Good choice by testing CS2. It's one of the better titles out there when it comes to latency optimizations. Thank you for this one! LE: Please, when you do find the time and resources do more of these, Michael. If latency on Linux is the same as on Windows or better, competitive gamers should definitely switch. Although I doubt Linux will ever beat Windows as the drivers problem is still a thing in 2024. I would love for you to make a Windows vs Linux test on latency. So guys, let's get Michael to 100k subs, shall we?
Optimum isn't going to touch Linux like gamers nexus wouldn't. They won't throw away perfomance for an OS. An OS is tool to most people, not a religion.
If I truly wanted to minimize gaming latency, I would probably start by using a lighter weight desktop like Xfce or even Mate. Or even more extreme--just go raw window manager like Openbox, Blackbox, i3, etc. Hell, I recall even running Doom 3 directly as a log in session--no window manager at all, just Doom 3 in full screen--many, many years ago. Back then I was stuck with a Pentium 4 PC that just didn't have the memory (only 256 MB) to properly run it and I was stuck on a system with PC-800 RDRAM which made upgrading so expensive it would not even be worth it. It was a pretty fun little experiment, and I have never seen the game run as well on that machine with any other configuration, and certainly not with WinXP or the OS it originally came with, WinME. It was pretty cool changing my session to DOOM 3, after entering my username and password the game would immediately start up, and when quitting the game I would immediately be dumped back to the display manager where I could pick my usual desktop environment or window manager at the time.
If it's not a secret, could you tell us how you managed to run Doom 3 that way? I have an old laptop that I upgraded with SSD and Linux, but any old game I try installing and playing runs just awful. I guess IntelHD without vulkan sucks 🥲
If you're looking to achieve that Doom3 trick in the current year, what you want is Gamescope in its embedded mode. Xfce, Mate, Openbox, Blackbox and i3 currently have Wayland support ranging from "not very good" to "not at all", and Wayland with its latest extensions is probably *better* in terms of latency than X11, so they are not an option anyone should consider.
@@brnddi The laptop in question has 4GB of ram and Intel Celeron 1000m, so even xfce is fine, but performance on any game I try is just horrific. I tried everything I could to increase it but it just won't 🥲
It's very nice to see latency testing done on Linux. I find couple of additional frames of input latency very bothersome even in office desktop work and have history of 90's twitch shooters. Though I have to admit that my reaction time has worn down over the years I've still been rather happy with X11 so far. Also the frame timing is more forgiving on higher refresh rates than 60Hz as after moving to 144Hz VRR monitors (VRR works fine in X11 Gnome with Nvidia) I really haven't felt bothering input latency on any game I've played. While a LED connected directly to a mouse switch activation would be nice if RGB capable mice could have a mode where the mouse flashes leds when any of its button are pressed. (this would be also helpful for determining switch chattering) Alternatively a USB extension cord with an indicator LED pushed buttons would be very helpful as well.
Fascinating and Thank You! Now you see what I hate to part with my Fatal1ty motherboard with a dedicated mouse port. Shrieking in the background, that system runs Windows 7 and and my legacy games like Unreal Tournament and the original StarCraft. I'm sure people will give me flack for running that old system, but at 70+ I get to play the games I like, and watch younger and smarter people like you to learn what I can... Bless You!
Good video. I play a lot of rhythm games where latency matters a lot, and I'm at the point where I can feel the difference of about 5-10ms, so thanks for confirming my priors lol I would say even for most competitive games, the difference is too small but when timing windows come down to less than 100ms (and my deviation being as little as 55-70ms), it's world's apart.
That's one interesting video I just stumbled upon! Reminds me of my weird experience, that (at least my main game) Hunt: Showdown runs noticebly smoother on Gnome Wayland than on X. Locked 100 fps on 144 Hz display.
I have personally found that basically all desktop environments have noticeably more input latency than window managers. Even if you don't like using them, while gaming you only use a single application in fullscreen, so imo it is worth it to learn the basics of a window manager to do that. Desktop environments and games just don't play nicely
I'm not sure if it's actually perceptable or just placebo, but technically you are a 100% correct as Desktop Environments have a lot more going on. Way more services, compositing and so on
For best performance, lowest latency possible and no stutter you should use an x11 window manager with no compositor (and vsync off). I compared dwm with no compositor to hyprland (on amd) and it ran faster on x11 and didn't have any stutter unlike hyprland. X11 with no compositor gives you even better result than windows.
Can you use x11 on nobara 39 Wayland or is that not worth the hassle to go through? I just started playing cs on Linux and it immediately felt off like everything was off when counterstrafing.
almost every cell phone (if not all of them) interpolates it's video when above 240 fps (turns out my A53 can't actually shoot about 120fps lol) you can test this by opening the "Pro" option in your camera app (according to how i set my Samsung up, it's in "More", which you can get to by swiping to the right, then selecting "Pro"), then you'll want to select "Speed" and slide it to the closest value to 1/480 (mine's 1/500), if you're not outdoors, this will make your image quite a bit darker, and a telltale sign that your Camera app isn't recording video at that frame rate, is if it looks brighter than what you're currently seeing.
I guess we're going past measuring 'input latency' only in 'ms', apparently there's something else to be found here, given that when you can feel it, it likely isn't a "placebo". Minor - yes, non-existent - probably not, at least not in every instance. Hence in some cases, on some days you may not feel any difference or even feel the opposite way about it (f.e. if you felt was better, now feel it might be negligible or worse). But my point is that the question isn't about the few 'single digit ms delay/difference' of 'total input lag', but rather more like how things are with stuff like 'pixel response' and 'ghosting'/'motion blur' etc. There you're talking about a 'few single digit ms'difference, often 'sub 1ms'. No one in their right mind would claim that a human can perceive such a small variation - 1-5 ms difference, alone 0.1 (tho most of these numbers are purely for marketing) any actual numbers wouldn't be much far either, but yet one can see and notice a perceivable and MEASURABLE difference in these cases. Anyone who have researched the topic on 'Motion Clarity' would know. So, I'm confident that there is a similar 'metric' to be found here. Only issue is that we are dealing with "double variables", as the human perception can vary, tho it is not to say that a monitor is an "absolute variable" either, as depending on various factor/conditions - stuff like 'Pixel Response Times' can vary quite a bit, even on single factor like 'Temperature'. Generally, Lower Temps = Slower Times. This debate brings memories of the old "You can't see more than 60Hz" misconception. We now know for certain that there is a noticeable AND great difference between 60 and higher Refresh Rates, regardless whether and what actual "Hz" the human eye can "see".
Great points. Its indeed not a black and white story, especially picture clarity which my monitor is certainly not the best if we compare it against those, double or even triple the price. Those things can have an impact on your performance, and a quite noticable one as well, however it becomes almost impossible to proper sample. Another debate that I like is DPI and pixel skipping. While this doesn't necessarily have an impact to your aiming ability in reality, I still wonder if pixel skips could be an advantage since you see one or two frames longer before the next skip. Latency it shouldn't be noticable, but since we can still "see" a difference between 240 and 144 Hz, it wouldn't surprise me
@@MichaelNROH I guess I haven't dug that deep into the Mouse DPI thing to notice the 'pixel skipping', but I can say for sure that different DPI is a different experience regardless of any other settings. 400 will 'perform/behave' differently to 800, 1200, 1600, etc., despite you can measure the same 'distance per pixels' (the specific distance you have to move your mouse from 1 point to another). Similarly, here one can measure identical ms or at most negligible difference in total input lag, tho it will still feel and seemingly behave differently. If one takes other aspects like frametime f.e., you can have identical FPS and input delay, but you'll definitely notice and be affected if your frametime is all over the place. I don't know if and to what extend one can rely on frametime here, as this likely comes down to the specifics of the rendering and frame generation of the OS, as you mentioned in the video, but I think it can likely provide useful information and give a much better idea of what's going. It's much more "sensitive" metric (so to say) compared to 'total input lag', as you can definitely notice even minor inconsistencies. Likely what was going on with the "Gnome stutters". Generally this is the main issue with Linux gaming currently, maybe save for Arch/SteamOS, all other distros seems to struggle with proving a stable frametime, even when they manage to outperform Windows in max/avg FPS.
