4080S blackouts, crashes and freezes

Graphics / Linux Linux
1

Back in May, I self-assembled a new desktop computer, and proceeded to install PopOS, migrating from Windows on my old laptop. I experienced a lot of bugs and glitches, but nothing so serious I could not comfortably use the computer for my main purpose, games, browsing, etc. Over a month ago, that was no longer possible. Now Minecraft (the only “native” game I have), is the only game that works. Every other game (Wine) blackouts (2-5 seconds long, approximately once every 30-60 seconds) or brings with it a display output crash of some sort. In fullscreen, the entire system freezes (although sounds keep playing, so other CPU cores aren’t blocked). I have given a more extensive report over at PopOS’s Github repository. PopOS developers cannot find the issue (I’ve also talked with them in chat). I thought there was a graphics driver update that started the problems, but the apt logs show that was not the case. It may have been caused by GTK/Gnome updates. That I’m not sure of.

I do find it a bit strange that eglinfo reports Mesa as the EGL vendor, but I’m not sure how to fix that;

EGL API version: 1.5
EGL vendor string: Mesa Project
EGL version string: 1.5
EGL client APIs: OpenGL OpenGL_ES

My hypothesis is that the DE services (OpenGL) are somehow fighting with the Wine games which seem to all map to Vulkan. For instance, in the few cases that I actually do manage to get a game running in fullscreen, it blackouts every time an overlay from the DE is rendered, such as volume up/down.

I don’t have the necessary familiarity with Linux to proceed with debugging and solving the issue on my own.

3

Your system changed, and is now misconfigured,

that’s the summary of every problem.

Please attempt to correctly install wine,

following instructions for ubuntu on winehq.org

and attempt to correctly install nvidia

following it’s instructions… that do not actually exist…

Make sure the display server is x11 first of all,

that’s what you want to use with nvidia

figure out what version of ubuntu your version

of pop os is based on, install all your systems updates,

then add support for 32 bit libraries, for proper
support of gaming libraries wine and nvidia,
will install for you

sudo dpkg --add-architecture i386
sudo apt update

follow the instructions for installing the wine
ubuntu repository, and installing wine and it’s
dependencies from
https://wiki.winehq.org/Ubuntu

and install nvidia like this…

Get the latest run file from this download link

drivers are also listed in a post stickied in this thread

It will automatically uninstall, your old bad install of nvidia, and then after rebooting, then running it again, it will install itself properly.

The way I install it, is to edit the grub command line at bootup, to get to a terminal, instead of the graphics environment.

You do that by highlighting the kernel you want to boot up with, at the grub menu, when you reboot your computer, and pressing the e key, in order to edit it’s command line, and then adding to the end of the line after where it says

/vmlinuz root=UUID=fa423d0a-6f35-4358-jr11-e2f3489k9fub ro

the number 3

so it looks like this…

/vmlinuz root=UUID=fb483d0a-6f34-4358-b211-e2f3389c9ffb ro 3

then press f10 to boot, and you will boot into a command line, where you can run the nvidia run file repeatedly to solve all your problems, by simply following it’s instructions…
(you gotta type in the location of the file,
and run it as root with sh like this)

sudo sh ~/Downloads/NVIDIA.run

You have to run it, to uninstall nvidia, then reboot, edit the grub command line again, to get to a terminal., then run it again, to install nvidia, and just keep following it’s instructions, and you’ll be good to go…

If it’s not a problem related to misconfiguration of
software which is the most typical, then you have
a hardware problem, and should make sure your hardware,
is seated properly, by reseating it.

If you follow all those steps, get everything installed,
and some game still fails to work, you can report back
here, about how you are trying to run it, and I can tell
you how to get it working too…

4

Thanks for the reply.

following it’s instructions… that do not actually exist…

See, that’s why I’m struggling! :P

The display server is X11/Xorg. Have never used WL. PopOS is based on Ubuntu 22.04 (as you can see in the log file).

Why would I install x32 on an x64 CPU (i9-14900K)?
That said, I’ve already run those two commands, as part of my attempt to swap the graphics driver (think I eventually gave up because I could not get it to work), see these chat messages for the troubleshooting process (pasted the link into the thread starter post).

