to reinvent Wine and/or Proton.
There's also DOSBox (which is quite capable at running win9x now, with Voodoo emulation) and 86box to fill those compatibility gaps too.
I mean, at least until last week, when I bought myself a new top-of-the-line laptop. I’ve been distro-hopping trying to find something that works and everything failed in its own annoying way. Part of it is because I stubbornly decided to stick to Wayland because I really wanted to use my laptop’s HDR display to the fullest.
Nobara KDE had serious issues handling hybrid GPU mode. The SDR color profile of my built-in display got completely borked - worked fine in HDR or plugged in to a display. But then I had serious graphical artifacts when I plugged in my display with VRR disabled! They went away when I enabled VRR, but the flickering was really bad. All of this went away if I switched my laptop to dGPU mode, but grub stopped showing anything and I couldn’t reach the UEFI anymore unless I removed the SSD.
Next I tried Garuda Dragonized Gaming. The styling is atrocious IMO, but I really liked the OS management tools. Unfortunately I couldn’t get it to recognize the dGPU, so I moved on.
Next I tried Bazzite. I was very impressed by how well everything worked and performed! Atomic Linux made some of my regular setup more complicated, but the challenge was interesting. But then I decided to unplug it from my dock, and I discovered that the kernel was rebooting the built-in keyboard constantly, making it impossible to type anything.
I decided to go back to my go-to safe choice, Pop!_OS. Installation went smoothly as usual, I even followed a tutorial to use Btrfs which I really like. Everything worked great until I plugged in my monitor and the whole system started stuttering.
I decided to give up for now, I installed Windows again and applied Atlas OS to it to trim down the annoying stuff. After some tweaking I got the battery life to something that seems reasonable. Games work as expected, and I’m mostly done finding alternatives to some of my personal setup quirks.
I want to be clear: my switch to Windows is temporary until fixes for the issues I experienced start to surface. My laptop model is very recent, and I don’t have the know how or time to dig deeply into all of these issues. I’ll probably be sick of Windows in 6 months, ready for round 2.
I am 100% team Linux but cross-platform benchmarking is rife with problems. Differences are often from not testing the exactly same thing or different defaults that trade-off safety/quality/perf and can in theory be changed. No point measuring only one side of a triangle whose ratios are a matter of taste.
A responsible benchmark would try and prove fidelity.
(this is Hacker News after all, seems like a reasonable question to me too)
I had not thought about Motion...good one.
Agreed regarding a responsible benchmark.
All that exploration could help illuminate (ha) the path to a user-first "Desired frames per second" Graphical Setting in game set-ups. "I just want 60 / 120 / don't-care 30 is okay". With nVidia going bonkers with new AI interpolation features the users may want a "minimum 60 please".
It's a soft feature that Consoles have, IMO, though with variable rate refresh panels I feel it's less of a draw.
Does Linux have workable VRR support?
I don't believe ReFS is contributing so much to the performance improvements seen when using Dev Drives. Removing storage filters from volumes can go a long way.
From my personal experience overall games run much much better on Windows ( 10 or 11 ).
Edit: ok I just noticed the title is missleading, it's for handled device not pc.
More news at 11.
That's absolutely noteworthy.
those games come with benchmark tools
Now, maybe you cut out Bongo Cat and Wallpaper Engine, those aren't games. Next on the list we'd have Apex Legends (2020) and Warframe (2013).
Few of these games tell you much about performance. And I don't think many of them have ways to get a consistent, quick performance measurement. I think the article authors made a pretty good set of choices, tbh.
Code and kernels that target known hardware doesn’t need dynamic conditional code to handle unpredictable hardware. This will be faster.
General purpose operating systems handle printing events, background updates, periodic online checks, network discovery, maintenance jobs etc, all these operations consume resources and time.
Yes, Steam deck on Linux will run faster than equivalent games on Windows. But Steam deck on a smaller OS like Haiku will run even faster than Linux.
Engineering is a compromise. A F1 car can corner faster than a passanger car. But it probably sucks to reverse park. Also, I cannot imagine using a sports car for grocery shopping and hauling furniture from Ikea.