I think, next time you want to mesure something from a mouse click, use either and equivalent to xdotool, for wayland (sending a mouse click from cli, so you can display it somewhere, film the screen and the output at same time, or evemu (records and replays device events), or if you have some arduinos lying around, it's super easy to make it send a mouse click and light a led at the same time (so, you ask your aduino to make a loop sending a click, lighting the led, keeping it for a few time, wait a second, do it again, for 10 times), otherwise, great video !
That is a good idea if we want to take compositors into the mix as well, but that wasn't my goal. I wanted to capture fullscreen applications whereas the compositor doesn't have to do much or anything at all, while also exerting CPU and GPU Load. But yeah, other testing methods could be done for other testing purposes and might show different variances
@@MichaelNROH That doesn’t really make sense because the whole reason I’ve seen people say Wayland can have more latency than gnome is due to not being able to disable the compositor on Wayland for full screen applications, which you can do on gnome. This was also a problem on windows in non full screen applications until they implemented flip mode into full screen optimizations. So part of the latency difference you might have felt could easily have been from the compositor especially since your test already had a ton of variability in it.
Yeah, I didn't consider anything else besides Wayland for this vid as its where the future is headed but yeah, other DEs that are still on X11 might have an advantage here
@@Robotta For me on amd side with arch, it always feels like being at least some frames behind in CS2 when using any wayland session (it feels like delayed multiple frames). I could NOT get CS2 to use VRR (fps_max lower than 144 on 144Hz display) on any wayland compositor (gnome, sway, Hyprland, plasma) , no matter what, other games that sync properly are completely fine and dont result in this delay for me.. Best input feel is still xorg with gnome and slight visual tearing for CS2. Got experimental tearing to work with Hyperland , but this was visually tearing but also had the same feel of multiple frames lagging behind feeling (200-300fps), retesting this every couple of weeks , want to ditch the xorg session.....
These are fantastic videos. What kills me is all the stutter in CS2. I have to work hard to iron that stuff out. Usually disabling multithreading in the game, which is silly.
Yeah, CS2 is terrible currently. Matchmaking (5v5) works fine luckily, but Deatmatch, and now Arms race drop around 200 - 300fps from my normal average. I found out that performance decays start, when enough people have joined the server. Probably some hefty networking tasks that pull performance down
Hey Michael! It's a great pleasure to meet you. I'm new here on your channel and I'm loving it. I haven't found a specification yet but which Linux distribution are you using? Thank you very much and a big hug!
There is no way it's a placebo. All my games feel like they have less input lag on gnome 46 Wayland compared to KDE Plasma 6. With 240hz and no VRR. And some games have more stable fps like overwatch 2. On gnome it stays at fps lock. On KDE it jumps from 236 to 200 very often. In World of Warcraft on KDE it feels like I'm moving my camera through water. On gnome is fine, no issues. I don't care for VRR because I use 2 monitors anyway. And I would only use it for single player games.
this is unfortunately because of valve. they use a library that uses the cpu to do a lot of calculations so you need a much more powerful cpu on linux when using vr than on windows. On windows they use the same library but run it with opencl, so it runs fast with the gpu. That library actually supports opencl on linux too but its just that valve hasn't enabled that. So its a simple mistake that unfortunately cant easily be fixed since the steamvr driver is not open source. Thats why proprietary software sucks
@@notuxnobux I wouldn't bother about paired performance...right now VR is not running after the new update of steamvr. I left Fedora to use Windows...used to run flawless with Arch.
@@luiz8755 yeah thats also unfortunately a steam vr driver issue. Other headsets dont have this issue when using openxr for example. Its so bad that if any application closes without telling steamvr to shutdown (for example if the application crashes) then steamvr crashes too.
I setup a VM with gpu passthrough running windows 10 for this reason. Somehow it runs even better than a dual boot and my input has never been more smooth.
VMs generally have more overhead, so I wouldn't trust in it, but it shouldn't be noticeable anyway as long as the output is hooked directly to your monitor
@@edwardecl You mean like run it without a Desktop Environment? I don't seem how that's practical, but the latency would probably be a bit less just because there is less background processing
@@MichaelNROH Gamescope is a Wayland window manager when run from the console, it supports VRR. Would be interesting to see if the compositor adds to latency or not or if there's nothing to call between any of them.
Because these settings only affect KDE Plasma and non-fullscreen apps. This is a setting that most don't seem to know what it does and that's why its removed in Plasma 6
@@MichaelNROH the real reason why it was removed from plasma 6 is because they found a solution that has the least amount of latency possible and also the smoothest, so the option is no longer needed.
It's easy, you just turn off V-Sync, look at that horrible screen tear. You open RTSS and set -1 woala, fluid image, without input lag and fixed to the monitor's fps.
Yeah, the Plasma 6 implementation of it is way better, but its still not Gnome polished since you still can't properly launch apps from it. You can search for them, but if you already have an instance running (e.g. a browser) then it just finds the window. I wish the could implement the taskbar / panel in it
i dont mind with gnome. really. my gripe beside the mandatory wayland thing(which still slower compare to x11) is the LACKS of customizations to its theme. the most lacking than every other desktop offerings in linux environment. its like if gnome shipped in one of the distros, those distros would implement an annoying "brand color" theming which may not be in favor to the users. and there is no options to change those out of the box. this is in every single distros coming with gnome. but i understand why. just look at their logo. the foot missing a toe. so its only natural that you get that feeling when you enter a gnome environment. something is missing. by the way, fellow fedora user here too. fedora has big potential. it could be the "it" from linux to compete head on with windows.
I have never used any since I always adjust these things in UEFI, but otherwise probably fancontrol (or fancontrol-gui), depending on how you want to do it. I have also seen Coolero in my Software Center, but I don't know if its just monitoring or also configuring.
@@minecraftpufferfish9066 I use Fan2Go. its a hassle to setup once. But the github of the guy is pretty straight forward. And the best thing is that you can setup combined fan curves. My botom fans react to my CPU and GPU independtly. So if GPU askes more. They spin up. And if CPU only asks more cooling they also spin up. Same setup with my exaust. And the best part. Its extremly low on resources. I know this post is late. But I just love this one.
My own personal hardware - brain, is too slow anyway, so such latency differences don't play any role for me. I'm gaming mostly for the story (RPG games) and I always hated the precision, speed or dexterity games. Although, achievement is also a game hook-up, so sometimes I do try such games, but if they are too challenging, they are not for me. If they can forgive clumsiness, there is a chance I may enjoy it. However, playing competitively? No interest at all.
@@MichaelNROH Because story matters most, I was able to play Baldur's Gate III with 10-21 FPS, on the lowest settings possible. It was still fun, and I was able to finish the game. Still, it's nice to have better performance, although for some games it's not crucial, because graphics or performance are not necessary for playing it. Anyway, I bought a new Tuxedo AMD gaming laptop, and once it arrives, I will be set for another 8-10 years (if the hardware will hold). With such needs, I don't have to upgrade the hardware too often. Frankly, I wouldn't change much, but the battery in my current, old Alienware laptop, is declining, and I already exchanged it twice. I'm not sure if it's now possible to buy another one, so I decided it's finally time for a new hardware. Gaming capabilities improvement is only a bonus.