Downloading the drivers from Nvidia directly, and setting up things on my own, doesn’t work. If I was to succeed at installing another driver, I’d need to pull PopOS’s graphics repository and build from that source. I tried it. Else, the RUN files from Nvidia fail with build errors, and I cannot proceed to the next steps you’ve listed.

I have used Wine from within Lutris, as well as from within Steam. In Steam, I’ve tried selecting (almost) every single available Proton version, including the Hotfix. They all pretty much behave identically.

Since I’ve assembled the computer myself, I’ve reseated the GPU several times. Been as careful as possible, so I’m positive I haven’t inflicted any damage as part of the process.

5

I was initially at 550.67 when things started failing. I then tried the not-yet-released PopOS-specific-install of 555.58. No changes. I then reverted to 550.67. No changes. After I came home from a vacation, 555.58 was now released for PopOS, available in Pop!_Shop. Made no difference. I’m not so sure if it’s the driver itself that’s the cause. It may be the way it interacts with other system services. For instance, should the EGL vendor not be Nvidia? It says Nvidia in the Nvidia Settings application, but eglinfo says Mesa. Are two versions installed and simultaneously in use?
eglinfo also reports two errors. Not sure if they’re relevant, but including them anyway:

MESA-LOADER: failed to open simpledrm: /usr/lib/dri/simpledrm_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib/x86_64-linux-gnu/dri:\$${ORIGIN}/dri:/usr/lib/dri, suffix _dri)
GBM platform:
MESA-LOADER: failed to open simpledrm: /usr/lib/dri/simpledrm_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib/x86_64-linux-gnu/dri:\$${ORIGIN}/dri:/usr/lib/dri, suffix _dri)

Also, small observation. Fallen Order’s main menu renders fine and responds to mouse movement, in fullscreen, when another application’s window is visible on top, (cmd+tab). As soon as FO becomes the front window again, it freezes. Strange.

6

follow the instructions for installing the wine
ubuntu repository, and installing wine and it’s
dependencies from
Ubuntu WineHQ Repository - WineHQ Wiki

Followed the steps, and chose the stable version. After that, restarted, tried Apex Legends. No change. Checked apt show wine, it still said 6.0.3. Then saw there were a set of Vulkan updates in Pop!_Shop. Installed them, to then notice Steam had already begun a download for an Apex Legends update. Eventually, Steam crashed (during the install I think: the fans seriously spun down). It did not want to open again. I found it still running in System Monitor, however. After a while, I force quit and relaunched it. Apex Legends wouldn’t run. Restarted.

On the other hand, Fallen Order (Lutris → EA App → FO) now started and ran in fullscreen. I played it for a few minutes on max graphics settings. It stuttered quite much, something I’d not expect for an older game on brand new high end hardware. There were no blackouts, but just after a minute or two, the display output crashed. While writing this post, the signal came back. It’s now had several short blackouts. Eventually, it crashed again, I reconnected the DP cable (yes, it’s in the GPU), and FO switched to windowed mode. At least with that, zero blackouts and crashes thus far. Overall, a great improvement for this game, although it ran pretty much flawlessly 2 months ago.

Battlefront II took a good while to startup, but eventually loaded up the fullscreen window and started the intro. Got a blackout just now while waiting to enter a match. Not expecting any different experience than with FO. (BF2 is additionally glitching with sound)

If I remember correctly, this is a pretty similar experience to when I was troubleshooting with the help from a PopOS developer. I updated the motherboard firmware. That slightly improved the game experience (from not working at all to the state I’m now experiencing), for a couple of days, before going straight back to not working at all.

May you correct a possible misunderstanding while we’re at it?
I was under the impression that the Wine install bundled with Ubuntu was not being used. I thought that Lutris and Steam use their own updated versions of it? In general, I don’t really understand how these components all work together, even after repeatedly reading explanations from various places.

