Live data from Hacker News

Linux gaming is faster because Windows APIs are becoming Linux kernel features

xda-developers.com

161–170 of 694 posts

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#161

Show me the numbers. Show me an identical gaming PC running Windows 11 and then Linux, and show not just FPS - but things like frametime pacing, latency, etc. This NTSync stuff is very impressive, but I haven't seen a lot of end-to-end numbers versus Windows. The last comparisons I saw showed pretty much every distribution on the order of 5-30% behind Windows, varying on the game. And Nvidia GPU support was still not…

A lot of the revolution is just getting within 5-30% of Windows! If you need every last bit of FPS maybe it is lagging, but 5-30% slower is roughly on par at a large sense, it's less than the difference of e.g. one NVidia GPU generation to the next, so it makes it playable.

One problem is that having better FPS stops mattering if the frame pacing and timing is bad, making the game feel like a juddery mess. Or if there is significant input delay differences.

That's why all the data matters for all of these dimensions; game performance is much more than FPS per watt over time.

When people see "linux gaming is great now, look at the fps" it comes across as potentially disengenuous because of all the other factors that matter and should be tested. Or rather, if a reviewer is talking entirely about framerate, then I just can't trust their opinion and expertise when it comes to the state of Linux gaming.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#162

Show me the numbers. Show me an identical gaming PC running Windows 11 and then Linux, and show not just FPS - but things like frametime pacing, latency, etc. This NTSync stuff is very impressive, but I haven't seen a lot of end-to-end numbers versus Windows. The last comparisons I saw showed pretty much every distribution on the order of 5-30% behind Windows, varying on the game. And Nvidia GPU support was still not…

I think the actual answer you are looking for is this paragraph:

> These old workarounds got subtle edge cases wrong in ways that produced occasional hitches, deadlocks, or weird behavior in specific games, which are bugs that don't show up on benchmark charts but can absolutely ruin individual experiences. NTSYNC fixes those at the source by matching Windows behavior exactly, and that means as soon as your favorite distro moves to the new kernel version, whether it be Bazzite, CachyOS, Fedora, or a flavor of Ubuntu, they all get this much-needed fix.

That's the crux of the article. NTSYNC isn't faster, it's more "correct". Most games are around the same level of performance, with certain outliers both ways. Right now there isn't anything performance wise that Linux has to do that would impact all games. Just tweaks and additions to the different layers [1][2][3] in the same way driver vendors do. Much of the poor performance is for API violations and other shenanigans.

1: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/uti...

2: https://github.com/doitsujin/dxvk/blob/master/src/util/confi...

3: https://github.com/HansKristian-Work/vkd3d-proton/blob/maste...

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#163

Feels like there is some real momentum on linux gaming now. I mostly play older games but I've gotten most of them working acceptably in proton on my old system 76 laptop (oryp5, with a nvidia 2060; ~7 years old). The laptop actually has plenty of power for the games I play, but I underclock to keep the heat/fan speeds down (been doing the same on the win10 install on the same system), still getting acceptable framer…

I'm looking to finally get off Windows for good. My experience with the SteamDeck started me, later I upgraded to a ROG Ally X for beefier performance but found Windows insufferable on a handheld, and installed SteamOS. I was blown away by the performance gains. A few months later I installed Kubuntu for the first time since 2013 or so, Steam shortly after, and while the desktop linux route is definitely more taxing…

There were specific games keeping me on windows, mostly online PVP. At some point I switched anyway and I don't regret it at all. Now when my friends suggest a game and I'm not able to play it, I just do something else or we choose a different game. There are so many great games out there now, and more release every week. Plus, as I've gotten older, it has become more apparent the fun is in socializing, not the game itself.

My point is, you may find the one or two games holding you back won't be missed much.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#164

Earlier quoted context omitted.

I'm looking to finally get off Windows for good. My experience with the SteamDeck started me, later I upgraded to a ROG Ally X for beefier performance but found Windows insufferable on a handheld, and installed SteamOS. I was blown away by the performance gains. A few months later I installed Kubuntu for the first time since 2013 or so, Steam shortly after, and while the desktop linux route is definitely more taxing…

What do you plan to do about firmware updates?

Believe it or not, it's actually easier to handle on linux than it is on windows now [1]. Normal caveats apply, it depends on your HW manufacturer. However, a lot of them are participating which makes it pretty slick.

And, assuming your are doing x86, you probably already have an EFI partition so even doing motherboard bios updates isn't much of a big deal. You just drop the update in the FAT32 EFI partition, reboot, and point the motherboard at that location. Some motherboards even support just doing that as part of an online update.

https://wiki.archlinux.org/title/Fwupd

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#165
post #25

If you purpose build a Linux gaming PC, would you lean more towards AMD GPUs over Nvidia?

AFAIK none of AMD's offerings match the 5090 for pure gaming performance, so personally that's what I would stick with regardless.

