In my purely anecdotal experience over the last few years, performance ranking is as follows:
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.
Interestingly enough, I've had games that had both a native Linux port and Windows version, and the Windows version through Proton ran better than the native Linux version. This ended up being true for Civ5, Civ6 and Cities Skylines (1).
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.
Linux port if there is one is usually done by a third party porting studio, which is not necessarily at the same quality as the original codebase. Also the devs just don't have the manpower/bandwidth to spare for Linux users given how small this community is.
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.
These ports are also not usually source ports, so they're not much "more native" than the Proton ports. They often use the same kind of API translation layer, probably also built on WINE. I think as Proton sees more investment and becomes more advanced, it's probably becoming difficult for competing compatibility layers to keep up.
A source port that is optimized as lovingly as its Windows counterpart will probably be faster than the Windows version running via Proton, but the incentives aren't generally there to justify the costs/difficulties. Maybe some day it will be! That would be wonderful.
But until then, Proton seems like an increasingly compelling option for these compatibility layer-based ports of Windows games.
Factorio and Minecraft (Java edition) are two of the few games that come to mind where the Linux port got comparable effort to the Windows port, and I don't think people are in a rush to play either of them in Proton.
Your latest AAA open world RPG on the other hand? Yeah, you're probably going to have better luck in Proton even if it gets a native Linux port.
> and get their game running perfectly under proton
Even better would be to compile for linux, but use DXVK-Native (https://github.com/doitsujin/dxvk#dxvk-native) if you think migrating from DirectX to Vulkan requires too much effort.
They probably QA mostly for windows, so they run into bottlenecks and edge-cases of windows APIs during QA. Linux-native APIs probably have different bottlenecks and edge-cases.
I think the reimplementations of Windows APIs in Linux, even though alternative to the original, should have similar bottlenecks and edge-cases. So the extra QA on Windows helps the Proton version more.
The answer is pretty simple here - hire CodeWeavers to work on supporting your game in Proton/Wine rather than some other porting shop doing an old rewrite-style port.
To be you should compare the windows version on windows, no proton against the Linux version.
DXVK, which proton uses, makes some games run better in windows than "native".
> 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.
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.
> 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.
> HDR
Already supported
> Nvidia GPUs
You have it the wrong way around. NVIDIA had issues supporting Linux, not Linux supporting NVIDIA. AMD drivers work fine, so its not a linux specific issue.
Is it though? I confess I haven’t tried in a few weeks but until last time I did, to get HDR in games you had to start a session with `gamescope` rather than a DE, and still had to set a bunch of flags - and in some ways have a very subpar experience with problems with mouse movements and other issues I can’t recall.
I exclusively game on Linux and I find the experience far superior than doing anything on the other OS, but last I checked HDR was not actually supported.
KDE has supported it for a while on the desktop. Still needed Gamescope to make it work in games. Like I mentioned, I haven't tried it in some time so it may have changed.
For gaming and general desktop on Linux AMD is best if you want a dedicated GPU.
If you want a laptop with good battery life Intel is generally the way to go.
A lot of this is due to the enormous amount of effort Valve put into improving the open source AMD drivers, which is what is used on their Steam platform.
Of course if you want CUDA you need Nvidia, but if you use Nvidia to drive your Linux desktop expect some suffering to go along with it.
For what it’s worth I’ve been using an RTX 3090 and it’s been mostly smooth sailing for a couple years now.
Running NixOS with a pretty vanilla configuration and it has been hassle free.
I did have to disable power management at the system level because framerate suffers severely if the system sleeps and wakes back up, but I shut the system down when I’m not using it, so this was a non factor for me.
It's getting there though. I own a high-end PC with nvidia GPU and I play VR on my Linux setup via ALVR (I own Quest 3)
It's not straightforward and full of workarounds I have to do, but once you're in the game it works great
I think Windows isn’t that different, just that there's more motivation for NVIDIA or Microsoft to fix those things. I recall not that long ago a combination of Windows 11, my NVIDIA RTX 40xx, my previous Dell Alienware monitor also had some issues with switching between SDR and HDR (and later Dolby Vision brought even more of a mess).
Meanwhile Android and iOS phones have been able to do it flawlessly for a while now…
There are indeed nvidia drivers for Linux and they're reasonably good for gaming, but the feature set sometimes lags far behind windows. There is no DLSS 3 for Linux, for instance. (as of a few months ago anyway - I haven't checked recently)
Nvidia support across the desktop ecosystem is poor, for example practically nonfunctional in Sway. And just buggy in other Wayland based desktop environments (kde seems to be the best in my experience).
WRT Nvidia+Sway this was certainly true not so long ago. But since the latest Ubuntu release and with a recent driver I am running this combo and it works flawlessly.
Really! I'm happy to hear it, I'll give it a try on my desktop this weekend. I would love to get all my machines on the same desktop environment once and for all.
I'm doing PCI passthrough of a 1080 to an archaic tiling X11 window manager and so far it Just Works with noeveau. XFCE also worked fine before I decided I don't want a full DE. Rock stable. I will move dists before I move to Wayland by the looks of it.
i3 should be pretty easy switch from sway if you haven't tried.
It doesn't have anything to do with graphics in the Linux ecosystem. Netflix specifically blocks Linux from having decent quality so it's kind of pointless to discuss. If you want high quality, you can pirate it or rip from disc. Dirt cheap n100 minipcs are capable of playing UHD bluray rips in Linux just fine for example, so Netflix's relatively low bitrate media aren't an issue.
Not an ecosystem issue at all. There's no "yet". Linux computers even on the very low end are already perfectly capable of playing 4k Netflix videos. You can easily prove this to yourself by downloading one via torrent (you can generally get exact stream rips with DRM removed if you want). Netflix just won't send UHD streams to Linux users. That's a political choice, not a technology problem, and it's easy enough to get the media elsewhere DRM-free if it's that important to you since Netflix evidently specifically does not want Linux users as customers.
I wonder of this might be due to your Linux nvidia driver (nouveau?) pinning the card on baseclock by default while the Windows one will allow it to scale up? Something I heard somewhere that seems applicable here.
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.
Happens with whisky and macos in my experience. It is like as soon as a game is installed and you do something like chuck a grenade, no explosion the first time ever you do that.
> 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.
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.
I have a laptop with the same GPU, and Diablo 4 runs really well out of Lutris. Graphics version 570, and the CPU is an AMD with a Radeon 680M integrated. I often play games with FSR on, which probably keeps performance higher?
Side note, Niri is a fantastic WM. When I saw the Phoronix article on HN talking about the addition of overview mode and more, I finally took the plunge and spent an afternoon converting over from Sway.¹ Anecdotally, I've seen less hangups on Niri around fullscreen games and floating windows, perhaps thanks to X11 running in xwayland-satellite.
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.
I've been so happy with Niri after many many years bouncing around other WMs. It addresses the main issues I've had with other tiling window managers and has been such a joy to use.
The scrollable aspect just feels so natural and intuitive to me.
Steam won't launch for me in xwayland-satellite here... I just assumed steam+wayland = broken. I have a kind of weird setup using sway with xwayland disabled and running xwayland-satellite though.
Try running xwayland-satellite in another WM like Niri and see if it works there – for example, Gamescope didn’t work well on Sway, crashing as soon as Steam tried to spawn another window, but it’s working fine in Niri.
Been a Linux gamer for years now and I think you are correct on your frame rate observations in general.
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.
Yep, compression is on and I think that's what does it considering how large games are these days.
I stay away from XFS, every time I have used it in the past my entire drive have ended up scrambled within a few months. It's by far the worst file system I have ever used, not even FAT32 was that unstable for me.
Most large games are already compressed in cabinet files or similar so OS/file system level compression wouldn't gain you anything. While there are games that ship with loose files, like the Elder Scrolls series, other games like Diablo IV or CoD use archives.
> The biggest thing I noticed when switching from X11/Xfce to Wayland/Niri was just an overall increase in framerate.
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).
Some of the games I play often that saw improvements: The Finals, Overwatch, Rocket League, Helldivers.
> (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:
Hm, i tried it and i could set up the higher refresh rate but Niri was forcing vsync on the game that capped the framerate to my monitor's refresh rate (165Hz) even though in the game i have vsync disabled. This seems to be an issue with Niri as KDE Plasma Wayland disabling vsync works fine.
At "ultra" settings i got around 115fps which is ~5 fps lower than 117-120fps i got from Xorg/Window Maker and KDE/Wayland, though i'm pretty sure that was just the forced vsync, so in practice it seems that the window system doesn't matter much.
Did you use XFCE's desktop compositor? AFAIK XFCE's compositor isn't particularly great, some years ago when i was working on a custom game engine i had to add an explicit option to use the X11 "override redirect" flag instead of the window hint for fullscreen windows because XFCE's compositor wouldn't disable itself otherwise and the game would feel a bit laggy/inconsistent. Not sure if this has been fixed nowadays but in general it gave me a bad impression for XFCE's compositor as other compositors didn't seem to have the same issue.
Is there a way of making wayland actually usable with Nvidia GPUs? I never manage to make it work, and it makes the whole system feel slow and sluggish compared to X11
> 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.
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.
The last missing piece for full Linux gaming is anticheat. Last I looked into it, the major vendors don’t want to support it due to lack of kernel security and the ones that do, game devs refuse to allow it (destiny for example)
One we can play AAA games I am literally ditching windows forever. Steamos is the best thing that has happened to gaming
Anti-cheat today is a stop-gap measure at best. For various reasons such as improved OS security and security concerns with this software, ring zero anti-cheat won't be around forever. Besides, it's a cat and mouse game where the vendor is the mouse.
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.
> Besides, it's a cat and mouse game where the vendor is the mouse.
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.
With the emergence of AI cheating, cheats don't even need access to memory anymore. The cheat can entirely run on mouse and screen peripherals and the computer will have totally no idea what's going on. The best you can do is behavior analysis. But it always comes with chance of misreports.
Although that looks like an obvious solution on first glance it's not really technically feasible. Things like gunshots or footstep sounds are not visible to the player, but still need to be relayed to the client.
As far as I see the only way around not sharing anything that's outside of the immediate perception of a player is to have the audio and graphics be entirely rendered server-side.
You can minimize that to some degree (Valorant does this), but due to movement prediction/network latency you do have to overshare a little bit.
I imagine that most game devs just look at the incredible amount of work this takes to implement and complexity it adds, and decide to not bother. Valorant can do it because the game itself is low complexity, the developer has deep pockets, and also the added competitive integrity is valuable.
Game servers are complicated and have a lot to manage.
It's infeasible for the server to keep track of each player and do frustum and raycasting to every other player to check who can see who every frame.
Culling out of view entities also has the problematic effect of when a player spins around you now have to stream in several big chunks of world state in the few milliseconds before the user clicks to get that 180 no-scope.
> 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.
Working on mostly server platforms, I had forgotten that IOMMU enablement (and, where relevant, enforcement) was not the default.
In my world, we won't let a system boot with production credentials unless the IOMMU is enabled.
This is enforced by a greatly enriched TPM (and it's willingness to unwrap credentials). We have trust several layers of firmware and OS software, but the same mechanism allows us to ensure that known-bad versions of those aren't part of the stack that booted.
If I wanted secure games (and the market would tolerate it), I'd push for enforcement of something similar in the consumer space.