Edit: even Minecraft (vanilla, 1.21.1), is now affected by short-burst blackouts. That was not the case before. I haven’t actually been playing MC yet on this computer, so last time I tested it was 0.5-1 month ago, after the issues with Wine games showed up.

(Video playback in FF is not affected whatsoever)

7

https://registry.khronos.org/EGL/sdk/docs/man/

EGL™ is an interface between Khronos rendering APIs such as OpenGL ES or OpenVG and the underlying native platform window system.

The reason it is ideal to update wine from winehq’s stable repo is twofold, first they are the actual provider of wine, unlike a third party website, the second is that different scenario’s will expect libraries to be on your system, installing all these gaming programs, and libraries will pull in all sorts of dependencies, that are useful for gaming in general, or broadly.

sudo dpkg --add-architecture i386
sudo apt update

wine (from it’s repo as it instructs us)
sudo apt install --install-recommends winehq-stable
lutris
sudo apt install --install-recommends lutris
nivida
sudo apt install --install-recommends steam
and even steam

Grabbing all these dependencies is what makes gaming on linux easy, and avoids the headaches you are describing.

Now, when I’m gaming, and even if I do not want to use steam, or do not want to use lutris, I follow those steps above in order to get appropriate dependencies for gaming.

adding 32bit support is essential because it lets you get 32 bit libraries that wine, and many of it’s dependencies (which are 32 bit packages) install on to your system. Some stuff for gaming is 32 bit, and we rely on it. You should have x86 support on your 64bit cpu, as that is typical, called: x86-64

Ok, so for making games work, take them 1 at a time first of all.

Use protondb, a website managed by steam-valve, for using wine-proton to make steam games work. There are user provided instructions for getting games to work.

this is the page for apex legends

One guy said all he needed to do was

Tinker Steps:Switch to experimental

which means he used the experimental proton to run the game

proton Experimental
Distro:Pop!_OS 22.04 LTS
Kernel:6.9.3-76060903-generic
RAM:32 GB
GPU Driver:NVIDIA 555.58.02
GPU:NVIDIA GeForce RTX 3070 Ti
CPU:11th Gen Intel Core i5-11600K @ 3.90GHz

He’s even using PopOS

You have a lot of weird errors in your error log, and not a lot of useful information to me. I’m more interested in information that isn’t so technical and complex. I would like to know:

  1. how you installed your system generally

  2. whether or not you updated your kernel from the distro or externally. (it looks like it wasn’t installed properly.)

(you have a lot of errors related to your kernel)

it looks like your kernel, and your driver both did not get installed correctly: which is a common problem.

For a quick fix, you might try installing a new kernel, and then installing nvidia again, being careful to pay attention to any error messages, and instructions.

If the nvidia.run file isn’t working, you have a major problem that you need to fix.

You might even try one of the projects that build the kernel, or use one provided from the distro itself

8

The reason it is ideal to update wine from winehq’s stable repo is twofold, first they are the actual provider of wine, unlike a third party website, the second is that different scenario’s will expect libraries to be on your system, installing all these gaming programs, and libraries will pull in all sorts of dependencies, that are useful for gaming in general, or broadly.

So Lutris and Steam don’t actually pull more recent versions of Wine themselves, but use the apt installation?

Lutris has been installed from Pop!_Shop. I first tried the “Pop!_OS (deb)” version (apt, I assume?), but that seemed like it was just outdated, and buggy, so I changed it to the “flathub (flatpack)” version instead.

I can’t remember how I installed Steam, if it was a download from Steam’s website, or with an installation script from ProtonDB. I may have reinstalled it from the terminal later on.

which means he used the experimental proton to run the game

I’ve already tested that by selecting it in the properties for AL in the Steam UI.

EGL™ is an interface between Khronos rendering APIs such as OpenGL ES or OpenVG and the underlying native platform window system.

Thanks, I know, but it doesn’t really explain how EGL interacts with X11/Xorg. For instance, it doesn’t explain how a different EGL vendor can impact non-OpenGL applications.