Sure, if you're made of money. For the rest of us, AMD gives you more bang for your buck. Though in this market, it's hard to argue that any of them give you good value.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#166

Feels like there is some real momentum on linux gaming now. I mostly play older games but I've gotten most of them working acceptably in proton on my old system 76 laptop (oryp5, with a nvidia 2060; ~7 years old). The laptop actually has plenty of power for the games I play, but I underclock to keep the heat/fan speeds down (been doing the same on the win10 install on the same system), still getting acceptable framer…

I'm looking to finally get off Windows for good. My experience with the SteamDeck started me, later I upgraded to a ROG Ally X for beefier performance but found Windows insufferable on a handheld, and installed SteamOS. I was blown away by the performance gains. A few months later I installed Kubuntu for the first time since 2013 or so, Steam shortly after, and while the desktop linux route is definitely more taxing…

Modding will continue to be a challenge, but doable, thing, until more mod devs get onboarded to Linux themselves. If the mod devs enjoy using Linux, they'll probably start building mods with UIs native to Linux.

I would say custom modding and online multiplayer anti-cheat systems are the last real hold outs, and even then it doesn't affect every game.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#167
post #158

Show me the numbers. Show me an identical gaming PC running Windows 11 and then Linux, and show not just FPS - but things like frametime pacing, latency, etc. This NTSync stuff is very impressive, but I haven't seen a lot of end-to-end numbers versus Windows. The last comparisons I saw showed pretty much every distribution on the order of 5-30% behind Windows, varying on the game. And Nvidia GPU support was still not…

If you want to swap, then just do it right now? As far as gaming is concerned Linux just works, and reaches speeds that are more than good enough to do so, even if they're not exactly the same as windows - the steam deck is pretty much proof of this. If Linux was measurably 5% slower on all benchmarks, would that mean you wouldn't do it even if you wanted to? Is every single nanosecond of performance really that impo…

I gave it a try. Got a steam deck, tries steam os on my desktop.

I kept running into issues that took me time to solve. I understand that the only reason it took me time to solve these issues is because I'm new to it and that people who have been gaming on Linux for years already know how to solve them all. But what would happen was is I would sit down to play a game spend maybe an hour or two fixing issues and then after that I ran out of time to play the game. I kept this up for a couple months but honestly at some point I just gave up. Now I'm playing games on Windows again.

To be clear, I'm a huge proponent of Linux gaming. I just unfortunately am too busy these days to spend the time to get it to work.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#168
post #158

Earlier quoted context omitted.

If you want to swap, then just do it right now? As far as gaming is concerned Linux just works, and reaches speeds that are more than good enough to do so, even if they're not exactly the same as windows - the steam deck is pretty much proof of this. If Linux was measurably 5% slower on all benchmarks, would that mean you wouldn't do it even if you wanted to? Is every single nanosecond of performance really that impo…

I gave it a try. Got a steam deck, tries steam os on my desktop. I kept running into issues that took me time to solve. I understand that the only reason it took me time to solve these issues is because I'm new to it and that people who have been gaming on Linux for years already know how to solve them all. But what would happen was is I would sit down to play a game spend maybe an hour or two fixing issues and then…

And I'll try again when I have more time.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#169
post #158

Show me the numbers. Show me an identical gaming PC running Windows 11 and then Linux, and show not just FPS - but things like frametime pacing, latency, etc. This NTSync stuff is very impressive, but I haven't seen a lot of end-to-end numbers versus Windows. The last comparisons I saw showed pretty much every distribution on the order of 5-30% behind Windows, varying on the game. And Nvidia GPU support was still not…

If you want to swap, then just do it right now? As far as gaming is concerned Linux just works, and reaches speeds that are more than good enough to do so, even if they're not exactly the same as windows - the steam deck is pretty much proof of this. If Linux was measurably 5% slower on all benchmarks, would that mean you wouldn't do it even if you wanted to? Is every single nanosecond of performance really that impo…

5% would already be well within the margin of difference for separate identical clean installations of windows on the same hardware.

But the issue is that it is many multiples of that, especially on the most common PC gaming hardware (Nvidia GPUs), often more than a 25% difference in framerates. Not so important at 144fps, but very important at a 60fps baseline and for genres like fighting games.

A lot of people don't mind, say, an extra 5 frames of input delay. They don't notice it. But a lot of people do notice even an extra 2 or 3.

I do think that frame pacing issues kinda do have a critical thin threshold where it's either bearable or an unacceptable difference. And the native windows version can often already be riding right on that line. So while it's not fair to the Linux version to demand better, it is unfortunately the case that it might tip over that line.

Re: Linux gaming is faster because Windows APIs are becoming Linux kernel features

#170
post #94

Earlier quoted context omitted.

if I implemented a couple of traps What does this mean? System calls?

Similar but traps are triggered automatically on attempts to execute a protected instruction. https://www.geeksforgeeks.org/operating-systems/traps-and-sy...

Why/when are traps used rather than explicit system calls? Is it just historical coevolution? Or is the idea that the user mode program doesn’t need to know that it’s unprivileged? Or is it just repurposing the error handler path to perform privileged operations?
Post reply on HN