It's your IOMMU, you can do what you want with it. Maybe you need to write heaps of stuff to take advantage of it, but what's new there?
The only thing you're getting by saying "no IOMMU" is "I want any devices in my machine to be able to do anything, not just what I want them restricted to".
Okay, but he's specifically brought it up in the context of a computer's owner doing something that the software vendor (and also myself as another gamer harmed by cheating) would prefer he did not.
> unless you find it terrifying that owners of hardware have control over their hardware
I mean that the presence or absence of an IOMMU doesn't impact whether owners of hardware have control over their hardware.
It just means that the owner of the machine is able to limit what memory the devices in their system are able to access, in the same way that MMUs limited what memory every process on your system could access.
As long as games are running on user hardware/OS, you'll always deal with cheating. Server-side checks and computation can only go so far.
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.
Exactly, nothing short of streaming the entire game fully rendered from the server will stop cheats. And even then you can probably still do aimbotting with modern day computer vision.
This reminds me of a discussion around 2 decades ago, where someone showed a picture of his "undetectable aimbot" for a turn-based artillery game: a ruler, a page of charts, and a handheld calculator; followed by a copious amount of discussion of whether that was considered cheating.
How exactly will it stop cheats? Any skill based game can still be cheated. Just analyze the video stream, or go even lower tech, point a camera at your screen. Many games can be effectively cheated like this. For eaxmple Aimbots in counter strike and peak human reflexes in dota/lol.
Surely the server can tell the client what sounds to play and what other environmental hints to do, just as well as the server can choose to tell the client where the other players are when they are in sight.
The storage read, memory bandwidth, load computation, and gamethred pause to add the object to the game world is far more expensive than sending a move.
I've always thought the line about whats cheating, and what's not is unfair and arbitrary. How is it ok that some players can play 4k 200fps and others 1080p at 30fps.
The only way to be really fair is for everybody to Stream the game at the same res, frame rate and latency.
In certain competitive environments framerate is definitely limited. Here [0] are the rules for Fallout 4 any% speed runs, framerate must be capped at 60FPS. AFAIK that rule applies to all games in this engine due to physics behavior. I don't follow tournament FPS games, but it wouldn't shock me if there are rules for competitive play there as well.
If you are asking why games like counterstrike don't have limits on online play, that's mostly a commercial question. Would those games be as popular if they limited performance to what was achievable for minimum specs? I certainly wouldn't want to play at 1920x1080 on my nice widescreen monitor, but setting the minimum to a $1500 monitor and the hardware to drive it would guarantee very few players.
Yeah and in real world people from different countries with vastly different economic backgrounds compete on the same stage, I think video games are fine.
This isn't exclusive to video games. Much of the improvements to world records in sports are due to improvements in gear, yet we don't consider those records to have been unfairly achieved.
Some games do impose limits though, for example Overwatch doesn't allow you to use an aspect ratio larger than 16:9 and selecting a wider aspect ratio actually cuts down on your vertical field-of-view rather than granting you more horizontal field-of-view. This lessens the potential advantage of ultra-wide monitors.
How would consoles be any more immune to computer vision based cheating? Instead of feeding the output to a spoofed keyboard & mouse, you'd just be feeding it to a controller input. I'm not really seeing any difference in technical challenge here, and you wouldn't even need esoteric hardware since console controllers are USB devices anyways.
Since the hardware is better controlled and secured, and hardware attestation is a solved problem these days, it's not particularly difficult to enforce security to the point where you'd need to hardware hack a controller and connect it to a physical camera to bot.
That's still gonna be annoying for players, but it'll greatly decrease incidence, and if reporting a player for botting requires buying and hacking a new controller... It should be quite effective.
They are often more convenient and secure. If you don't mind a single-purpose device that severely limits your ability to modify your experience. Better is subjective after all.
> Anti-cheat today is a stop-gap measure at best. For various reasons such as improved OS security and security concerns with this software, ring zero anti-cheat won't be around forever.
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.
The goal isn’t to stop 100% of cheats but the majority of them and that’s fine. Either way, it’s the only thing stopping me from playing the rest of my games on steamos.
Multiplayer games without dedicated servers is dead end anyways. I dont need a "anti-cheat" daemon hooking into kernel scanning files and other memory while playing a game. Communities in dedicated servers are much more efficient at moderating the player base than centralized match making ever will be.
This is where I'm at with gaming. Even outside of cheating, it's not fun to me to be dumped in a game with screaming children/manchildren. If I'm playing a game I want it to be with my actual friends. And then I don't have to worry about them running cheats because I trust them.
Once you get to match making, global ranks, etc it's just getting too sweaty and ruined by cheating/low trust/etc.
At least in TF2 the sweats are all on the community servers. I can get like 3 kills a round on skial servers. I’m trash there. When I use the valve matchmaking servers I dominate the lobby about half the time.
Communities with dedicated servers include anti-cheat though. Most people aren't interested in spending time moderating a player base: they'd rather just play the game. So server admins use anti-cheat.
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.
Yeah, but it's very time consuming/impossible to find similarly skilled players for a fun lobby. The only competitive game I care to play on Linux is Rocket League, which is nearly impossible to cheat at, so it doesn't currently have anti-cheat, but I wouldn't be surprised if Epic decides to put their beloved EAC in it at some point anyway, maybe even just because they hate Linux so much.
They really aren’t. I used to think that but I actually enjoy skill based matchmaking. It makes game availability better and faster and I don’t have to deal with 1 outlier absolutely stomping or overzealous admins or whatever. I like both approaches though for something like battlefield I think dedicated servers are better but for things like cod, siege, etc we need sbmm
It's also an anticompetitive red herring, at least for Epic.
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.
This is problem for me and my brother right now. He's up at a remote job site and we want to play Siege or Apex together¹, but both require anti-cheat and don't support Linux. And I'm loathe to devote space on my SSD to Windows.
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
There are rumours of next xbox generation supporting steam platform and 386 architecture. I know it's a bit off topic, but it could be an elegant solution to the cheating problem, gradually move to standardised consoles. This could solve the dma problem too
AFAIK most anti cheats such as BattlEye actually work under Proton, but the Game Developers have to write to the Anticheat vendor that they want to opt-in into allowing Proton.
Don't forget the ancillary applications that gamers want. If you follow Discord's website, you're gonna end up installing a DEB file manually. Then, every couple weeks, Discord won't launch until you go download another DEB file and install that. Oh, and good luck getting Discord screen sharing working on Wayland. I tried for hours, gave up, and switched to X11. So, just in Discord, we've already run into two hideous workflows that no Windows native is going to take in stride.
And certain types of games have a _ton_ of ancillary applications. For flight simulation, I rely on 2-3 additional contollers, some of which I am fairly certain either won't have driver support, or at the very least will have some major issues with the GUI and configuration.
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.
Discord is shipped in a number of package managers (I don’t know the status for mainline apt repos).
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.
Think about it from the POV of a Windows user, especially one who has never used Linux before, and especially one who doesn't know what HN is. To install a program, the first thing you're going to do is type "discord" into your browser, and go to their website. Discord's website doesn't suggest that there's a better option. It just gives you a DEB file.
It's funny because one of the OG draws to Steam back in the day was because Counter-Strike (et al) had the superior Valve Anti-Cheat (VAC) in Steam copies than the myriad of raw CD installs of HL1 plus the mods to run Counter-Strike.
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).
interestingly, I have no issues with the anti-cheat within Marvel Rivals; however, games that embedded an anti-cheat prior to the steam deck popularity don’t work as you described (PUBG, apex legends).
Given that Windows games run faster via Proton on SteamOS, developers should prioritize targeting SteamOS APIs—not Windows. This ensures compatibility with Windows while maximizing performance. Game engines like Unity and Unreal must adopt SteamOS as the primary target, with CI systems rigorously testing both platforms. SteamOS, not Windows, should be the baseline for optimization.
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.
I'm not sure that makes sense since the Windows API is the source of truth for how something works. If you make a game that works on Windows but not in Proton, Valve will push a fix that makes Proton work the same as Windows. But if you make your game work with Proton, but not Windows, you are relying on some quirk of Proton which isn't guaranteed to work in to the future and as soon as something else needs it to work the same as Windows, your game will break.
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.
You would need to test on both of course. I am arguing that one should target the fast happy-path on Proton as Proton is a subset of the Windows APIs that runs faster than Windows.
While they are wrong, in that Proton is absolutely a subset of Windows APIs, it is describing aiming for an implementation detail.
Yes call X is faster than call Y in Proton, but that could also be because its only 50% written and skips a bunch of side effects that will be created after you have written your game.
Therefore you need to view Proton as a potential moving target. Not that Windows isn't too, but its just not as clear cut as is being claimed.
I know this because I pin some games against older versions of Proton because they work better/faster.
Thank you. I didn't do a very good job of explaining why I think Proton is worth targeting specifically if perf tests show it running faster than Windows.
The Windows API perf topology is astronomically large.
That Proton is faster than Windows can't be universally true, but if you stick to that subset that has good conformance with Windows APIs and is also faster, that should be the target (and everything is moving of course).
Cross benchmarking games between Windows and Proton, taking API traces and finding those subsets of APIs can usage styles will enable a game developer to target those Proton APIs that are both conformant and fast. It is that subset of Windows APIs (temporal spatial structural usage) that I am suggesting being the target. The API linting tool might itself compare API traces. In fact for those API traces from performant AAA games, I could see generating a header file that only exposes the same API as used by that game.
Valve actually ships "stable ABI" Linux runtimes that Linux games and Proton are built against, so that they don't have to rely on distribution packages, similarly to Flatpak. They renew them every few years but they stay stable so that games built against a specific runtime keep working.
Ignoring that, you know, 99% of users are running Windows and not SteamOS. Test on what your users run. Proton is just an implementation of Win32, you're still just targeting Windows.
Over what time period? 99 -> 95.45 is a huge absolute number. Is most of that attributable to SteamDeck? Would that mean that 4.5% of active Steam users own a SteamDeck?
If you click on Linux only and expand the details, SteamOS accounts for 31%. "Other" is the next biggest at 23%, and Arch (which SteamOS is built on) at 10%. Someone probably has made a comprehensive historical spreadsheet but the internet archives save the page frequently if you want to see trends. Even historically though Mac has commonly been known for being around 2%, at times more, it's only recently (like in the last couple years) that it fell below 2% and Linux passed it. SteamDeck did have a big impact for sure, though. I've seen estimates for around 4 million sold so far after three years, that's not bad. Putting some extra effort into your game's release to get it "SteamDeck Certified" seems worth it to me just to get that little bit of extra attention from those customers.
Back in the Windows XP days, I discovered that running Windows in a VMWare VM, hosted on Linux, was faster than running the same version of Windows bare on the same machine.
I am guessing that they're implying that a VM effectively loaded the "hard drive" that Windows uses into a RAM-disk, so one would be comparing loading dlls and whatnot from RAM vs loading it from spinning rust.
I'm not sure how true that is, because in the Windows XP days most of us wouldn't have had enough RAM to spare to do that.
Depending on the VM technology they use they offer a variety of different caching mode and configurations, but the basic three approaches that most everybody offers are going to be something like;
"writeback" means that when the guest's storage data ends up in the host's cache it is reported back to the guest as 'written'. This means that from the guest's perspective the disk is written to, but in actuality the data is still floating around in memory. If the guest wants data to be 'safe' they need to issue additional flush commands.