You have a lot of weird errors in your error log, and not a lot of useful information to me. I’m more interested in information that isn’t so technical and complex. I would like to know:

  1. how you installed your system generally

That is something I can’t really remember every detail of, but here goes a probably more detailed than necessary description:

Had a HDMI cable (Televison) plugged into the motherboard for a few days (I ordered in bulk, so the monitor and chassis fans arrived a few days later). Don’t think I’d plugged the GPU in yet. Before installing W11 and afterwards PopOS, I enabled one or two basic Intel overclocking technologies in the UEFI (If these are relevant, I can check the exact ones).
I initially put Windows 11 on an exFat USB stick, and proceeded to install that. The setup process thereafter wasn’t as smooth as expected. Some halfway through the setup process, I changed my mind, and decided I wanted to make a proper attempt at migrating to Linux. Waited until the rest of the components arrived, plugged in the GPU, remaining chassis fans, put the side panels on, and plugged the monitor’s DP cable into the motherboard, and checked the UEFI (everything seemed fine). Picked PopOS, downloaded the latest from their website (Nvidia), flashed it to the USB stick with BalenaEtcher on W10. Plugged it in, and ran the installation, with a complete overwrite/formatting of the disk. (“Clean Install”, I think, not “Custom (Advanced)”) The exact options I chose at the setup process, I can’t remember. System language: nb/no. Keyboard layout, honestly not sure which I picked, and which ones I added to the list later on. Did not enable disk encryption, simply because I didn’t trust Linux not to bring me some problems, so I wanted easy disk access from USB. After the installation was complete, and I was booted into the desktop, I installed the Nvidia graphics drivers from Pop!_Shop, along with complementary system and DE libraries, before eventually moving the DP cable to the GPU.

  1. whether or not you updated your kernel from the distro or externally. (it looks like it wasn’t installed properly.)

I don’t think I have touched that at all…
Actually, I have. The PopOS developer told me to boot up with an older kernel that PopOS bundled, to check if I experienced the same issues with that. Yes, I did. Other than that, I don’t think I’ve touched the current kernel. If it’s changed since my initial install, that’s because of an update either in Pop!_Shop or the Settings application.

it looks like your kernel, and your driver both did not get installed correctly: which is a common problem.

That is very much a possibility, as far as I’m aware.

If the nvidia.run file isn’t working, you have a major problem that you need to fix.

As far as I can understand, that’s not correct. As you can see in the discussion on Mattermost, that file is not expected to execute successfully by default. Nvidia driver installations are apparently handled differently on PopOS.

9

If I do copy my data off the disk, a complete reinstall with a disk wipe followed by “manually” downloading and installing the 1 TB of application data again might work out, but I might also just run into the exact same issues again, possibly after another PopOS update.

10

Lutris and steam use their own custom wine packages. Whereas using the wine repo, you will get the latest wine, including it’s dependencies. I call proton now, wine-proton to make sense of the situation, because proton is just custom wine package. Lutris makes it’s own custom wine packages, and both of these are so good, that sometimes we even try to use them outside of their applications to get stuff working too, like using proton from lutris, or like you would typically using regular wine.

Gaming on linux is not like gaming on windows, where we can just keep updating everything, and all our hardware/software performs as we expect it to. in fact, it’s the complete opposite situation, but it has seen major improvements over the last ten years. Take a look at different guides for getting games to work and the situation will become clearer to you, it often takes a lot of tweaking. The plug and play experience does exist now, thanks to the custom lutris, and steam wine packages actually, but there are still many problems.

Ya know what, it looks like pop-os has an iso that has nvidia preinstalled for you, do you know which iso you used?

screenshot
screenshot612×521 125 KB

I would recommend, since it seems that you are new to linux, to do a clean reinstall, using the nvidia iso they actually provide to give yourself the best experience. Make sure you get your system updated fully through their package manager, and see that everything is working, and then you can try doing things like branching out to third party resources to expand your systems capabilities.

To set yourself up for gaming the procedure is the same for all the distro’s.(the commands are different)

First add 32 bit library support

sudo dpkg --add-architecture i386
sudo apt update
sudo apt install lutris
sudo apt install steam

That will pull in all the different dependencies you need for basic gaming.