No, and that's why I showed it. The thing is, that this settings is meant for just KDE Plasma and Windows that are not running in Fullscreen. Fullscreen applications get a special treatment and does not need any compositing, so it doesn't matter for gaming at all. It's even removed in Plasma 6, as I can recall.
@@MichaelNROH That's great, AFAIK Gnome disables composting on fullscreen apps by default, and plasma users had to do this manually with an option to disable it. Glad that they also picked this up.
For some reason when I was testing 4090 out under Plasma WL recently, I couldn't get VRR to work at all. Maybe it was the games I was testing, but oddly enough I couldn't get VRR working under X11, and I even disabled my other monitors to test this. I think there could be a bug or three in 545 drivers that fails to use VRR with certain monitor configurations or something.
I think NVIDIA still doesn't support VRR on Wayland that's why it doesn't work. Could be that I missed a release but it should be inaccessible at the moment
I haven't gotten into changing to use wayland because many issues are still present, but I am using x11 kde plasma 6. I was curious if you used kde plasma 6 or not.
I use GNOME because when i installed KDE Plasma on my system and played minecraft i couldnt reach 30 fps with a good laptop but when i reinstalled GNOME i was easily getting over a 150 fps
That sounds like something has gone during the installation, since the Desktop Environment doesn't have much of an impact on gaming performance. In fact, Plasma even has a slight edge over Gnome because it tends to be optimized more quickly and is partially funded by Valve
Which Graphics card did you use and did you compare it to another? even lets say an older nvidia card vs a newer one (when i can believe comments from some other users the newer cards ar way better supported than the (now) old ones... I Think - which graphics card you use can make a difference. AMD vs Nvidia and potential Bugs.. From own experience with a Nvidia 1080 it couldnt even move a normal windows lag free aroud on the Desktop - thats obviously a bug... but still...
I used an AMD RX 6800XT, but I own a GTX 1080 as well. The thing that I noticed when I was on NVIDIA well over a year ago was, that it doesn't run as nicely with DX12 games in comparison to DX11 (besides the regular performance difference). Nvidia doesn't seem to bother supporting older cards on Linux, as they do on Windows which unfortunately comes with a performance penalty. But it's not that bad as I could still play some games like Far Cry 6 with like 85 to 90% of its performance on Windows.
Seeing my 5 year old video pop up at th-cam.com/video/JJZeByBEoZY/w-d-xo.html was quite a surprised haha! Thanks for the video, currently trying to switch my gaming + streaming setup to linux and this helped a lot.
@@MichaelNROH Actually maybe I'm confused you little bit. I was talking about dropdown placed in Display -> Compositor tab, not the animation speed on Effects tab When I set it to Balanced mode - it worked with "nominal" perfomance (about 120-130 FPS on my setup). When I set "Lowest delay" (there is warning about "tearing will be possible") - it drops my FPS то 80-100 and made game process very chunky and rough... like some frames are dropped... But actually this option must made gaming smoother... So best compositor mode is Balanced, any other ruins gaming perfomance for me. I think you need to check that aspects and show the difference.
Besides the lack of some software solutions like NVIDIA Reflex, the latency is practically the same. Comparing the latency between Linux and Windows is not as easy as comparing between Desktop Environments, since it depends a lot on your hardware (e.g. Proprietary AMD driver on Windows vs Open Source Driver on Linux) and the games you play. Every test result on each system could be different, leading to non-reproducable results. Practically speaking, there is no difference whatsoever, and if there was a latency difference then it would be way below 50ms, which is the time that most can notice that something is off when constantly dragging the mouse.
@@MichaelNROH Nvidia actually has sent in some pending patches to dxvk/vkd3d to enable reflex in proton, although latencyflex is already pretty much the same thing.
@@PeakKissShot I'm glad that some effort comes from them, as LatencyFlex often gets flagged by Anti-Cheats. Having some option until this is patched would be nice
question: i am running Ubuntu on my pc and i sometimes have an issue where it will completely freeze, mouse and keyboard dont work, even the disk activity light stops blinking. it seems to happen when i close a game that was running from steam and i send a discord message right after it closes. then all web browser related apps (chrome, steam, discord) all say not responding, then everything freezes a few seconds later. any idea what is causing this?
Mh, that's a tough one. Usually freezes occur when a GPU driver or the compositor crashes, but that shouldn't take down everything. It could be a bug in the compositor itself since for fullscreen apps there is no compositing going on. Maybe closing the application triggers some issue. You could try reading through the latest syslog files in /var/log/, or open a bug report with the log attached
@@MichaelNROH Hmm. this morning, i turned on my pc and there was a red minus symbol in the task bar (Gnome extension) and it said "There is installed software with missing dependencies." this has never happened before. i havent installed anything new for at least a month.
@@wibwabz If you generally haven't installed anything then your drive might be broken and keeps breaking bit by bit. You should take a look at that with some online tutorials
I think that many use gamescope wrong, as running it on top your Desktop Environment can even introduce more overhead if you don't run it fullscreen. It's basically Steam Game -> Gamescope -> Mutter, Kwin, etc. -> Output What you want is to eliminate the Desktop Environments for Gamescope to properly take control of your screen. Otherwise, Gamescope just provides better scaling (e.g. Global FSR), Limiting Framerates, custom resolutions, a sandbox that doesn't interact with other things on your system and so forth.
3 หลายเดือนก่อน
Can you update with commentary for Gnome 46 enabling VRR.
From time to time there are reports, but it's rare in browsers. The much bigger problem is games, though VRR or not, Gnome is refining the way how they synchronize (or check) timings of the GPU as well for no more drops. Btw. I guess we are talking about dropped frames to the monitor, not actually lower fps in a game. This would be its own issue
@@MichaelNROH I reread my original question and realized I completely botched the explanation. I'll fix it below. For my 60fps monitor (no VRR) with AMD Ryzen 6000 - Proton games under Gnome Wayland run at 30fps, no matter what I do. I have enabled the Steam overlay to show this information. My monitor is configured at 60 hz, and the game settings are at the lowest they go. Doesn't matter if fractional scaling is on or off. These games look like they're running at 30fps too. - Proton games under KDE Plasma Wayland run at 60fps easily with the same setup. These games are smooth. Now for any browser that I use: - Under Gnome Wayland, I pull up any 60fps video on TH-cam, and about half the frames are reported to have been dropped. Oddly enough, it looks like 60fps content, not the 30fps I get with Proton games. - Under KDE Plasma Wayland, browsers don't report frames as being dropped, and look the same. My solution thus far has been to use KDE Plasma / Windows. I really hate running X11, but that does make all issues go away in Gnome.
@@MichaelNROH Apologies. I botched my comment and it's not accurate after re-reading. I've corrected it below: Under Gnome Wayland with a 60 Hz monitor, all Proton games run at a steady 30fps. I can tell just by how choppy the games are, which is confirmed by the Steam overlay. All other apps appear to run at 60 fps because of how smooth they are comparitively. The other odd thing is that Firefox and Chromium browsers under Gnome Wayland, when playing TH-cam videos that have 60 fps content, reportedly drop ~ half their frames. But it looks like 60 fps content: it's not choppy like the games. And it doesn't matter if I run those browsers under XWayland or Wayland. Under KDE Plasma Wayland, I don't have these problems, but I have other issues: Any kind of popup (like the volume control bar) causes the Proton game graphics to freeze so that I have to force quit and restart. And Chromium browsers (IIRC, Firefox too) freeze when full screen if they're running in their Wayland modes. Under Gnome X11, all these problems go away. I have a hybrid graphics setup (it's a laptop). Both are AMD chips, and the dedicated Ryzen card is a 6000 series (I don't recall off the top of my head). No matter which distro I run, I consistently get the same results as the above. It's odd, because AMD should have the best support. I have no idea, and I can't find my problems reported anywhere else. Just curious as to if you've run into this before?