"writethrough" means that the host is using its memory for caching file system, but that writes are reported as 'competed' only when they have been committed to actual disk.
and "none" means the cache is used as little as possible.
So if your guest's virtual disk is in 'writeback' mode it isn't actually writing to real disk. It is writing to memory. Which is going to be very fast up to the point were the cache on the host is exhausted.
Certainly Windows could lie to applications and not write information to disk and keep it in memory much longer then it normally would but that would defeat some of the assurances that file systems are supposed to offer to applications.
'writeback' would be closer to what Windows already implements on the OS level, but because Linux is just plain faster it should improve performance somewhat. Microsoft can only work in improving the performance of Windows to get the same results, but Linux is pretty hard to beat.
'none' is what I use when running Linux on Linux because having two layers of cache is just kinda wasteful and doesn't result in real improved performance.
So if I understand correctly, you're suggesting that in essence the VM lied to Windows to improve performance? It's been many years, but I don't imagine I would have chosen an option like that. I am very surprised VMWare would default to such behavior.
> So if I understand correctly, you're suggesting that in essence the VM lied to Windows to improve performance?
Pretty much. Just speculating because I don't know how your systems was configured back in the day.
But all of what I said applies to most VM solutions.
If you are dealing with enterprise-grade hardware it isn't a bad things. Keep in mind that typically OSes are going to assume that they are the only things operating on storage. So if you have like 30 windows boxes all trying to write to shared disk at the same time it can lead to some bad behavior if you are not using a write cache.
The system has battery backup and typically they expect you to use shared SAN or NAS with multipath and/or bonded network interfaces for redundancy as well as having backups. So the chances of data loss is a lot less then typical consumer hardware and it gives the hosting OS better chances at optimizing and scheduling writes properly.
Also remember the guest OS still has the option to send a 'flush' command to disk, which would ensure the writes complete regardless.
> It's been many years, but I don't imagine I would have chosen an option like that. I am very surprised VMWare would default to such behavior.
This is pretty normal. You'll see the same sort of options with hardware RAID devices and such things with their own internal battery-backed cache.
Recently switched as well (Arch not steamOS, which is arch based) and it's been pretty solid.
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
Some of these games are practically neck and neck for performance. I'm wondering if it's a similar situation to early Proton comparisons, where framerates were higher in Proton but when comparing still for still you could tell it just wasn't actually doing certain effects. Are there features that are being attempted in the Windows version that are just not functional and thus effectively disabled on the Proton one?
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.
Might be unfair to call Proton a "translation layer" because the Win32 API is not defined in terms of system calls but rather a set of functions exported from a DLL.
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.
Clearly they don't want to come out and say that Wine is an "implementation of windows apis", because that would invite legal issues. But clearly this is what Wine really is for a large part. Some stuff are just shallow shims to Linux apis, while other stuff they need to make more of their own implementation of.
"WINE Is Not an Emulator" has always said something exactly like that. It hasn't exactly been a "secret" that Wine is an implementation of Windows APIs (and not an emulator/translator).
I think that's how it started out, and also how a lot of developers still conceptualize it. Wine has had to massively expand that scope to reach the maturity it has now. I think it's kind of straddling the line between "Implementation" and "translation".
Philosophically its still a translation layer though. It doesn't really care about correctness if the no apps depend on it. Success is in meaningfully running client software. The implementation of the Windows Libraries are just a way to get there.
I really have to respect Wine's persistence (with Proton's help).
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.
Wine is translating Windows ABIs (not APIs) into underlying Linux OS and userland. Translation simply means that normally Windows ABIs are meant to be used on Windows, they aren't native on Linux.
Also, most games are still 32-bit EXEs, so on most PCs today they are running not just in the Win32 subsystem, but the WOW64 subsystem (Windows-on-Windows 64) the compatibility layer of running 32-bit Windows applications on top of 64-bit Windows and its Win32 subsystem (it's still called Win32 on 64-bit, an unfortunate naming bug from massive compatibility break differences between Win16 and Win32 that didn't exist in the 64-bit transition, 64-bit didn't get a new API subsystem, it just upgraded the existing one, mostly).
AAA releases have worked for years with the exception being about 60% of online games with anti cheat. Linux has been a first class gaming citizen for years if you use steam client and AMD gpus.
We've been there since the release of the Steam deck. Really, the only things I can't run on my Linux machine are the games that intentionally break the game with anticheat mechanisms. And it's not all anticheat - there are plenty of titles with anticheat that run just fine.
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).
I have always wanted Windows to move to Unix so that I can have the best of both worlds(software development and video games). Glad that we are close to that reality.
> Because the NT OS/2 I/O system is asynchronous by nature, the ability to make a
request and then have it completed at a later time makes it natural for implementing
oplocks. Further, because synchronization is required by the file system to determine
when the caller has completed its oplock update transfers, the file system can use this
feature to block open requests to a file by queueing the I/O Request Packet (IRP) to its
internal file control structure until the oplock owner lets it know that it is finished.
Tbf, this mainly seems to compare shitty/outdated 3rd-party Windows drivers against a vertically integrated system where Valve has complete control over both the hardware and software.
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.
Valve has done an amazing job turning Linux gaming a reality, I really admire their efforts. Unfortunately I had to switch back to Windows on my gaming rig due to a Steam bug which unnecessarily downloads the shader cache of several game every single day, disabling the cache hurts the performance. That’s been plaguing many users for years and surprisingly Valve hasn’t fixed it: https://github.com/ValveSoftware/steam-for-linux/issues/1028...
i wonder how much of that is simply due to how incredible dxvk is. I think intel even uses it as part of their windows drivers to translate some older dx version games to vulkan because it performs so much better on their cards.
I’m hearing that Microsoft has done a shitty job of tuning windows for the Lenovo device. Someone last week cited commentary that Microsoft has been working on it and they think they can reduce the Windows overhead by TWO GIGABYTES. Why is the fucking operating system using 2 gigabytes of memory in the first place, let alone enough more than that to be able to reduce memory usage by 2 GB.
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 I recall there were services that were part of the desktop experience that they didn’t shut off. One of them being the desktop wallpaper. Which with high res desktops being the rage these days may actually be a large part of that size win.
Seems more like a test of the hardware than Windows 11 and SteamOS since they ran into driver issues immediately. Not to mention those frame rates are terrible across the board. Just not very good hardware.
It's the same hardware on each test. The only difference are the drivers and OS in question. Lenovo has been slow to officially ship updated GPU drivers for this device, but the exact same SoC is used on a number of handhelds.
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.
Windows 11 is also bloated as hell by default. Curuios how they compare to a very optimized and debloated windows 11?
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)
and 3: its also windows 11 on the handheld - its not comparing a desktop (edit- or many desktops for that matter) with steamos on it vs some windows. (though i can see somebody debloating 11 and dropping it on the device - why not?)
> 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.
You can be sure that gamers are going to install SteamOS onto their desktops once it supports more kinds of hardware.
Yes gamers could install Bazzite right now, but those that are open to switching away from Windows aren't going to if they don't have a large company that can fund the support focused primarily on the issues that gamers are going to experience.
Related to 1, Microsoft did announce a gaming-focused "debloated" Windows 11 experience especially for gaming handhelds to release "sometime next year". It will be interesting to see what it is like as a standard, but obviously it is not out yet.
Stupid default Apps or an extra Windows Service or few isn't going to make any difference in how the OS performs (or the out of date AMD drivers they used on their Windows install).
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.
Microsoft is partnering with Asus to make the ROG Xbox Ally, which will run a stripped down Windows 11 that boots straight into the "Xbox" app, and you can switch to "desktop mode" much like how SteamOS / Steam Deck works. At least on Deck, it only boots the KDE desktop environment up when you switch to desktop mode so you aren't wasting resources on a windowing system you'll never see. It sounds like Microsoft is planning a similar setup, but only time will tell how much they manage to avoid enshittifying the plan.
What's the advantage for a consumer here though? I honestly don't get the selling point of why I would want Win11 specifically on a handheld. The desktop UI sucks with a touchscreen, so having a windows vs a Linux desktop on the hardware doesn't seem like a difference. I get why MS wants to try to compete here, but I just don't get what they could possibly offer. I don't believe Win11 can be stripped down enough to compete on performance, currently I can't get a machine with 4GB of RAM and a SATA SSD to perform adequately with a web browser in Win11. The OS just consumes huge amounts of resources on stupid background tasks that can't be disabled without registry tweaks and the undefined behavior that comes with that.
The person you're responding to wasn't entirely accurate, as it's not just a regular desktop session with the Xbox app set to auto launch. Early reports (admittedly from Microsoft PR) say the gamepad-centric window manager they're shipping on the Xbox ROG, a replacement for explorer.exe, shaves off 2GB of memory usage specifically because they're disabling desktop-centric services during the session. They do seem to be addressing a lot of your points.
The selling point is compatibility with anti-cheat and Game Pass. These aren't targeted for me, I'm not big into competitive shooters, I prefer a la carte and I main Linux, but I can imagine it would be for a lot of people.
I'm on Windows 10 LTSC which will receive updates until 2032. You'll likely have to add stuff to the OS to install the Windows Store and UWP apps but otherwise regular apps just work.
On the one hand I hope with the proliferation of such articles and sentiment that Microsoft would start paying more positive attention to Windows as an Operating System instead of an AI and Advertisement Machine.
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.
By Microsoft's loss and gamers' gain I mean that Linux used to be nonviable to play games, but Valve put a lot of effort into making games work on Linux (and continues to do so). They saw (and maybe still see?) Microsoft as a threat to Valve's future.
So now gamers have gained another platform to play games!
> but Valve put a lot of effort into making games work on Linux
Yes! this is competition at its finest, right? It is also what I was talking about in my original comment.
> So now gamers have gained another platform to play games!
Yup, this is definitely a good thing. Yes gamers get another platform and hopefully gets MS to focus on Windows to make it better and less hostile to users.
Gamers are also quite vicious and they will tarnish the reputation of anything that doesn't work well for them. Let's see how Linux will fare in their hands.
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
This story has been floating around for weeks already. I’ve seen two different videos talking about this already including one claiming Microsoft has already stated they are working on it. It’s not just that the games run faster but that battery life is also better. On a handheld system battery life will change minds more than frame rates.
Correction: it’s been a month and they had both devices:
The choice of games is also super strange. Why include borderlands 3? A 2019 cel-shaded release that isn't super demanding seems odd. Homeworld and returnal also seem odd, they're a little more modern but I don't typically see them used as graphical benchmarks. The only game on their list that I typically see for benchmarking is cyberpunk 2077, the rest of the list honestly looks like they did some cherry picking with their selection to massage the data.
It's not a terrible description of what Borderlands complex web of shaders are doing. A base layer part of it is effectively a cell-shader, but Borderlands is doing more other things, too.
Relatedly, cell-shading and other shaders are sometimes very heavy GPU compute workloads (which would also be far more expensive if they needed to run on the CPU due to a bad graphics card or driver). It seems funny to dismiss cell-shading as "not super demanding", when cell-shading was a technical dream or expensive technical demo thing to do for quite a while in the time before modern GPU shaders.
I don't think this is because Steam, Proton, or Linux are particularly good on their own (they are, extremely so) but more that windows is just so awful.
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.
I want games to run optimally too, but I don't think they should release a gaming version of Windows. Ultimately gamers want to do more with their PC's than just game.