Using the stock system will allow to to benefit from the experience of other users of PopOS too, and their resources, such as how-to guides ect.

11

Yes, the one in the middle.

But, I am using the stock option. I haven’t installed any other kernel, etc, and all drivers, system services, DE, etc, are selected by System76.

12

Ok perfect, now we have more information,

I would recommend you, to step away from stock-pop, and take control of your system to get it working.

Use the instructions I gave you, to install a liquorix kernel, and a brand new nvidia driver, straight from nvidia themselves…

That’s what I would do in your situation.

Get those two things working correctly, treat your system like it’s ubuntu, and don’t rely on the pop-os specific stuff, and you should be good to go.

You have to be patient, carefully read instructions, don’t rush, take notes of error messages if you find them.

13

I’ve done an OS reinstall while keeping the home folder (the reinstall option in the Settings application). I then tested the various games, same issues. Then installed available updates (mostly graphics or DE ones) in Pop!_Shop. Then tested again, same issues. Created a new user account (admin), then set it up with all defaults (except I chose the white UI). Proceeded to install Steam, then downloaded Apex Legends fresh within Steam (4 hours). Just like I’ve had issues with at various stages before, the game won’t even start now (same with various Proton versions), just gets stuck taking forever at some percentages of “Processing Vulkan shaders”, while the chassis fans are spinning fast due something generating heat. The GPU fans remain disabled (by design). I’ll see if this shader “processing” ever completes. Unless I update the post, there’s been no progress or noteworthy change. I’ll try to see if I can copy my data off the disk, do a wipe, and test again with a clean install, tomorrow. If not, try Ubuntu 24.

14

No, this is somehow worse. Switching from 60Hz to 143,99Hz in Settings somehow crashed the DE or windowing system, Settings now says the changes cannot be used due to possible limitations in the hardware, and Nvidia X Server Settings blanks the field to “auto”.

Nevermind that, since I could change it by setting a specific resolution instead of “auto”. That too freezes up the system, blacking out parts of the screen (for instance the background image) for a good while. I don’t know if the change was reverted (the UI says not) (the freeze lasted longer than the confirmation deadline), as the system is generally lagging now. The screen isn’t tearing like it usually does without “Force Composition Pipeline” on, and it feels more like 10-20 fps when moving an application window around.

15

Andf no one in thread even point on it…

You problem - may be - not linux, not Nvidia but your Intel CPU.

I myself for last 6 months - in this Nvidia forum saw multiple “bugreports” - where people assumed “their Nvidia GPU is bugged” but it was Intel.
For example look this thread RTX4090, torch, kernel tried to execute NX-protected page - #5 by s48gs.w

I would recomend you to install Windows and see if you have same stability problems there - if you do - run to RMA your damaged CPU.

16

Eh, seems it’s somewhat fixed now? See the last three paragraphs.

  • Disabled X.M.P.
  • Switched Intel Default Settings from Extreme to Performance
  • Changed Intel(R) Adaptive Boost Technology from Enabled to Auto
  • Disabled Hyper-Threading

Upon exiting, it presented me the full list of changes:

  • Intel Default Settings [Extreme] → [Performance]
  • Ai Overclock Tuner [XMP I] → [Auto]
  • Intel(R) Adaptive Boost Technology [Enabled] → [Auto]
  • DRAM Frequency [DDR5-5600MHz] → [Auto]
  • Performance Core Ratio [By Core Usage] → [Auto]
  • Efficient Core Ratio [By Core Usage] → [Auto]
  • DRAM CAS# Latency [34] → [Auto]
  • DRAM RAS# to CAS# Delay Read [45] → [Auto]
  • DRAM RAS# to CAS# Delay Write [45] → [Auto]
  • DRAM RAS# PRE Time [45] → [Auto]
  • DRAM RAS# ACT Time [108] → [Auto]
  • CPU Core/Cache Current Limit Max. [400.00] → [Auto]
  • CPU Graphics Current Limit [400.00] → [Auto]
  • DRAM VDD Voltage [1.40000] → [Auto]
  • DRAM VDDQ Voltage [1.40000] → [Auto]
  • PMIC Voltages [Sync All PMICs] → [Auto]
  • Hyper-Threading [Enabled] → [Disabled]