@@MichaelNROHalso how do you know that cs2 is running in full screen? Most newer games have a “fullscreen” option that is really just fullscreen windowed but forcing mouse to stay on screen until alt tabbed. AFAIK that would mean Linux would still see it as a windowed application
@@MichaelNROHI just tested cs2 while changing the Wayland settings and they do in fact 100% affect cs2 in fullscreen due to the application itself not ever being in true fullscreen. If you set to lowest latency possible the game will get extremely stuttery like it’s dropping frames and when you go just one above the lowest latency, the latency is definitely there but it’s smooth and no issues with dropping frames
Slightly off topic: I tested the keyboard latency in Windows and in Linux, in Linux I had constant 250ms of key repeat rate (because I set it as such with xset), but on Windows I had from 220 to 280ms, Windows scheduler/IO suck so much.
As of right now yes, but not because Wayland is slower, but because almost every game is being run with XWayland and not every DE supports Tearing and VRR yet. Once that's all in Place, Wayland should be faster since its more "cleaned up" protocol
@@MichaelNROH wayland wont be faster. Both x11 and wayland work pretty much exactly the same way if you run a fullscreen game (and enable tearing in wayland). They even use the exact same kernel (drm) and mesa (gbm) functions.
@@notuxnobux Probably not measurable, but by how the protocol itself is build. Less overhead means more resources and better latency though this is probably even below a ms.
@@MichaelNROH there isn't less overhead with wayland. As I said, wayland and x11 works pretty much exactly the same way when it comes to fullscreen gaming. When you have a fullscreen application that uses opengl/vulkan (and no vsync) it bypasses the display server. A lot of people seem to misunderstand x11 and think it has overhead for gaming "because it's old and bloated", which isn't true. That "bloat" in X11 is only touched when you dont use opengl/vulkan.
Variable Refresh Rate technologies only really comes into play if the fps are lower than your monitors refresh rate and it depends on your setup. It can theoretically improve latency, but that depends on how good your monitor is. The lower tiers (there are several) can slightly add latency since the synchronisation takes some processing speed. But this has a very small impact. Some TH-camrs measured this
tearing is not supported by your kernel yet (unless ur rocking 6.8 for whatever reason) and therefore the tearing option won't make any difference right now
How did you come up with that? Tearing is working as a Wayland protocol and doesn't have anything to do with the kernel at all. KDE Plasma and Gamescope have it successfully implemented
@@MichaelNROH The tearing protocol also required changes in mesa, the kernel, and I believe setting an environmental variable. So unless all of those were present, KDE didn't actually have tearing enabled.
@@that_leaflet Oh, you mean "KWIN_DRM_NO_AMS" on Plasma right? That is set for this video, however there is still a difference how Kwin handles it. From playing without the setting, KWIN seems to behave similar to Gnome with more stutters though it's most likely a combination of all the patches in place (Kwin, Mesa, XWayland, etc.)
To have better latency you need to have tearing at least but it have a lot of issues under wayland even on kde. Gnome is completely unusable for me from this point of view. On kde it work stable under the x11 session only. Under wayland it sometimes works, something doesn't.
Long story short. If you're into gaming, get a monitor that supports VRR and play in KDE while limiting the FPS slightly under the maximum refresh rate of your monitor.
I wouldn't recommend this typically. If you want a smooth experience then yes, but for competitive gameplay, you want higher frames and utilize tearing for a faster response time.
Thats why I just left Linux gaming. Linux is great (for now) for programming, browsing web, listening to music all things what Windows and Mac does minus gaming. Proton, DXVK, VKD3D are great thing, and they'll fight Microsoft monopoly in gaming, but for now when Linux NVIDIA drivers are terrible and many games are not launching after a freaking driver update, I'll stay on my Windows VM with NVIDIA GPU for gaming and AMD GPU on Linux for displaying desktop and that's it. It's my setup. Arch + KDE Plasma (wayland) and KVM with GPU passthrough for gaming
Is it me or the tone of the guy talking is super annoying (fake)? I couldn't finished the video even though the topic really interest me. I imagine an old news anchor guy talking with that tone, fake af.
9 หลายเดือนก่อน +5
it's not just you. His videos looks interesting, but every video is a pain to watch because of the way how he is talking, how every few words his voice pitch goes very high, it's like on a rollercoaster.
Garbage. If you're wanting to game or get the most out of your office hardware, then use Windows. If you're wanting to just do general tasks, surf the Internet, watch movies, write texts, and don't care about whether your latest hardware and all its features will be fully supported by an operating system, then use Linux.
bah que tonterias, yo juego en linux perfectamente, al menos los juegos que me gustan funcionan bien. Genshin, honkai star rail, HZD, spiderman remastered, god of war ragnarok, doom eternal, star wars jedi survivor, callisto protocol.
Kernel 6.8 comes with an I/O optimization of ~6%
For Gaming or throughput? I haven't looked into it yet
I've been waiting for someone to look into this, very interesting!
Great video, MASTER!!
That was interesting tests! I have never thought that the DE could affect gaming experience in any way!
Thanks for the video!
Good choice by testing CS2. It's one of the better titles out there when it comes to latency optimizations. Thank you for this one!
LE: Please, when you do find the time and resources do more of these, Michael. If latency on Linux is the same as on Windows or better, competitive gamers should definitely switch. Although I doubt Linux will ever beat Windows as the drivers problem is still a thing in 2024. I would love for you to make a Windows vs Linux test on latency. So guys, let's get Michael to 100k subs, shall we?
Should make a collab with optimum and test Windows vs Linux
Optimum isn't going to touch Linux like gamers nexus wouldn't. They won't throw away perfomance for an OS. An OS is tool to most people, not a religion.
If I truly wanted to minimize gaming latency, I would probably start by using a lighter weight desktop like Xfce or even Mate. Or even more extreme--just go raw window manager like Openbox, Blackbox, i3, etc.
Hell, I recall even running Doom 3 directly as a log in session--no window manager at all, just Doom 3 in full screen--many, many years ago. Back then I was stuck with a Pentium 4 PC that just didn't have the memory (only 256 MB) to properly run it and I was stuck on a system with PC-800 RDRAM which made upgrading so expensive it would not even be worth it.
It was a pretty fun little experiment, and I have never seen the game run as well on that machine with any other configuration, and certainly not with WinXP or the OS it originally came with, WinME.
It was pretty cool changing my session to DOOM 3, after entering my username and password the game would immediately start up, and when quitting the game I would immediately be dumped back to the display manager where I could pick my usual desktop environment or window manager at the time.
If it's not a secret, could you tell us how you managed to run Doom 3 that way? I have an old laptop that I upgraded with SSD and Linux, but any old game I try installing and playing runs just awful. I guess IntelHD without vulkan sucks 🥲
If you're looking to achieve that Doom3 trick in the current year, what you want is Gamescope in its embedded mode. Xfce, Mate, Openbox, Blackbox and i3 currently have Wayland support ranging from "not very good" to "not at all", and Wayland with its latest extensions is probably *better* in terms of latency than X11, so they are not an option anyone should consider.
You can create users for every game))
@@brnddi The laptop in question has 4GB of ram and Intel Celeron 1000m, so even xfce is fine, but performance on any game I try is just horrific. I tried everything I could to increase it but it just won't 🥲
@@KajzerD if your hardware gets no Vulkan support it's a waste of time to even try to run any game that doesn't support OpenGL. It just won't work.
Awesome video Michael as always!!