They should focus on optimizing the current OS to more optimally run games rather than segmenting the OS.
About 10-15 years ago I played a game on Windows system and was dual booting to Linux.
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.
Gamepass is the only thing holding me on Windows, and in October, Windows 10 hits EoL. I think that puts a pretty clear end date on the end of my Gamepass subscription as I don't want to upgrade to a new Windows version.
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.
As a semi-related aside: I often read of people complaining that Fallout New Vegas crashes constantly for them, but under GNU/Linux using Proton I get none of that, I can play it for hours on end without any notable instability.
Can't say that Windows 11 compatibility on this regard is any way better. Anything pre Dx-9 is somewhat broken one way or another.
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.
It's definitely game-dependent. In a few cases I've found it was easier to rebuy a game on GOG, like Dino Crisis or '98 Resident Evil 2 (whose re-releases indeed run on win11 flawlessly while they're quite problematic on Proton) rather than messing with their original cd/dvds.
On the other hand the Sega arcade ports from the same time period (Sega Rally 1&2, Daytona USA Deluxe, Manx TT Superbike Virtua Striker, etc.) are definitely less problematic on Wine/Proton rather than Win11.
I've actually found the opposite in my experience, though I would assume it's very much case-by-case. A recent example was Hyperbowl, a turn of the century game that broke on every version of Windows post-XP. Works fine on Linux with recent Proton and dgvoodoo2, though.
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, I keep a physical vintage winXP machine around for games of that vintage because I find they don't tend to play super nice with modern hardware on windows either. I haven't switched my main personal desktop away from win10 yet due to compatibility with my Cad program, but playing games from that vintage was a nightmare IME. I dunno, maybe I'm unlucky with my selection not playing nice, but I found it way easier to just have a decent 2000's vintage PC hooked up with a KVM to my 3rd 1080p monitor for that. Bonus is that, since I'm playing those games for nostalgia anyway, it's better running them under XP anyway. I haven't gone as far as hooking up a CRT, just due to the space. But the second desktop is tucked away under the desk, so the only real downside is having a low red monitor hooked up. No big loss there, the tertiary monitor is mostly for slack and a media player otherwise so it doe3need to be nicer. Just something to consider, since PCs of that vintage aren't that expensive unless you want a high-end example.
I’ve been gaming on Linux for quite a while, and it’s overall been a great experience.
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.
Shame you have been downvoted. That's a reasonable question no matter where you stand. Motion is important too, not just static screen-shots.
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.
But on Windows you’re guaranteed that any game can be run without any extra configuration, which still is a big win. I know that proton handles most cases but it’s still frustrating when corner issue happens.
You can also achieve nearly the same level of performance by adding exclusions to Windows defender. I typically exclude the game process as I still would want folders scanned when updates / downloads are running. There's a noticable improvement to load times, streaming texture latency, and CPU utilisation.
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.
According to some benchmarking removing the storage filters offers a mild boost when compared to adding an exclusion and you don't have the benefit of some AV protection when other processes are accessing the volume like steam etc.
It doesn't make sense to benchmark a handheld gaming device w/ a native screen resolution of 1920x1200 at 2k or 4k. Likewise setting any graphics settings "extreme" like raytracing - it will run at slideshow speeds no matter what operating system. Besides, DOOM: The Dark Ages is a top tier graphics title released a month and a half ago; I think the selection of titles is decent for capability of the device.
You recognize, of course, that the conventional wisdom up until today (and even still in most people's minds) was that Windows was the OS to install if you want to play games. There's genuinely a movement right now that's saying, "Linux might actually be the all around better desktop OS if you are a gamer".
"To test the performance impact of this operating system choice, we started with the SteamOS version of the Legion Go S (provided by Lenovo) and tested five high-end 3D games released in the last five years using built-in benchmarking tools..."
Borderlands 3 was 2019. Homeworld 3 has a 38% on Steam and sold poorly. I highly doubt it ever received patches or optimizations. Again, these feel unbelievably arbitrary, as if someone just wanted to push a narrative.
Steam favorability percentages are famously vulnerable to review bombing. HW3 got swept up in culture war nonsense around LGBTQ representation.
I'm a hardcore Homeworld fan. I've run campaigns of their TTRPG, modeled their ships, played the old games to death. I found my own experience with 3 to be "mixed", it's hardly the best entry in the series, but the reviews absolutely are artificially low due to brigading.
Aside, an unoptimized game is actually one I'd want included in my benchmark. Games that have the teams and budgets to really polish will likely perform well no matter what. But how does OS level changes affect those other games, games where the developers didn't put in the care? Does one OS make those games worse? Or does it help with the shortcomings? It's valuable to have entries like that in your dataset.
Homeworld 3 received a terrible rating because its story and delivery was a massive departure to what users wanted, not because of it's gameplay or tech.
It would have been chosen for the same reason ashes of the singularity was chosen as a benchmark for so long: because it looks good, it comes with a benchmark, and it's really good at stressing out a particular part of the computer (for AoS: async rendenring, for HW3: CPU).
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.
Every simple code base / API is faster than mature API’s due to the fact they do less. A simple string handling library which isnt biderectional, doesnt cater to people with accessibility, doesnt take locale formatting into account, doesnt cover 100% unicode spec will always be faster than complex code that does.
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.
In my purely anecdotal experience over the last few years, performance ranking is as follows:
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.
Interestingly enough, I've had games that had both a native Linux port and Windows version, and the Windows version through Proton ran better than the native Linux version. This ended up being true for Civ5, Civ6 and Cities Skylines (1).
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.
Linux port if there is one is usually done by a third party porting studio, which is not necessarily at the same quality as the original codebase. Also the devs just don't have the manpower/bandwidth to spare for Linux users given how small this community is.
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.
These ports are also not usually source ports, so they're not much "more native" than the Proton ports. They often use the same kind of API translation layer, probably also built on WINE. I think as Proton sees more investment and becomes more advanced, it's probably becoming difficult for competing compatibility layers to keep up.
A source port that is optimized as lovingly as its Windows counterpart will probably be faster than the Windows version running via Proton, but the incentives aren't generally there to justify the costs/difficulties. Maybe some day it will be! That would be wonderful.
But until then, Proton seems like an increasingly compelling option for these compatibility layer-based ports of Windows games.
Factorio and Minecraft (Java edition) are two of the few games that come to mind where the Linux port got comparable effort to the Windows port, and I don't think people are in a rush to play either of them in Proton.
Your latest AAA open world RPG on the other hand? Yeah, you're probably going to have better luck in Proton even if it gets a native Linux port.
Factorio is the only game that I know that have Linux (and Mac) exclusive feature, which is non-blocking saving: https://www.factorio.com/blog/post/fff-408
> and get their game running perfectly under proton
Even better would be to compile for linux, but use DXVK-Native (https://github.com/doitsujin/dxvk#dxvk-native) if you think migrating from DirectX to Vulkan requires too much effort.
They probably QA mostly for windows, so they run into bottlenecks and edge-cases of windows APIs during QA. Linux-native APIs probably have different bottlenecks and edge-cases.
I think the reimplementations of Windows APIs in Linux, even though alternative to the original, should have similar bottlenecks and edge-cases. So the extra QA on Windows helps the Proton version more.
But maybe figure out how to start getting those third party Linux porting studios paid to work on Proton...
The answer is pretty simple here - hire CodeWeavers to work on supporting your game in Proton/Wine rather than some other porting shop doing an old rewrite-style port.
To be you should compare the windows version on windows, no proton against the Linux version. DXVK, which proton uses, makes some games run better in windows than "native".
> 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.
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.
To the extent that Linux doesn’t support nvidia gpu it is actually Nvidia not supporting Linux and keeping their drivers proprietary.
But that doesn't matter. If the feature isn't there, Linux is non-viable if you want/rely on said feature. It doesn't matter whose 'fault' it is.
It does to me. I don't buy their products, simple as that.
> 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.
> HDR
Already supported
> Nvidia GPUs
You have it the wrong way around. NVIDIA had issues supporting Linux, not Linux supporting NVIDIA. AMD drivers work fine, so its not a linux specific issue.
> VR
SteamVR works though?
> > HDR > Already supported
Is it though? I confess I haven’t tried in a few weeks but until last time I did, to get HDR in games you had to start a session with `gamescope` rather than a DE, and still had to set a bunch of flags - and in some ways have a very subpar experience with problems with mouse movements and other issues I can’t recall.
I exclusively game on Linux and I find the experience far superior than doing anything on the other OS, but last I checked HDR was not actually supported.
GNOME and KDE support HDR now too, so if you use those DEs you dont need gamescope anymore.
KDE has supported it for a while on the desktop. Still needed Gamescope to make it work in games. Like I mentioned, I haven't tried it in some time so it may have changed.
It works on the Deck.
The Deck runs Gamescope.
For gaming and general desktop on Linux AMD is best if you want a dedicated GPU.
If you want a laptop with good battery life Intel is generally the way to go.
A lot of this is due to the enormous amount of effort Valve put into improving the open source AMD drivers, which is what is used on their Steam platform.
Of course if you want CUDA you need Nvidia, but if you use Nvidia to drive your Linux desktop expect some suffering to go along with it.
For what it’s worth I’ve been using an RTX 3090 and it’s been mostly smooth sailing for a couple years now.
Running NixOS with a pretty vanilla configuration and it has been hassle free.
I did have to disable power management at the system level because framerate suffers severely if the system sleeps and wakes back up, but I shut the system down when I’m not using it, so this was a non factor for me.
An AMD APU without a separate graphics chip is a much better option on laptops these days than Intel.
> SteamVR works though?
Not at a level where the experience is more fun than frustation.
It's getting there though. I own a high-end PC with nvidia GPU and I play VR on my Linux setup via ALVR (I own Quest 3) It's not straightforward and full of workarounds I have to do, but once you're in the game it works great
By the time it gets there there will be tons of other new features that Linux won't support.
Doubtful but ok.
It's not lacking features for me, it's lacking polish
Feature-wise the main missing feature is kernel level anticheat which I personally don't care about
AFAICT HDR is supported, like on the Steam Deck
HDR on Nvidia with Linux is still very glitchy, I've had the driver crash a few times trying to use it.
Ah good to know.
I think Windows isn’t that different, just that there's more motivation for NVIDIA or Microsoft to fix those things. I recall not that long ago a combination of Windows 11, my NVIDIA RTX 40xx, my previous Dell Alienware monitor also had some issues with switching between SDR and HDR (and later Dolby Vision brought even more of a mess).
Meanwhile Android and iOS phones have been able to do it flawlessly for a while now…
Doesn't support NVIDIA GPU's!? Is this a display or gaming specific thing?
All the ML people are using NVIDIA GPU's on Linux.
There are indeed nvidia drivers for Linux and they're reasonably good for gaming, but the feature set sometimes lags far behind windows. There is no DLSS 3 for Linux, for instance. (as of a few months ago anyway - I haven't checked recently)
There is though. I'm playing Hitman 3 with all RTX options enabled, DLSS and frame generation
Ahh rightio. That's a shame.
Nvidia support across the desktop ecosystem is poor, for example practically nonfunctional in Sway. And just buggy in other Wayland based desktop environments (kde seems to be the best in my experience).
WRT Nvidia+Sway this was certainly true not so long ago. But since the latest Ubuntu release and with a recent driver I am running this combo and it works flawlessly.