I appreciate it
It's very nice to see latency testing done on Linux. I find couple of additional frames of input latency very bothersome even in office desktop work and have history of 90's twitch shooters. Though I have to admit that my reaction time has worn down over the years I've still been rather happy with X11 so far. Also the frame timing is more forgiving on higher refresh rates than 60Hz as after moving to 144Hz VRR monitors (VRR works fine in X11 Gnome with Nvidia) I really haven't felt bothering input latency on any game I've played.
While a LED connected directly to a mouse switch activation would be nice if RGB capable mice could have a mode where the mouse flashes leds when any of its button are pressed. (this would be also helpful for determining switch chattering) Alternatively a USB extension cord with an indicator LED pushed buttons would be very helpful as well.
Fascinating and Thank You! Now you see what I hate to part with my Fatal1ty motherboard with a dedicated mouse port. Shrieking in the background, that system runs Windows 7 and and my legacy games like Unreal Tournament and the original StarCraft. I'm sure people will give me flack for running that old system, but at 70+ I get to play the games I like, and watch younger and smarter people like you to learn what I can... Bless You!
Good video. I play a lot of rhythm games where latency matters a lot, and I'm at the point where I can feel the difference of about 5-10ms, so thanks for confirming my priors lol I would say even for most competitive games, the difference is too small but when timing windows come down to less than 100ms (and my deviation being as little as 55-70ms), it's world's apart.
That's one interesting video I just stumbled upon!
Reminds me of my weird experience, that (at least my main game) Hunt: Showdown runs noticebly smoother on Gnome Wayland than on X. Locked 100 fps on 144 Hz display.
I have personally found that basically all desktop environments have noticeably more input latency than window managers. Even if you don't like using them, while gaming you only use a single application in fullscreen, so imo it is worth it to learn the basics of a window manager to do that. Desktop environments and games just don't play nicely
I'm not sure if it's actually perceptable or just placebo, but technically you are a 100% correct as Desktop Environments have a lot more going on. Way more services, compositing and so on
Great video bro...
For best performance, lowest latency possible and no stutter you should use an x11 window manager with no compositor (and vsync off). I compared dwm with no compositor to hyprland (on amd) and it ran faster on x11 and didn't have any stutter unlike hyprland. X11 with no compositor gives you even better result than windows.
Yep, X11 is still the way to go.
Can you use x11 on nobara 39 Wayland or is that not worth the hassle to go through? I just started playing cs on Linux and it immediately felt off like everything was off when counterstrafing.
Nice review!
Good video!
Thanks
almost every cell phone (if not all of them) interpolates it's video when above 240 fps (turns out my A53 can't actually shoot about 120fps lol) you can test this by opening the "Pro" option in your camera app (according to how i set my Samsung up, it's in "More", which you can get to by swiping to the right, then selecting "Pro"), then you'll want to select "Speed" and slide it to the closest value to 1/480 (mine's 1/500), if you're not outdoors, this will make your image quite a bit darker, and a telltale sign that your Camera app isn't recording video at that frame rate, is if it looks brighter than what you're currently seeing.
I guess we're going past measuring 'input latency' only in 'ms', apparently there's something else to be found here, given that when you can feel it, it likely isn't a "placebo".
Minor - yes, non-existent - probably not, at least not in every instance.
Hence in some cases, on some days you may not feel any difference or even feel the opposite way about it (f.e. if you felt was better, now feel it might be negligible or worse).
But my point is that the question isn't about the few 'single digit ms delay/difference' of 'total input lag', but rather more like how things are with stuff like 'pixel response' and 'ghosting'/'motion blur' etc.
There you're talking about a 'few single digit ms'difference, often 'sub 1ms'.
No one in their right mind would claim that a human can perceive such a small variation - 1-5 ms difference, alone 0.1 (tho most of these numbers are purely for marketing) any actual numbers wouldn't be much far either, but yet one can see and notice a perceivable and MEASURABLE difference in these cases. Anyone who have researched the topic on 'Motion Clarity' would know.
So, I'm confident that there is a similar 'metric' to be found here.
Only issue is that we are dealing with "double variables", as the human perception can vary, tho it is not to say that a monitor is an "absolute variable" either, as depending on various factor/conditions - stuff like 'Pixel Response Times' can vary quite a bit, even on single factor like 'Temperature'. Generally, Lower Temps = Slower Times.
This debate brings memories of the old "You can't see more than 60Hz" misconception.
We now know for certain that there is a noticeable AND great difference between 60 and higher Refresh Rates, regardless whether and what actual "Hz" the human eye can "see".
Great points.
Its indeed not a black and white story, especially picture clarity which my monitor is certainly not the best if we compare it against those, double or even triple the price.
Those things can have an impact on your performance, and a quite noticable one as well, however it becomes almost impossible to proper sample.
Another debate that I like is DPI and pixel skipping. While this doesn't necessarily have an impact to your aiming ability in reality, I still wonder if pixel skips could be an advantage since you see one or two frames longer before the next skip. Latency it shouldn't be noticable, but since we can still "see" a difference between 240 and 144 Hz, it wouldn't surprise me
@@MichaelNROH I guess I haven't dug that deep into the Mouse DPI thing to notice the 'pixel skipping', but I can say for sure that different DPI is a different experience regardless of any other settings. 400 will 'perform/behave' differently to 800, 1200, 1600, etc., despite you can measure the same 'distance per pixels' (the specific distance you have to move your mouse from 1 point to another).
Similarly, here one can measure identical ms or at most negligible difference in total input lag, tho it will still feel and seemingly behave differently.
If one takes other aspects like frametime f.e., you can have identical FPS and input delay, but you'll definitely notice and be affected if your frametime is all over the place. I don't know if and to what extend one can rely on frametime here, as this likely comes down to the specifics of the rendering and frame generation of the OS, as you mentioned in the video, but I think it can likely provide useful information and give a much better idea of what's going.
It's much more "sensitive" metric (so to say) compared to 'total input lag', as you can definitely notice even minor inconsistencies. Likely what was going on with the "Gnome stutters".
Generally this is the main issue with Linux gaming currently, maybe save for Arch/SteamOS, all other distros seems to struggle with proving a stable frametime, even when they manage to outperform Windows in max/avg FPS.
I think, next time you want to mesure something from a mouse click, use either and equivalent to xdotool, for wayland (sending a mouse click from cli, so you can display it somewhere, film the screen and the output at same time, or evemu (records and replays device events), or if you have some arduinos lying around, it's super easy to make it send a mouse click and light a led at the same time (so, you ask your aduino to make a loop sending a click, lighting the led, keeping it for a few time, wait a second, do it again, for 10 times), otherwise, great video !
That is a good idea if we want to take compositors into the mix as well, but that wasn't my goal.
I wanted to capture fullscreen applications whereas the compositor doesn't have to do much or anything at all, while also exerting CPU and GPU Load.
But yeah, other testing methods could be done for other testing purposes and might show different variances
@@MichaelNROH That doesn’t really make sense because the whole reason I’ve seen people say Wayland can have more latency than gnome is due to not being able to disable the compositor on Wayland for full screen applications, which you can do on gnome.
This was also a problem on windows in non full screen applications until they implemented flip mode into full screen optimizations.
So part of the latency difference you might have felt could easily have been from the compositor especially since your test already had a ton of variability in it.
@@AvocadoBondage There is no compositing in fullscreen.
i use lxqt or lxde they are the best they dont has any composition and thx for video we need more videos for latency on linux
Yeah, I didn't consider anything else besides Wayland for this vid as its where the future is headed but yeah, other DEs that are still on X11 might have an advantage here
@@MichaelNROH Nowadays the difference with X11 is not really that much noticeable so limiting yourself to GNOME and KDE makes sense. Great video btw.
@@Robotta For me on amd side with arch, it always feels like being at least some frames behind in CS2 when using any wayland session (it feels like delayed multiple frames).
I could NOT get CS2 to use VRR (fps_max lower than 144 on 144Hz display) on any wayland compositor (gnome, sway, Hyprland, plasma) , no matter what, other games that sync properly are completely fine and dont result in this delay for me..
Best input feel is still xorg with gnome and slight visual tearing for CS2. Got experimental tearing to work with Hyperland , but this was visually tearing but also had the same feel of multiple frames lagging behind feeling (200-300fps), retesting this every couple of weeks , want to ditch the xorg session.....
These are fantastic videos. What kills me is all the stutter in CS2. I have to work hard to iron that stuff out. Usually disabling multithreading in the game, which is silly.
Yeah, CS2 is terrible currently.
Matchmaking (5v5) works fine luckily, but Deatmatch, and now Arms race drop around 200 - 300fps from my normal average.
I found out that performance decays start, when enough people have joined the server. Probably some hefty networking tasks that pull performance down
Good video! New subsciber here
Welcome ❤️
Hey Michael! It's a great pleasure to meet you. I'm new here on your channel and I'm loving it.
I haven't found a specification yet but which Linux distribution are you using?
Thank you very much and a big hug!
It's Fedora 39
@@MichaelNROH Thank you Michael!
There is no way it's a placebo. All my games feel like they have less input lag on gnome 46 Wayland compared to KDE Plasma 6. With 240hz and no VRR.
And some games have more stable fps like overwatch 2. On gnome it stays at fps lock. On KDE it jumps from 236 to 200 very often.
In World of Warcraft on KDE it feels like I'm moving my camera through water. On gnome is fine, no issues.
I don't care for VRR because I use 2 monitors anyway. And I would only use it for single player games.
Damn i can't wait to see VR smooth on linux too
i am on windows now just because of that.
Agreed
this is unfortunately because of valve. they use a library that uses the cpu to do a lot of calculations so you need a much more powerful cpu on linux when using vr than on windows. On windows they use the same library but run it with opencl, so it runs fast with the gpu. That library actually supports opencl on linux too but its just that valve hasn't enabled that. So its a simple mistake that unfortunately cant easily be fixed since the steamvr driver is not open source. Thats why proprietary software sucks
@@notuxnobux I wouldn't bother about paired performance...right now VR is not running after the new update of steamvr. I left Fedora to use Windows...used to run flawless with Arch.
@@luiz8755 yeah thats also unfortunately a steam vr driver issue. Other headsets dont have this issue when using openxr for example. Its so bad that if any application closes without telling steamvr to shutdown (for example if the application crashes) then steamvr crashes too.
@@notuxnobux do you think it's worth to initialize a Windows 10 Virtual Machine and use Single GPU Passthrough to play with VR?
I setup a VM with gpu passthrough running windows 10 for this reason. Somehow it runs even better than a dual boot and my input has never been more smooth.
VMs generally have more overhead, so I wouldn't trust in it, but it shouldn't be noticeable anyway as long as the output is hooked directly to your monitor
@@MichaelNROH It is! Honestly thank god for multiple sources.
@@juliansvidal can you make a tutorial or just link one on how to do that? thanks in advance
Did you have gamescope running as a nested session in both the cases? If yes, then that can invalidate all the tests.
I don't use Gamescope as it doesn't result in any improvements that I could notice
@@MichaelNROH What about gamescope with no other window manager running.
@@edwardecl You mean like run it without a Desktop Environment?
I don't seem how that's practical, but the latency would probably be a bit less just because there is less background processing
@@MichaelNROH Gamescope is a Wayland window manager when run from the console, it supports VRR.
Would be interesting to see if the compositor adds to latency or not or if there's nothing to call between any of them.
Now a comparison with a light windowmanager please which is written in C or Rust. ;)
Had the best results with I3 and XFCE
That's a very Battle(non)sense-like video. Love it 😁👍
Too deep, so complicated but fun :) information.
I have a lot of freezes/stattering at kde. Don't tryed x11, and don't want even try it. In gnome all pretty smooth.
Why not use the "prefer lower latency" (or force lowest) setting?
Because these settings only affect KDE Plasma and non-fullscreen apps.
This is a setting that most don't seem to know what it does and that's why its removed in Plasma 6
@@MichaelNROH the real reason why it was removed from plasma 6 is because they found a solution that has the least amount of latency possible and also the smoothest, so the option is no longer needed.
My results with the 180 hz monitor and CS2:
Kwin Wayland 179 fps - 14ms
Kwin Wayland unlimited (~350 fps) - 20ms
Windows 11 179 fps - 14ms
Windows 11 unlimited (~400 fps) - 12ms
Could you please share the distro that was used in the experiment?
Fedora 39, as it usually offers the latest releases of each Desktop Environment.
Thanks a lot 😇
It's easy, you just turn off V-Sync, look at that horrible screen tear. You open RTSS and set -1 woala, fluid image, without input lag and fixed to the monitor's fps.
I just switched to plasma... and seeing that workspace switcher for the first time since made me sad :( overview is like half as good
Yeah, the Plasma 6 implementation of it is way better, but its still not Gnome polished since you still can't properly launch apps from it.
You can search for them, but if you already have an instance running (e.g. a browser) then it just finds the window. I wish the could implement the taskbar / panel in it
Oh I guess I'm sticking with kde lol. I pretty much only use my pc for gaming these days.
Interesting, interesting... I'll still keep turning off the compositor in KDE 8D
i dont mind with gnome. really. my gripe beside the mandatory wayland thing(which still slower compare to x11) is the LACKS of customizations to its theme. the most lacking than every other desktop offerings in linux environment. its like if gnome shipped in one of the distros, those distros would implement an annoying "brand color" theming which may not be in favor to the users. and there is no options to change those out of the box. this is in every single distros coming with gnome. but i understand why. just look at their logo. the foot missing a toe. so its only natural that you get that feeling when you enter a gnome environment. something is missing. by the way, fellow fedora user here too. fedora has big potential. it could be the "it" from linux to compete head on with windows.
You can easily theme GNOME with Gradience, a GNOME Circle app
Hi, Michael! What software would you recommend for controlling fan/pump speeds on Linux?
I have never used any since I always adjust these things in UEFI, but otherwise probably fancontrol (or fancontrol-gui), depending on how you want to do it.
I have also seen Coolero in my Software Center, but I don't know if its just monitoring or also configuring.
@@MichaelNROHI will check those out. Thanks a lot for the reply!
@@minecraftpufferfish9066 I use Fan2Go. its a hassle to setup once. But the github of the guy is pretty straight forward. And the best thing is that you can setup combined fan curves. My botom fans react to my CPU and GPU independtly. So if GPU askes more. They spin up. And if CPU only asks more cooling they also spin up. Same setup with my exaust. And the best part. Its extremly low on resources. I know this post is late. But I just love this one.
My own personal hardware - brain, is too slow anyway, so such latency differences don't play any role for me. I'm gaming mostly for the story (RPG games) and I always hated the precision, speed or dexterity games. Although, achievement is also a game hook-up, so sometimes I do try such games, but if they are too challenging, they are not for me. If they can forgive clumsiness, there is a chance I may enjoy it. However, playing competitively? No interest at all.
That's probably the healthier approach anyway 😅
@@MichaelNROH Because story matters most, I was able to play Baldur's Gate III with 10-21 FPS, on the lowest settings possible. It was still fun, and I was able to finish the game. Still, it's nice to have better performance, although for some games it's not crucial, because graphics or performance are not necessary for playing it. Anyway, I bought a new Tuxedo AMD gaming laptop, and once it arrives, I will be set for another 8-10 years (if the hardware will hold). With such needs, I don't have to upgrade the hardware too often. Frankly, I wouldn't change much, but the battery in my current, old Alienware laptop, is declining, and I already exchanged it twice. I'm not sure if it's now possible to buy another one, so I decided it's finally time for a new hardware. Gaming capabilities improvement is only a bonus.