Really! I'm happy to hear it, I'll give it a try on my desktop this weekend. I would love to get all my machines on the same desktop environment once and for all.
I'm doing PCI passthrough of a 1080 to an archaic tiling X11 window manager and so far it Just Works with noeveau. XFCE also worked fine before I decided I don't want a full DE. Rock stable. I will move dists before I move to Wayland by the looks of it.
i3 should be pretty easy switch from sway if you haven't tried.
> noeveau
Hasn't Nvidia locked most functionality away from the open source drivers?
You either have the choice of using drivers that work well with Linux, or drivers that are fully featured.
I have Debian with KDE Plasma 6 and I have enabled HDR
Can you even watch decent Netflix on Linux yet?
Up to full HD, depending on what Netflix streams out. But this has nothing to do with graphics drivers or GPU performance.
Yeah, but it does have to do with graphics in the linux desktop ecosystem and is particularly relevant to those without a dedicated gaming machine.
It doesn't have anything to do with graphics in the Linux ecosystem. Netflix specifically blocks Linux from having decent quality so it's kind of pointless to discuss. If you want high quality, you can pirate it or rip from disc. Dirt cheap n100 minipcs are capable of playing UHD bluray rips in Linux just fine for example, so Netflix's relatively low bitrate media aren't an issue.
Not an issue except for any of the couple hundred million people who subscribe to Netflix at any rate.
Not an ecosystem issue at all. There's no "yet". Linux computers even on the very low end are already perfectly capable of playing 4k Netflix videos. You can easily prove this to yourself by downloading one via torrent (you can generally get exact stream rips with DRM removed if you want). Netflix just won't send UHD streams to Linux users. That's a political choice, not a technology problem, and it's easy enough to get the media elsewhere DRM-free if it's that important to you since Netflix evidently specifically does not want Linux users as customers.
I wonder of this might be due to your Linux nvidia driver (nouveau?) pinning the card on baseclock by default while the Windows one will allow it to scale up? Something I heard somewhere that seems applicable here.
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.
Happens with whisky and macos in my experience. It is like as soon as a game is installed and you do something like chuck a grenade, no explosion the first time ever you do that.
> 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.
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.
I have a 65 watt ryzen 9 system on chip (8945hs, I think) minipc and make heavy use of it for linux gaming.
My guess is that Nvidia’s linux video drivers are still substandard.
I have a laptop with the same GPU, and Diablo 4 runs really well out of Lutris. Graphics version 570, and the CPU is an AMD with a Radeon 680M integrated. I often play games with FSR on, which probably keeps performance higher?
FSR does indeed increase the performance a lot, before FSR 4 with a significant cost to image quality though.
For your system, the integrated graphics should also be quite capable. More so on Linux, thanks to the driver advantage AMD has here.
Side note, Niri is a fantastic WM. When I saw the Phoronix article on HN talking about the addition of overview mode and more, I finally took the plunge and spent an afternoon converting over from Sway.¹ Anecdotally, I've seen less hangups on Niri around fullscreen games and floating windows, perhaps thanks to X11 running in xwayland-satellite.
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.
> X11 running in xwayland-satellite
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.
I've been so happy with Niri after many many years bouncing around other WMs. It addresses the main issues I've had with other tiling window managers and has been such a joy to use.
The scrollable aspect just feels so natural and intuitive to me.
Steam won't launch for me in xwayland-satellite here... I just assumed steam+wayland = broken. I have a kind of weird setup using sway with xwayland disabled and running xwayland-satellite though.
Try running xwayland-satellite in another WM like Niri and see if it works there – for example, Gamescope didn’t work well on Sway, crashing as soon as Steam tried to spawn another window, but it’s working fine in Niri.
Been a Linux gamer for years now and I think you are correct on your frame rate observations in general.
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.
Why do you think ZFS helps? I’m guessing you have compression turned on? IME, ZFS is rarely better in terms of raw performance, compared to e.g. XFS.
Yep, compression is on and I think that's what does it considering how large games are these days.
I stay away from XFS, every time I have used it in the past my entire drive have ended up scrambled within a few months. It's by far the worst file system I have ever used, not even FAT32 was that unstable for me.
Most large games are already compressed in cabinet files or similar so OS/file system level compression wouldn't gain you anything. While there are games that ship with loose files, like the Elder Scrolls series, other games like Diablo IV or CoD use archives.
From experience having an L2ARC SSD, especially if it's nvme, can really help with zfs performance. I'm curious if they have that in their setup.
nope, but I use LZ4 compression and have plenty of ram.
Has he tried DevDrive in async mode or with filters disabled entirely?
> The biggest thing I noticed when switching from X11/Xfce to Wayland/Niri was just an overall increase in framerate.
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).
Some of the games I play often that saw improvements: The Finals, Overwatch, Rocket League, Helldivers.
> (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:
Hm, i tried it and i could set up the higher refresh rate but Niri was forcing vsync on the game that capped the framerate to my monitor's refresh rate (165Hz) even though in the game i have vsync disabled. This seems to be an issue with Niri as KDE Plasma Wayland disabling vsync works fine.
At "ultra" settings i got around 115fps which is ~5 fps lower than 117-120fps i got from Xorg/Window Maker and KDE/Wayland, though i'm pretty sure that was just the forced vsync, so in practice it seems that the window system doesn't matter much.
Did you use XFCE's desktop compositor? AFAIK XFCE's compositor isn't particularly great, some years ago when i was working on a custom game engine i had to add an explicit option to use the X11 "override redirect" flag instead of the window hint for fullscreen windows because XFCE's compositor wouldn't disable itself otherwise and the game would feel a bit laggy/inconsistent. Not sure if this has been fixed nowadays but in general it gave me a bad impression for XFCE's compositor as other compositors didn't seem to have the same issue.
Is there a way of making wayland actually usable with Nvidia GPUs? I never manage to make it work, and it makes the whole system feel slow and sluggish compared to X11
Try avoiding the open source kernel modules, I had a similar issues.
> 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.
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.
Try winewayland + Wayland.
The last missing piece for full Linux gaming is anticheat. Last I looked into it, the major vendors don’t want to support it due to lack of kernel security and the ones that do, game devs refuse to allow it (destiny for example)
One we can play AAA games I am literally ditching windows forever. Steamos is the best thing that has happened to gaming
Anti-cheat today is a stop-gap measure at best. For various reasons such as improved OS security and security concerns with this software, ring zero anti-cheat won't be around forever. Besides, it's a cat and mouse game where the vendor is the mouse.
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.
> Besides, it's a cat and mouse game where the vendor is the mouse.
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.
With the emergence of AI cheating, cheats don't even need access to memory anymore. The cheat can entirely run on mouse and screen peripherals and the computer will have totally no idea what's going on. The best you can do is behavior analysis. But it always comes with chance of misreports.
Direct Memory Access cheats will always perform better as they can reveal the location of an opponent before they're even visible on the screen.
Why are players who aren't on screen even in the client memory? The server shouldn't be sharing sensitive information.
Although that looks like an obvious solution on first glance it's not really technically feasible. Things like gunshots or footstep sounds are not visible to the player, but still need to be relayed to the client.
As far as I see the only way around not sharing anything that's outside of the immediate perception of a player is to have the audio and graphics be entirely rendered server-side.
You can minimize that to some degree (Valorant does this), but due to movement prediction/network latency you do have to overshare a little bit.
I imagine that most game devs just look at the incredible amount of work this takes to implement and complexity it adds, and decide to not bother. Valorant can do it because the game itself is low complexity, the developer has deep pockets, and also the added competitive integrity is valuable.
Game servers are complicated and have a lot to manage.
It's infeasible for the server to keep track of each player and do frustum and raycasting to every other player to check who can see who every frame.
Culling out of view entities also has the problematic effect of when a player spins around you now have to stream in several big chunks of world state in the few milliseconds before the user clicks to get that 180 no-scope.
How else would the game render the enemy player's sound around corners and in adjacent rooms without knowing their location for instance?
DMA hardware and cheats are getting more and more accessible. It's not just chosen few anymore
> 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.
Working on mostly server platforms, I had forgotten that IOMMU enablement (and, where relevant, enforcement) was not the default.
Consumer hardware and software is terrifying.
Not sure how that's relevant, unless you find it terrifying that owners of hardware have control over their hardware.
In my world, we won't let a system boot with production credentials unless the IOMMU is enabled.
This is enforced by a greatly enriched TPM (and it's willingness to unwrap credentials). We have trust several layers of firmware and OS software, but the same mechanism allows us to ensure that known-bad versions of those aren't part of the stack that booted.
If I wanted secure games (and the market would tolerate it), I'd push for enforcement of something similar in the consumer space.
It's your IOMMU, you can do what you want with it. Maybe you need to write heaps of stuff to take advantage of it, but what's new there?
The only thing you're getting by saying "no IOMMU" is "I want any devices in my machine to be able to do anything, not just what I want them restricted to".
Okay, but he's specifically brought it up in the context of a computer's owner doing something that the software vendor (and also myself as another gamer harmed by cheating) would prefer he did not.
And if they want complete control, they can choose not to use a vendor and do it themself, for all the control they could want.
Hooray, freedom!
I have no idea what you're trying to say
> unless you find it terrifying that owners of hardware have control over their hardware
I mean that the presence or absence of an IOMMU doesn't impact whether owners of hardware have control over their hardware.
It just means that the owner of the machine is able to limit what memory the devices in their system are able to access, in the same way that MMUs limited what memory every process on your system could access.
> they now use Direct Memory Access over the PCI-E bus
Do you have any good resources with keeping up with this kind of thing? Seems like a fun topic to learn about
As long as games are running on user hardware/OS, you'll always deal with cheating. Server-side checks and computation can only go so far.
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.
Exactly, nothing short of streaming the entire game fully rendered from the server will stop cheats. And even then you can probably still do aimbotting with modern day computer vision.
This reminds me of a discussion around 2 decades ago, where someone showed a picture of his "undetectable aimbot" for a turn-based artillery game: a ruler, a page of charts, and a handheld calculator; followed by a copious amount of discussion of whether that was considered cheating.
I hope this was for gunbound, lovers that turn based artillery game.
How exactly will it stop cheats? Any skill based game can still be cheated. Just analyze the video stream, or go even lower tech, point a camera at your screen. Many games can be effectively cheated like this. For eaxmple Aimbots in counter strike and peak human reflexes in dota/lol.
Surely the server can tell the client what sounds to play and what other environmental hints to do, just as well as the server can choose to tell the client where the other players are when they are in sight.
If the server says "there are footsteps from these coordinates" then it's telling the client there's a person at these coordinates.
Due to latency, the client has to at least do a little prediction.
Maybe, but I think that would've been done already if it was feasible.
The storage read, memory bandwidth, load computation, and gamethred pause to add the object to the game world is far more expensive than sending a move.
I've always thought the line about whats cheating, and what's not is unfair and arbitrary. How is it ok that some players can play 4k 200fps and others 1080p at 30fps.
The only way to be really fair is for everybody to Stream the game at the same res, frame rate and latency.
In certain competitive environments framerate is definitely limited. Here [0] are the rules for Fallout 4 any% speed runs, framerate must be capped at 60FPS. AFAIK that rule applies to all games in this engine due to physics behavior. I don't follow tournament FPS games, but it wouldn't shock me if there are rules for competitive play there as well.