So did the smoothness of the window animation had any mearuable difference in latency?
No, and that's why I showed it.
The thing is, that this settings is meant for just KDE Plasma and Windows that are not running in Fullscreen.
Fullscreen applications get a special treatment and does not need any compositing, so it doesn't matter for gaming at all.
It's even removed in Plasma 6, as I can recall.
@@MichaelNROH That's great, AFAIK Gnome disables composting on fullscreen apps by default, and plasma users had to do this manually with an option to disable it. Glad that they also picked this up.
Yep me too, and it turns out that... IT'S NOT 😜!
For some reason when I was testing 4090 out under Plasma WL recently, I couldn't get VRR to work at all. Maybe it was the games I was testing, but oddly enough I couldn't get VRR working under X11, and I even disabled my other monitors to test this. I think there could be a bug or three in 545 drivers that fails to use VRR with certain monitor configurations or something.
I think NVIDIA still doesn't support VRR on Wayland that's why it doesn't work.
Could be that I missed a release but it should be inaccessible at the moment
do like that dashboard wall paper, were is is from?
KDE Store. I marked its name in my KDE Plasma Video
For me in particular does Gnome Crash for anyone randomly when gaming ?
I haven't encountered an issue like this for years. Which distribution are you using?
I haven't gotten into changing to use wayland because many issues are still present, but I am using x11 kde plasma 6. I was curious if you used kde plasma 6 or not.
I use Plasma 6 exclusively on Wayland.
Hi, thx for this vidéo, can you do TKG kernel vs Stock kernel :)
I use GNOME because when i installed KDE Plasma on my system and played minecraft i couldnt reach 30 fps with a good laptop but when i reinstalled GNOME i was easily getting over a 150 fps
That sounds like something has gone during the installation, since the Desktop Environment doesn't have much of an impact on gaming performance.
In fact, Plasma even has a slight edge over Gnome because it tends to be optimized more quickly and is partially funded by Valve
thats what i thought would of happend. but im happy right now with what i have@@MichaelNROH
Which Graphics card did you use and did you compare it to another? even lets say an older nvidia card vs a newer one (when i can believe comments from some other users the newer cards ar way better supported than the (now) old ones...
I Think - which graphics card you use can make a difference. AMD vs Nvidia and potential Bugs..
From own experience with a Nvidia 1080 it couldnt even move a normal windows lag free aroud on the Desktop - thats obviously a bug... but still...
I used an AMD RX 6800XT, but I own a GTX 1080 as well.
The thing that I noticed when I was on NVIDIA well over a year ago was, that it doesn't run as nicely with DX12 games in comparison to DX11 (besides the regular performance difference).
Nvidia doesn't seem to bother supporting older cards on Linux, as they do on Windows which unfortunately comes with a performance penalty.
But it's not that bad as I could still play some games like Far Cry 6 with like 85 to 90% of its performance on Windows.
Wayland is garbage. I used to think it was just a matter of time before it was usable, but after using it again this week I realized it never will be.
Seeing my 5 year old video pop up at th-cam.com/video/JJZeByBEoZY/w-d-xo.html was quite a surprised haha! Thanks for the video, currently trying to switch my gaming + streaming setup to linux and this helped a lot.
It's a good video 🙂
I found interesting bug in my KDE - when I change animations speed - my game starts to be smoother/rougher, and FPS increase or drops too..
Have you verified this or might is just be placebo?
I initially thought that the compositor settings also had an impact, but it actually doesn't
@@MichaelNROH Actually maybe I'm confused you little bit.
I was talking about dropdown placed in Display -> Compositor tab, not the animation speed on Effects tab
When I set it to Balanced mode - it worked with "nominal" perfomance (about 120-130 FPS on my setup). When I set "Lowest delay" (there is warning about "tearing will be possible") - it drops my FPS то 80-100 and made game process very chunky and rough... like some frames are dropped... But actually this option must made gaming smoother...
So best compositor mode is Balanced, any other ruins gaming perfomance for me.
I think you need to check that aspects and show the difference.
hows the difference in windows vs linux?
Besides the lack of some software solutions like NVIDIA Reflex, the latency is practically the same.
Comparing the latency between Linux and Windows is not as easy as comparing between Desktop Environments, since it depends a lot on your hardware (e.g. Proprietary AMD driver on Windows vs Open Source Driver on Linux) and the games you play.
Every test result on each system could be different, leading to non-reproducable results.
Practically speaking, there is no difference whatsoever, and if there was a latency difference then it would be way below 50ms, which is the time that most can notice that something is off when constantly dragging the mouse.
@@MichaelNROH Nvidia actually has sent in some pending patches to dxvk/vkd3d to enable reflex in proton, although latencyflex is already pretty much the same thing.
@@PeakKissShot I'm glad that some effort comes from them, as LatencyFlex often gets flagged by Anti-Cheats.
Having some option until this is patched would be nice
question: i am running Ubuntu on my pc and i sometimes have an issue where it will completely freeze, mouse and keyboard dont work, even the disk activity light stops blinking. it seems to happen when i close a game that was running from steam and i send a discord message right after it closes. then all web browser related apps (chrome, steam, discord) all say not responding, then everything freezes a few seconds later. any idea what is causing this?
Mh, that's a tough one.
Usually freezes occur when a GPU driver or the compositor crashes, but that shouldn't take down everything.
It could be a bug in the compositor itself since for fullscreen apps there is no compositing going on. Maybe closing the application triggers some issue.
You could try reading through the latest syslog files in /var/log/, or open a bug report with the log attached
@@MichaelNROH Hmm. this morning, i turned on my pc and there was a red minus symbol in the task bar (Gnome extension) and it said "There is installed software with missing dependencies." this has never happened before. i havent installed anything new for at least a month.
@@wibwabz If you generally haven't installed anything then your drive might be broken and keeps breaking bit by bit. You should take a look at that with some online tutorials
@@MichaelNROH Broken drive? like corruption?
How do Valve implement it on their Steam Deck? Can you explain?
When you use game mode, the kde plasma is disabled and it switched to the game scope micro compositor (Wayland)
this video was done on x11?
Wayland
I wonder if this will change with kde plasma 6 thats about to come out
5:15 Halflife 2 yay
Now make one with xmonad and hyprland
And what about ..... Launching the games with gamescope?
I think that many use gamescope wrong, as running it on top your Desktop Environment can even introduce more overhead if you don't run it fullscreen.
It's basically Steam Game -> Gamescope -> Mutter, Kwin, etc. -> Output
What you want is to eliminate the Desktop Environments for Gamescope to properly take control of your screen.
Otherwise, Gamescope just provides better scaling (e.g. Global FSR), Limiting Framerates, custom resolutions, a sandbox that doesn't interact with other things on your system and so forth.
Can you update with commentary for Gnome 46 enabling VRR.
VRR doesn't decrease latency unless you are below the refresh rate limit
Proton games under Gnome Wayland drops frames like mad. Even Chrome and Firefox report this on TH-cam. Anyone else ran into this?
From time to time there are reports, but it's rare in browsers. The much bigger problem is games, though VRR or not, Gnome is refining the way how they synchronize (or check) timings of the GPU as well for no more drops.
Btw. I guess we are talking about dropped frames to the monitor, not actually lower fps in a game. This would be its own issue
@@MichaelNROH I reread my original question and realized I completely botched the explanation. I'll fix it below.
For my 60fps monitor (no VRR) with AMD Ryzen 6000
- Proton games under Gnome Wayland run at 30fps, no matter what I do. I have enabled the Steam overlay to show this information. My monitor is configured at 60 hz, and the game settings are at the lowest they go. Doesn't matter if fractional scaling is on or off. These games look like they're running at 30fps too.
- Proton games under KDE Plasma Wayland run at 60fps easily with the same setup. These games are smooth.
Now for any browser that I use:
- Under Gnome Wayland, I pull up any 60fps video on TH-cam, and about half the frames are reported to have been dropped. Oddly enough, it looks like 60fps content, not the 30fps I get with Proton games.
- Under KDE Plasma Wayland, browsers don't report frames as being dropped, and look the same.
My solution thus far has been to use KDE Plasma / Windows. I really hate running X11, but that does make all issues go away in Gnome.
@@MichaelNROH Apologies. I botched my comment and it's not accurate after re-reading. I've corrected it below:
Under Gnome Wayland with a 60 Hz monitor, all Proton games run at a steady 30fps. I can tell just by how choppy the games are, which is confirmed by the Steam overlay. All other apps appear to run at 60 fps because of how smooth they are comparitively. The other odd thing is that Firefox and Chromium browsers under Gnome Wayland, when playing TH-cam videos that have 60 fps content, reportedly drop ~ half their frames. But it looks like 60 fps content: it's not choppy like the games. And it doesn't matter if I run those browsers under XWayland or Wayland.
Under KDE Plasma Wayland, I don't have these problems, but I have other issues: Any kind of popup (like the volume control bar) causes the Proton game graphics to freeze so that I have to force quit and restart. And Chromium browsers (IIRC, Firefox too) freeze when full screen if they're running in their Wayland modes.
Under Gnome X11, all these problems go away.
I have a hybrid graphics setup (it's a laptop). Both are AMD chips, and the dedicated Ryzen card is a 6000 series (I don't recall off the top of my head). No matter which distro I run, I consistently get the same results as the above. It's odd, because AMD should have the best support. I have no idea, and I can't find my problems reported anywhere else. Just curious as to if you've run into this before?
Also, thanks for any help you're able to provide! I really don't like being on Windows for the moment, but I'd rather be using Gnome if possible.
Plugged into another monitor, and this issue is gone. It's got to be hardware related.
Why not test kde force lowest latency?
Because it makes no difference on full screen applications. Wayland compositing isn't used for it
@@MichaelNROH Ok, but there was a difference between balanced and smoothest animations? Why test both of them?
@@alvesvaren The difference is margin on error and the showcase was only to show that it doesn't do anything
@@MichaelNROHalso how do you know that cs2 is running in full screen? Most newer games have a “fullscreen” option that is really just fullscreen windowed but forcing mouse to stay on screen until alt tabbed.
AFAIK that would mean Linux would still see it as a windowed application
@@MichaelNROHI just tested cs2 while changing the Wayland settings and they do in fact 100% affect cs2 in fullscreen due to the application itself not ever being in true fullscreen.
If you set to lowest latency possible the game will get extremely stuttery like it’s dropping frames and when you go just one above the lowest latency, the latency is definitely there but it’s smooth and no issues with dropping frames
Slightly off topic:
I tested the keyboard latency in Windows and in Linux, in Linux I had constant 250ms of key repeat rate (because I set it as such with xset), but on Windows I had from 220 to 280ms, Windows scheduler/IO suck so much.
X11 latency is better i think
As of right now yes, but not because Wayland is slower, but because almost every game is being run with XWayland and not every DE supports Tearing and VRR yet.
Once that's all in Place, Wayland should be faster since its more "cleaned up" protocol
@@MichaelNROH wayland wont be faster. Both x11 and wayland work pretty much exactly the same way if you run a fullscreen game (and enable tearing in wayland). They even use the exact same kernel (drm) and mesa (gbm) functions.
@@notuxnobux Probably not measurable, but by how the protocol itself is build. Less overhead means more resources and better latency though this is probably even below a ms.
@@MichaelNROH there isn't less overhead with wayland. As I said, wayland and x11 works pretty much exactly the same way when it comes to fullscreen gaming. When you have a fullscreen application that uses opengl/vulkan (and no vsync) it bypasses the display server. A lot of people seem to misunderstand x11 and think it has overhead for gaming "because it's old and bloated", which isn't true. That "bloat" in X11 is only touched when you dont use opengl/vulkan.
nice video!! do we know how free sync/gsync addresses this?
Variable Refresh Rate technologies only really comes into play if the fps are lower than your monitors refresh rate and it depends on your setup.
It can theoretically improve latency, but that depends on how good your monitor is. The lower tiers (there are several) can slightly add latency since the synchronisation takes some processing speed.
But this has a very small impact. Some TH-camrs measured this
@@MichaelNROHgotcha, thanks for the response!
Wow
tearing is not supported by your kernel yet (unless ur rocking 6.8 for whatever reason) and therefore the tearing option won't make any difference right now
How did you come up with that? Tearing is working as a Wayland protocol and doesn't have anything to do with the kernel at all.
KDE Plasma and Gamescope have it successfully implemented
@@MichaelNROH The tearing protocol also required changes in mesa, the kernel, and I believe setting an environmental variable. So unless all of those were present, KDE didn't actually have tearing enabled.
@@MichaelNROH well read up some more, there's a kernel aspect to it as well. google "atomic async pageflips"
@@that_leaflet Oh, you mean "KWIN_DRM_NO_AMS" on Plasma right?
That is set for this video, however there is still a difference how Kwin handles it. From playing without the setting, KWIN seems to behave similar to Gnome with more stutters though it's most likely a combination of all the patches in place (Kwin, Mesa, XWayland, etc.)
@@MichaelNROHNO_AMS falls back to DRI2, what you really want is DRI3. DRI3 async pageflips were only implemented recently in kernel 6.8
like
To have better latency you need to have tearing at least but it have a lot of issues under wayland even on kde. Gnome is completely unusable for me from this point of view. On kde it work stable under the x11 session only. Under wayland it sometimes works, something doesn't.
Long story short. If you're into gaming, get a monitor that supports VRR and play in KDE while limiting the FPS slightly under the maximum refresh rate of your monitor.
I wouldn't recommend this typically.
If you want a smooth experience then yes, but for competitive gameplay, you want higher frames and utilize tearing for a faster response time.
Thats why I just left Linux gaming. Linux is great (for now) for programming, browsing web, listening to music all things what Windows and Mac does minus gaming. Proton, DXVK, VKD3D are great thing, and they'll fight Microsoft monopoly in gaming, but for now when Linux NVIDIA drivers are terrible and many games are not launching after a freaking driver update, I'll stay on my Windows VM with NVIDIA GPU for gaming and AMD GPU on Linux for displaying desktop and that's it. It's my setup. Arch + KDE Plasma (wayland) and KVM with GPU passthrough for gaming
Is it me or the tone of the guy talking is super annoying (fake)? I couldn't finished the video even though the topic really interest me. I imagine an old news anchor guy talking with that tone, fake af.
it's not just you. His videos looks interesting, but every video is a pain to watch because of the way how he is talking, how every few words his voice pitch goes very high, it's like on a rollercoaster.
Garbage. If you're wanting to game or get the most out of your office hardware, then use Windows. If you're wanting to just do general tasks, surf the Internet, watch movies, write texts, and don't care about whether your latest hardware and all its features will be fully supported by an operating system, then use Linux.
Windows itself draining all the horsepower that's the issue
@@publicname No, it's not. Games run perfectly fine under Windows 10, for me.
@@atlantic_love 7, 8.1 and 10 were fine tho but 11
bah que tonterias, yo juego en linux perfectamente, al menos los juegos que me gustan funcionan bien. Genshin, honkai star rail, HZD, spiderman remastered, god of war ragnarok, doom eternal, star wars jedi survivor, callisto protocol.