If you are asking why games like counterstrike don't have limits on online play, that's mostly a commercial question. Would those games be as popular if they limited performance to what was achievable for minimum specs? I certainly wouldn't want to play at 1920x1080 on my nice widescreen monitor, but setting the minimum to a $1500 monitor and the hardware to drive it would guarantee very few players.
[0] https://www.speedrun.com/fallout_4?h=Any-Full-game&rules=gam...
Edit:typo
Yeah and in real world people from different countries with vastly different economic backgrounds compete on the same stage, I think video games are fine.
This isn't exclusive to video games. Much of the improvements to world records in sports are due to improvements in gear, yet we don't consider those records to have been unfairly achieved.
Some games do impose limits though, for example Overwatch doesn't allow you to use an aspect ratio larger than 16:9 and selecting a wider aspect ratio actually cuts down on your vertical field-of-view rather than granting you more horizontal field-of-view. This lessens the potential advantage of ultra-wide monitors.
so consoles are better
How would consoles be any more immune to computer vision based cheating? Instead of feeding the output to a spoofed keyboard & mouse, you'd just be feeding it to a controller input. I'm not really seeing any difference in technical challenge here, and you wouldn't even need esoteric hardware since console controllers are USB devices anyways.
Since the hardware is better controlled and secured, and hardware attestation is a solved problem these days, it's not particularly difficult to enforce security to the point where you'd need to hardware hack a controller and connect it to a physical camera to bot.
That's still gonna be annoying for players, but it'll greatly decrease incidence, and if reporting a player for botting requires buying and hacking a new controller... It should be quite effective.
Controller-based cheats have been around for awhile, like Cronus which allows scripted actions; reduced recoil in CoD, for example.
Sure, but that's because they're not protected. They could easily be protected.
Someone doesn’t know about the chronus zen or how big console cheating is!
Let’s just say that my finals experience isn’t the same as yours! ;)
They are often more convenient and secure. If you don't mind a single-purpose device that severely limits your ability to modify your experience. Better is subjective after all.
> Anti-cheat today is a stop-gap measure at best. For various reasons such as improved OS security and security concerns with this software, ring zero anti-cheat won't be around forever.
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.
It might be a cat and mouse game, but that doesn't mean we shouldn't be trying.
Any anti-malware software ends up ultimately being a cat and mouse game, but that doesn't mean we stop updating our signature updates.
The goal isn’t to stop 100% of cheats but the majority of them and that’s fine. Either way, it’s the only thing stopping me from playing the rest of my games on steamos.
Multiplayer games without dedicated servers is dead end anyways. I dont need a "anti-cheat" daemon hooking into kernel scanning files and other memory while playing a game. Communities in dedicated servers are much more efficient at moderating the player base than centralized match making ever will be.
This is where I'm at with gaming. Even outside of cheating, it's not fun to me to be dumped in a game with screaming children/manchildren. If I'm playing a game I want it to be with my actual friends. And then I don't have to worry about them running cheats because I trust them.
Once you get to match making, global ranks, etc it's just getting too sweaty and ruined by cheating/low trust/etc.
At least in TF2 the sweats are all on the community servers. I can get like 3 kills a round on skial servers. I’m trash there. When I use the valve matchmaking servers I dominate the lobby about half the time.
Communities with dedicated servers include anti-cheat though. Most people aren't interested in spending time moderating a player base: they'd rather just play the game. So server admins use anti-cheat.
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.
Yeah, but it's very time consuming/impossible to find similarly skilled players for a fun lobby. The only competitive game I care to play on Linux is Rocket League, which is nearly impossible to cheat at, so it doesn't currently have anti-cheat, but I wouldn't be surprised if Epic decides to put their beloved EAC in it at some point anyway, maybe even just because they hate Linux so much.
EAC has a Proton-compatible version, actually. Even if they enable it, if they use that version, it'll work fine via wine/Proton.
They really aren’t. I used to think that but I actually enjoy skill based matchmaking. It makes game availability better and faster and I don’t have to deal with 1 outlier absolutely stomping or overzealous admins or whatever. I like both approaches though for something like battlefield I think dedicated servers are better but for things like cod, siege, etc we need sbmm
Ho do you do proper matchmaking and ranking up and progression etc with dedicated servers?
I want good balanced matches with players of my similar skill level via matchmaking.
It's also an anticompetitive red herring, at least for Epic.
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.
This is especially relevant to note because Epic bought and owns Easy Anti-Cheat (EAC) one of the currently most popular anti-cheats in AAA.
ETA: EAC still supports Linux gaming today, but the rumors remain that Epic could remove that at their whim.
Epic are a hostile company.
Just avoid it.
Missing anti-cheat is a feature, not a bug. Linux not allowing games to install kernel malware is a good thing.
Linux definitely allows it.
This is problem for me and my brother right now. He's up at a remote job site and we want to play Siege or Apex together¹, but both require anti-cheat and don't support Linux. And I'm loathe to devote space on my SSD to Windows.
¹: Rainbow Six: Siege and Apex Legends, respectively.
>the major vendors don’t want to support it
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
There are rumours of next xbox generation supporting steam platform and 386 architecture. I know it's a bit off topic, but it could be an elegant solution to the cheating problem, gradually move to standardised consoles. This could solve the dma problem too
AFAIK most anti cheats such as BattlEye actually work under Proton, but the Game Developers have to write to the Anticheat vendor that they want to opt-in into allowing Proton.
Don't forget the ancillary applications that gamers want. If you follow Discord's website, you're gonna end up installing a DEB file manually. Then, every couple weeks, Discord won't launch until you go download another DEB file and install that. Oh, and good luck getting Discord screen sharing working on Wayland. I tried for hours, gave up, and switched to X11. So, just in Discord, we've already run into two hideous workflows that no Windows native is going to take in stride.
And certain types of games have a _ton_ of ancillary applications. For flight simulation, I rely on 2-3 additional contollers, some of which I am fairly certain either won't have driver support, or at the very least will have some major issues with the GUI and configuration.
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.
Discord is shipped in a number of package managers (I don’t know the status for mainline apt repos).
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.
...why? Discord is available via flatpak.
It Just Works.
Tell Discord that!
Think about it from the POV of a Windows user, especially one who has never used Linux before, and especially one who doesn't know what HN is. To install a program, the first thing you're going to do is type "discord" into your browser, and go to their website. Discord's website doesn't suggest that there's a better option. It just gives you a DEB file.
Made the mistake the poster above did with discord did for years.
The package system is very important to learn in Linux. People have 12 ways to install an app, and they are far from equal.
It's funny because one of the OG draws to Steam back in the day was because Counter-Strike (et al) had the superior Valve Anti-Cheat (VAC) in Steam copies than the myriad of raw CD installs of HL1 plus the mods to run Counter-Strike.
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).
Don't discount peripheral support. I've got some pieces of hardware that are only kinda-sorta supported in Linux.
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.
interestingly, I have no issues with the anti-cheat within Marvel Rivals; however, games that embedded an anti-cheat prior to the steam deck popularity don’t work as you described (PUBG, apex legends).
Given that Windows games run faster via Proton on SteamOS, developers should prioritize targeting SteamOS APIs—not Windows. This ensures compatibility with Windows while maximizing performance. Game engines like Unity and Unreal must adopt SteamOS as the primary target, with CI systems rigorously testing both platforms. SteamOS, not Windows, should be the baseline for optimization.
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.
I'm not sure that makes sense since the Windows API is the source of truth for how something works. If you make a game that works on Windows but not in Proton, Valve will push a fix that makes Proton work the same as Windows. But if you make your game work with Proton, but not Windows, you are relying on some quirk of Proton which isn't guaranteed to work in to the future and as soon as something else needs it to work the same as Windows, your game will break.
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.
You would need to test on both of course. I am arguing that one should target the fast happy-path on Proton as Proton is a subset of the Windows APIs that runs faster than Windows.
Proton isn't a subset of the Windows APIs though. It's very likely that you could end up depending on behaviors which only exist in Proton.
That's very much not what I expected. Do you mind pointing out something Proton implements that windows doesn't?
While they are wrong, in that Proton is absolutely a subset of Windows APIs, it is describing aiming for an implementation detail.
Yes call X is faster than call Y in Proton, but that could also be because its only 50% written and skips a bunch of side effects that will be created after you have written your game.
Therefore you need to view Proton as a potential moving target. Not that Windows isn't too, but its just not as clear cut as is being claimed.
I know this because I pin some games against older versions of Proton because they work better/faster.
Thank you. I didn't do a very good job of explaining why I think Proton is worth targeting specifically if perf tests show it running faster than Windows.
The Windows API perf topology is astronomically large.
That Proton is faster than Windows can't be universally true, but if you stick to that subset that has good conformance with Windows APIs and is also faster, that should be the target (and everything is moving of course).
Cross benchmarking games between Windows and Proton, taking API traces and finding those subsets of APIs can usage styles will enable a game developer to target those Proton APIs that are both conformant and fast. It is that subset of Windows APIs (temporal spatial structural usage) that I am suggesting being the target. The API linting tool might itself compare API traces. In fact for those API traces from performant AAA games, I could see generating a header file that only exposes the same API as used by that game.
The only stable ABI on SteamOS is Win32. Targeting anything else is asking for it to break over time.
Valve actually ships "stable ABI" Linux runtimes that Linux games and Proton are built against, so that they don't have to rely on distribution packages, similarly to Flatpak. They renew them every few years but they stay stable so that games built against a specific runtime keep working.
https://github.com/ValveSoftware/steam-runtime
That's not an ABI.
The ABI lives below Steam.
https://en.wikipedia.org/wiki/Application_binary_interface
I only referred to the API as exposed by Proton, no argument was made for targeting a different ABI.
Not sure Epic, who owns Unreal engine would be happy with optimizing for SteamOS and its API's after the whole debacle with Epic Store vs Steam.
Ignoring that, you know, 99% of users are running Windows and not SteamOS. Test on what your users run. Proton is just an implementation of Win32, you're still just targeting Windows.
Only 95.45% now (https://store.steampowered.com/hwsurvey) but yeah.
Over what time period? 99 -> 95.45 is a huge absolute number. Is most of that attributable to SteamDeck? Would that mean that 4.5% of active Steam users own a SteamDeck?
If you click on Linux only and expand the details, SteamOS accounts for 31%. "Other" is the next biggest at 23%, and Arch (which SteamOS is built on) at 10%. Someone probably has made a comprehensive historical spreadsheet but the internet archives save the page frequently if you want to see trends. Even historically though Mac has commonly been known for being around 2%, at times more, it's only recently (like in the last couple years) that it fell below 2% and Linux passed it. SteamDeck did have a big impact for sure, though. I've seen estimates for around 4 million sold so far after three years, that's not bad. Putting some extra effort into your game's release to get it "SteamDeck Certified" seems worth it to me just to get that little bit of extra attention from those customers.
Back in the Windows XP days, I discovered that running Windows in a VMWare VM, hosted on Linux, was faster than running the same version of Windows bare on the same machine.
I never came up with a good explanation for that.
Cache.
Disk cache, to be precise.
please elaborate
I am guessing that they're implying that a VM effectively loaded the "hard drive" that Windows uses into a RAM-disk, so one would be comparing loading dlls and whatnot from RAM vs loading it from spinning rust.
I'm not sure how true that is, because in the Windows XP days most of us wouldn't have had enough RAM to spare to do that.
More to the point, if the VM could do that caching, why wouldn't Windows be able to do the same level of caching on bare metal?
Sorta. The devil in the details here.
Depending on the VM technology they use they offer a variety of different caching mode and configurations, but the basic three approaches that most everybody offers are going to be something like;
"writeback" means that when the guest's storage data ends up in the host's cache it is reported back to the guest as 'written'. This means that from the guest's perspective the disk is written to, but in actuality the data is still floating around in memory. If the guest wants data to be 'safe' they need to issue additional flush commands.
"writethrough" means that the host is using its memory for caching file system, but that writes are reported as 'competed' only when they have been committed to actual disk.
and "none" means the cache is used as little as possible.
So if your guest's virtual disk is in 'writeback' mode it isn't actually writing to real disk. It is writing to memory. Which is going to be very fast up to the point were the cache on the host is exhausted.
Certainly Windows could lie to applications and not write information to disk and keep it in memory much longer then it normally would but that would defeat some of the assurances that file systems are supposed to offer to applications.
'writeback' would be closer to what Windows already implements on the OS level, but because Linux is just plain faster it should improve performance somewhat. Microsoft can only work in improving the performance of Windows to get the same results, but Linux is pretty hard to beat.
'none' is what I use when running Linux on Linux because having two layers of cache is just kinda wasteful and doesn't result in real improved performance.
So if I understand correctly, you're suggesting that in essence the VM lied to Windows to improve performance? It's been many years, but I don't imagine I would have chosen an option like that. I am very surprised VMWare would default to such behavior.
> So if I understand correctly, you're suggesting that in essence the VM lied to Windows to improve performance?
Pretty much. Just speculating because I don't know how your systems was configured back in the day.
But all of what I said applies to most VM solutions.
If you are dealing with enterprise-grade hardware it isn't a bad things. Keep in mind that typically OSes are going to assume that they are the only things operating on storage. So if you have like 30 windows boxes all trying to write to shared disk at the same time it can lead to some bad behavior if you are not using a write cache.
The system has battery backup and typically they expect you to use shared SAN or NAS with multipath and/or bonded network interfaces for redundancy as well as having backups. So the chances of data loss is a lot less then typical consumer hardware and it gives the hosting OS better chances at optimizing and scheduling writes properly.
Also remember the guest OS still has the option to send a 'flush' command to disk, which would ensure the writes complete regardless.
> It's been many years, but I don't imagine I would have chosen an option like that. I am very surprised VMWare would default to such behavior.
This is pretty normal. You'll see the same sort of options with hardware RAID devices and such things with their own internal battery-backed cache.
Recently switched as well (Arch not steamOS, which is arch based) and it's been pretty solid.
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
Some of these games are practically neck and neck for performance. I'm wondering if it's a similar situation to early Proton comparisons, where framerates were higher in Proton but when comparing still for still you could tell it just wasn't actually doing certain effects. Are there features that are being attempted in the Windows version that are just not functional and thus effectively disabled on the Proton one?
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.
To my knowledge, this hasn’t been the case for years, and I’ve never noticed any extra visual glitching on Linux.
Might be unfair to call Proton a "translation layer" because the Win32 API is not defined in terms of system calls but rather a set of functions exported from a DLL.
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.
https://www.winehq.org/ calls it a compatibility layer that translates calls on the fly.
so 'translation layer' is not that unfair.
Clearly they don't want to come out and say that Wine is an "implementation of windows apis", because that would invite legal issues. But clearly this is what Wine really is for a large part. Some stuff are just shallow shims to Linux apis, while other stuff they need to make more of their own implementation of.
"WINE Is Not an Emulator" has always said something exactly like that. It hasn't exactly been a "secret" that Wine is an implementation of Windows APIs (and not an emulator/translator).
If it is forwarding to libc() as opposed to syscalls directly than maybe ‘translation’ is fair.
I think that's how it started out, and also how a lot of developers still conceptualize it. Wine has had to massively expand that scope to reach the maturity it has now. I think it's kind of straddling the line between "Implementation" and "translation".
Philosophically its still a translation layer though. It doesn't really care about correctness if the no apps depend on it. Success is in meaningfully running client software. The implementation of the Windows Libraries are just a way to get there.
I suppose you could say that Wine has "aged" well :-)
I really have to respect Wine's persistence (with Proton's help).
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.
Does it implement sscanf() with accidental complexity of O(n^2) for compatibility?
Wine is translating Windows ABIs (not APIs) into underlying Linux OS and userland. Translation simply means that normally Windows ABIs are meant to be used on Windows, they aren't native on Linux.
Proton/wine also implements many of those NT syscalls because windows programs do use them directly as well
There is also NanoX. Not fully Win32 replacement, but I think it is a good start.
Wine/Proton it's just another Win32 implementation for Unix. Win32 it's a subsystem on top of Windows NT too.
Also, most games are still 32-bit EXEs, so on most PCs today they are running not just in the Win32 subsystem, but the WOW64 subsystem (Windows-on-Windows 64) the compatibility layer of running 32-bit Windows applications on top of 64-bit Windows and its Win32 subsystem (it's still called Win32 on 64-bit, an unfortunate naming bug from massive compatibility break differences between Win16 and Win32 that didn't exist in the 64-bit transition, 64-bit didn't get a new API subsystem, it just upgraded the existing one, mostly).
Are we surprised?
https://blog.zorinaq.com/i-contribute-to-the-windows-kernel-...
https://news.ycombinator.com/item?id=38936868
When SteamOS and Ganoo/L00nockz become first-class gaming citizens, that's when I'm building a gaming PC for the first time since 2012.
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.
AAA releases have worked for years with the exception being about 60% of online games with anti cheat. Linux has been a first class gaming citizen for years if you use steam client and AMD gpus.
And that online part is the part that really matters. That's where the games make their money.
We've been there since the release of the Steam deck. Really, the only things I can't run on my Linux machine are the games that intentionally break the game with anticheat mechanisms. And it's not all anticheat - there are plenty of titles with anticheat that run just fine.
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).
I have always wanted Windows to move to Unix so that I can have the best of both worlds(software development and video games). Glad that we are close to that reality.
You'd also need a sane filesystem.
Windows has NTFS and ReFS. Both are quite sane and performant.
I know. Neither are either. git is noticeably slower, the locking is untenable, the case-insensitivity is a source of bugs.
From the NT Design Workbook, oplocks.
> Because the NT OS/2 I/O system is asynchronous by nature, the ability to make a request and then have it completed at a later time makes it natural for implementing oplocks. Further, because synchronization is required by the file system to determine when the caller has completed its oplock update transfers, the file system can use this feature to block open requests to a file by queueing the I/O Request Packet (IRP) to its internal file control structure until the oplock owner lets it know that it is finished.
Both file systems support case sensitivity.
Use DevDrive in a virtual disk or secondary volume, there are significant performance gains for things like git, nodejs modules, etc.
It helps to know the system. The perf would be an issue with any file system on Windows due to the file system filter architecture.
Locking is a function of the NT executive and not of the file system. It was a design decision. I’ll see if I can dig up the reasoning later.
[dead]
Tbf, this mainly seems to compare shitty/outdated 3rd-party Windows drivers against a vertically integrated system where Valve has complete control over both the hardware and software.
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.
They even have the audacity to use quotes for "official" drivers. At least it's an obvious indicator of the author's bias...
Where do they use quotes for the official drivers?
afaik the Steam Deck is by no means "vertically integrated" as e.g. Apple Products are. This is really more a quality thing on the side of Proton.
Valve has done an amazing job turning Linux gaming a reality, I really admire their efforts. Unfortunately I had to switch back to Windows on my gaming rig due to a Steam bug which unnecessarily downloads the shader cache of several game every single day, disabling the cache hurts the performance. That’s been plaguing many users for years and surprisingly Valve hasn’t fixed it: https://github.com/ValveSoftware/steam-for-linux/issues/1028...
i wonder how much of that is simply due to how incredible dxvk is. I think intel even uses it as part of their windows drivers to translate some older dx version games to vulkan because it performs so much better on their cards.
I’m hearing that Microsoft has done a shitty job of tuning windows for the Lenovo device. Someone last week cited commentary that Microsoft has been working on it and they think they can reduce the Windows overhead by TWO GIGABYTES. Why is the fucking operating system using 2 gigabytes of memory in the first place, let alone enough more than that to be able to reduce memory usage by 2 GB.
Compatibility is a big answer.
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 I recall there were services that were part of the desktop experience that they didn’t shut off. One of them being the desktop wallpaper. Which with high res desktops being the rage these days may actually be a large part of that size win.
I wonder how much better DXVK-native (without wine) Games would work.
Seems more like a test of the hardware than Windows 11 and SteamOS since they ran into driver issues immediately. Not to mention those frame rates are terrible across the board. Just not very good hardware.
It's the same hardware on each test. The only difference are the drivers and OS in question. Lenovo has been slow to officially ship updated GPU drivers for this device, but the exact same SoC is used on a number of handhelds.
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.
Windows 11 is also bloated as hell by default. Curuios how they compare to a very optimized and debloated windows 11?
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)
There's little point benchmarking a debloated windows 11 since:
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
and 3: its also windows 11 on the handheld - its not comparing a desktop (edit- or many desktops for that matter) with steamos on it vs some windows. (though i can see somebody debloating 11 and dropping it on the device - why not?)
> 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.
You can be sure that gamers are going to install SteamOS onto their desktops once it supports more kinds of hardware.
Yes gamers could install Bazzite right now, but those that are open to switching away from Windows aren't going to if they don't have a large company that can fund the support focused primarily on the issues that gamers are going to experience.
Related to 1, Microsoft did announce a gaming-focused "debloated" Windows 11 experience especially for gaming handhelds to release "sometime next year". It will be interesting to see what it is like as a standard, but obviously it is not out yet.
So they are scared of the Steam Deck. Good, they need the competition.
And the Switch 2.
https://www.xbox.com/en-us/handhelds/rog-xbox-ally
2. Many are if this makes their game playable and there is an easy way to debloat
It's usually easier to just install Linux from what I've seen.
Stupid default Apps or an extra Windows Service or few isn't going to make any difference in how the OS performs (or the out of date AMD drivers they used on their Windows install).
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.
There are issues in Windows going beyond just added bloat. I don't think its kernel can compete or keep up with Linux's, as an example.
Microsoft is partnering with Asus to make the ROG Xbox Ally, which will run a stripped down Windows 11 that boots straight into the "Xbox" app, and you can switch to "desktop mode" much like how SteamOS / Steam Deck works. At least on Deck, it only boots the KDE desktop environment up when you switch to desktop mode so you aren't wasting resources on a windowing system you'll never see. It sounds like Microsoft is planning a similar setup, but only time will tell how much they manage to avoid enshittifying the plan.
What's the advantage for a consumer here though? I honestly don't get the selling point of why I would want Win11 specifically on a handheld. The desktop UI sucks with a touchscreen, so having a windows vs a Linux desktop on the hardware doesn't seem like a difference. I get why MS wants to try to compete here, but I just don't get what they could possibly offer. I don't believe Win11 can be stripped down enough to compete on performance, currently I can't get a machine with 4GB of RAM and a SATA SSD to perform adequately with a web browser in Win11. The OS just consumes huge amounts of resources on stupid background tasks that can't be disabled without registry tweaks and the undefined behavior that comes with that.
The person you're responding to wasn't entirely accurate, as it's not just a regular desktop session with the Xbox app set to auto launch. Early reports (admittedly from Microsoft PR) say the gamepad-centric window manager they're shipping on the Xbox ROG, a replacement for explorer.exe, shaves off 2GB of memory usage specifically because they're disabling desktop-centric services during the session. They do seem to be addressing a lot of your points.
The selling point is compatibility with anti-cheat and Game Pass. These aren't targeted for me, I'm not big into competitive shooters, I prefer a la carte and I main Linux, but I can imagine it would be for a lot of people.
Idk I’m not a Microsoft enjoyer. I gave it a good shot with Surface Pro for a bit but it’s just not for me.
i've been daily drive ltsc windows for a while, don't see any software installation problem.
I'm on Windows 10 LTSC which will receive updates until 2032. You'll likely have to add stuff to the OS to install the Windows Store and UWP apps but otherwise regular apps just work.
yeah i did not install windows store, but i recall installing Steam and other software like that is smooth.
On the one hand I hope with the proliferation of such articles and sentiment that Microsoft would start paying more positive attention to Windows as an Operating System instead of an AI and Advertisement Machine.
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.
Microsoft's loss is gamers' gain in this area I think.
Besides, MSFT is almost $500, so they're doing something right.
> Microsoft's loss is gamers' gain in this area I think.
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.
By Microsoft's loss and gamers' gain I mean that Linux used to be nonviable to play games, but Valve put a lot of effort into making games work on Linux (and continues to do so). They saw (and maybe still see?) Microsoft as a threat to Valve's future.
So now gamers have gained another platform to play games!
> but Valve put a lot of effort into making games work on Linux
Yes! this is competition at its finest, right? It is also what I was talking about in my original comment.
> So now gamers have gained another platform to play games!
Yup, this is definitely a good thing. Yes gamers get another platform and hopefully gets MS to focus on Windows to make it better and less hostile to users.
Gamers are also quite vicious and they will tarnish the reputation of anything that doesn't work well for them. Let's see how Linux will fare in their hands.
> 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.
While I agree with the basic sentiment, racism really doesn't deserve a place here, please don't do this.
That's not racism. That's a play on his last name. Just because actual Nutella has a dark color doesn't make my comment racist.
Grow up and stop finding racism in everything.
Quote:
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
This story has been floating around for weeks already. I’ve seen two different videos talking about this already including one claiming Microsoft has already stated they are working on it. It’s not just that the games run faster but that battery life is also better. On a handheld system battery life will change minds more than frame rates.
Correction: it’s been a month and they had both devices:
https://youtube.com/watch?v=CJXp3UYj50Q
And it’s $130 cheaper.
The choice of games is also super strange. Why include borderlands 3? A 2019 cel-shaded release that isn't super demanding seems odd. Homeworld and returnal also seem odd, they're a little more modern but I don't typically see them used as graphical benchmarks. The only game on their list that I typically see for benchmarking is cyberpunk 2077, the rest of the list honestly looks like they did some cherry picking with their selection to massage the data.
> cel-shaded
Borderlands is not cel-shaded. People keep making that mistake for quite a while though.
It's not a terrible description of what Borderlands complex web of shaders are doing. A base layer part of it is effectively a cell-shader, but Borderlands is doing more other things, too.
Relatedly, cell-shading and other shaders are sometimes very heavy GPU compute workloads (which would also be far more expensive if they needed to run on the CPU due to a bad graphics card or driver). It seems funny to dismiss cell-shading as "not super demanding", when cell-shading was a technical dream or expensive technical demo thing to do for quite a while in the time before modern GPU shaders.
I don't think this is because Steam, Proton, or Linux are particularly good on their own (they are, extremely so) but more that windows is just so awful.
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.
I was expecting something like 20-30% of difference not 200-300%.
Microsoft really needs to release a gaming version of Windows without the bloat. I only use Windows to launch Steam these days.
I want games to run optimally too, but I don't think they should release a gaming version of Windows. Ultimately gamers want to do more with their PC's than just game.
They should focus on optimizing the current OS to more optimally run games rather than segmenting the OS.
About 10-15 years ago I played a game on Windows system and was dual booting to Linux.
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.
Better headline: "Lenovo Windows Drivers Bad for Gaming"
Gamepass is the only thing holding me on Windows, and in October, Windows 10 hits EoL. I think that puts a pretty clear end date on the end of my Gamepass subscription as I don't want to upgrade to a new Windows version.
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.
Anecdote: In the days of Quakeworld, for important games, I would reboot into linux for the match. Framerate and ping were better.
As a semi-related aside: I often read of people complaining that Fallout New Vegas crashes constantly for them, but under GNU/Linux using Proton I get none of that, I can play it for hours on end without any notable instability.
Those are likely people that don't check the pcgamingwiki, one of the first things it mentions is installing dxvk and the 4gb patch.
The main pain point I have with SteamOS is game compatibility, in particular with older games (90s/2000s).
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.
Can't say that Windows 11 compatibility on this regard is any way better. Anything pre Dx-9 is somewhat broken one way or another.
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.
https://dreamm.aarongiles.com
I would have to disagree. Those same games that didn't work on SteamOS worked flawlessly on Windows 11.
It's definitely game-dependent. In a few cases I've found it was easier to rebuy a game on GOG, like Dino Crisis or '98 Resident Evil 2 (whose re-releases indeed run on win11 flawlessly while they're quite problematic on Proton) rather than messing with their original cd/dvds.
On the other hand the Sega arcade ports from the same time period (Sega Rally 1&2, Daytona USA Deluxe, Manx TT Superbike Virtua Striker, etc.) are definitely less problematic on Wine/Proton rather than Win11.
I've actually found the opposite in my experience, though I would assume it's very much case-by-case. A recent example was Hyperbowl, a turn of the century game that broke on every version of Windows post-XP. Works fine on Linux with recent Proton and dgvoodoo2, though.
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, I keep a physical vintage winXP machine around for games of that vintage because I find they don't tend to play super nice with modern hardware on windows either. I haven't switched my main personal desktop away from win10 yet due to compatibility with my Cad program, but playing games from that vintage was a nightmare IME. I dunno, maybe I'm unlucky with my selection not playing nice, but I found it way easier to just have a decent 2000's vintage PC hooked up with a KVM to my 3rd 1080p monitor for that. Bonus is that, since I'm playing those games for nostalgia anyway, it's better running them under XP anyway. I haven't gone as far as hooking up a CRT, just due to the space. But the second desktop is tucked away under the desk, so the only real downside is having a low red monitor hooked up. No big loss there, the tertiary monitor is mostly for slack and a media player otherwise so it doe3need to be nicer. Just something to consider, since PCs of that vintage aren't that expensive unless you want a high-end example.
Surprisingly big difference, is there a more detailed overview of the root causes?
I’ve been gaming on Linux for quite a while, and it’s overall been a great experience.
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.
Is this testing the same graphics performance / i.e. screenshot-accurate comparison vs Windows?
Shame you have been downvoted. That's a reasonable question no matter where you stand. Motion is important too, not just static screen-shots.
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.
<3 and, sharing internet +1 coinage to you too.
(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?
But on Windows you’re guaranteed that any game can be run without any extra configuration, which still is a big win. I know that proton handles most cases but it’s still frustrating when corner issue happens.
How about if installed on Microsoft Dev Drive instead of plain NTFS?
You can also achieve nearly the same level of performance by adding exclusions to Windows defender. I typically exclude the game process as I still would want folders scanned when updates / downloads are running. There's a noticable improvement to load times, streaming texture latency, and CPU utilisation.
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.
The file system filter is still queried using exclusions. Eliminating all file system filters yields a faster read/write.
According to some benchmarking removing the storage filters offers a mild boost when compared to adding an exclusion and you don't have the benefit of some AV protection when other processes are accessing the volume like steam etc.
https://blog.maartenballiauw.be/post/2023/11/22/test-driving...
In WSL 1, unix find and git status still run about 4x faster on Dev Drive than on NTFS with defender off.
That's all well and good, but until I can actually play the games I want to on Linux it's really a non-starter to me.
Ok, but what about indie games,m? Does every game just work on SteamOS now? Or is that game dependent?
I did some tests a while ago on w8 vs w11 for bootup time and w8 was faster. This doesnt surprise me.
Those benchmark don't show anything interesting, where is the 2k, 4k, raytracing etc.. show us modern UE5 games, Ubisoft / EA games ect ...
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.
It doesn't make sense to benchmark a handheld gaming device w/ a native screen resolution of 1920x1200 at 2k or 4k. Likewise setting any graphics settings "extreme" like raytracing - it will run at slideshow speeds no matter what operating system. Besides, DOOM: The Dark Ages is a top tier graphics title released a month and a half ago; I think the selection of titles is decent for capability of the device.
This just in, Games on a Gaming OS developed by a Gaming Business perform better than an OS not explicitly for that.
More news at 11.
You recognize, of course, that the conventional wisdom up until today (and even still in most people's minds) was that Windows was the OS to install if you want to play games. There's genuinely a movement right now that's saying, "Linux might actually be the all around better desktop OS if you are a gamer".
That's absolutely noteworthy.
Borderlands 3? Homeworld 3? Who chose these games? Why not just use the current top 10 on Steam atm?
"To test the performance impact of this operating system choice, we started with the SteamOS version of the Legion Go S (provided by Lenovo) and tested five high-end 3D games released in the last five years using built-in benchmarking tools..."
those games come with benchmark tools
Borderlands 3 was 2019. Homeworld 3 has a 38% on Steam and sold poorly. I highly doubt it ever received patches or optimizations. Again, these feel unbelievably arbitrary, as if someone just wanted to push a narrative.
Steam favorability percentages are famously vulnerable to review bombing. HW3 got swept up in culture war nonsense around LGBTQ representation.
I'm a hardcore Homeworld fan. I've run campaigns of their TTRPG, modeled their ships, played the old games to death. I found my own experience with 3 to be "mixed", it's hardly the best entry in the series, but the reviews absolutely are artificially low due to brigading.
Aside, an unoptimized game is actually one I'd want included in my benchmark. Games that have the teams and budgets to really polish will likely perform well no matter what. But how does OS level changes affect those other games, games where the developers didn't put in the care? Does one OS make those games worse? Or does it help with the shortcomings? It's valuable to have entries like that in your dataset.
Homeworld 3 received a terrible rating because its story and delivery was a massive departure to what users wanted, not because of it's gameplay or tech.
It would have been chosen for the same reason ashes of the singularity was chosen as a benchmark for so long: because it looks good, it comes with a benchmark, and it's really good at stressing out a particular part of the computer (for AoS: async rendenring, for HW3: CPU).
Counter Strike 2 (lol, steam says 2012), DOTA 2 (2013), PUBG (2017), Bongo Cat, Elden Ring: Nightreign (2025), Peak (2025), Rainbow Six Siege (2015), Dune Awakening (2025), Marvel Rivals (2024), aaaaaand Wallpaper Engine.
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.
Every simple code base / API is faster than mature API’s due to the fact they do less. A simple string handling library which isnt biderectional, doesnt cater to people with accessibility, doesnt take locale formatting into account, doesnt cover 100% unicode spec will always be faster than complex code that does.
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.