Starting up the computer, it quickly shut down and started itself again. Logged in, realized the DP cable was still in the motherboard, shut down, moved the cable to the GPU, started the computer. Before the login options appeared, the screen was blinking with the normal grey, and a black background. As usual, I was stuck on a long black screen after logging in; not sure why it takes up to 10 seconds to load in the desktop. Immediately attempted to launch Apex Legends. Apparently, there was a 4.66 GB update.

I also noticed that Pop!_Shop had updates, specifically (read: “Version”, it’s in NO):

  • org.gnome.Platform.Compat.i386
    • Versjon: 45
  • org.gnome.Platform.Locale
    • Versjon: 45
  • org.gnome.Platform
    • Versjon: 45

I tested two games without first installing these. Apex Legends once again stuck at “Processing Vulkan shaders”. Clicked “Skip”. The game then eventually launched. Seemed to have the same issues as before. The display output then crashed when I used the keyboard to adjust the volume. Reconnected the DP cable. Now the screen blacks out every time I adjust the volume (volume indicator renders on the screen. I think that’s why it’s blacking out; it can’t handle both sources). The gameplay is a bit more smooth than before.

Fallen Order had no issues launching, but there were multiple blackouts even before the launch screen had finished. Eventually, display output crashed here, too.

Installed the updates in Pop!_Shop. Restarted. Not much difference other than that the framerate is more unstable and much more unpleasant in AL (not FO).

Can be the CPU, but I’m not entirely convinced. Is Nvidia X Server Settings supposed to display incorrect information? Thermal settings, for instance, only says my GPU has 2 fans, while in fact, it has 3. At the same time, the “Enable GPU Fan Settings” has never worked. It does report the temperature, though.

It’s worth noting that disabling “Allow G-SYNC/G-SYNC Compatible” in Nvidia X Server Settings stopped the screen from blacking out when I change the volume (and the volume indicator is rendered just fine). Additionally, no more display output crashes.

Actually… Everything seems fine now after disabling that… Apex Legends ran flawlessly (well, a tiny bit of lower framerate sometimes, but I’ve always experienced that in this game, even on the old Windows laptop).
Has the solution all this time been that simple? But I thought I tested disabling that option almost a month ago? Apex Legends is still struggling to launch. I simply have to click “Skip” while “Processing Vulkan shaders”. I can remember that didn’t always before either.

Is this a bug? My monitor is G-sync compatible. Samsung G7. Maybe it’s a combination of changes?

17

Fallen Order also ran flawlessly now. I even have much less screen tearing on the desktop, without enabling “Force Composition Pipeline”. Let me try enabling that. (Once enabled in the past, disabling it has not been sufficient to revert the impacts it makes on the games. Only a restart was sufficient)

Enabled “Force Composition Pipeline” while FO was running. The entire system froze. Takes at least 10 seconds for it to unfreeze, at which point it’s only unfrozen when another application’s window is rendered topmost (bringing it up with cmd+tab). The game renders fine in the background. When putting FO firstmost (clicking on it, or minimizing the overlaying window), it continues rendering for 0.5 seconds before freezing up. Seems it renders a new frame on a rare occasion. Shut it down, and started up AL instead. Now too, my system is frozen while the game is still on the splash screen.

Considering the changes by disabling G-sync, can I consider this a bug?

18

Yeah, I’ve located the offending settings. I may have enabled these by myself some while ago; not sure if they default to on or not. Either way I feel very stupid for spending such a long time figuring something as basic as that out, especially as I may have caused all this by enabling the offending setting in the first place.

  • Nvidia X Server Settings
    • X Screen 0
      • OpenGL Settings:
        • Allow Flipping

If this setting is enabled, I experience the behaviour I’ve experienced when either “Allow G-SYNC/G-SYNC Compatible” is enabled (blackouts and video output crash), or when “Force Composition Pipeline” is enabled (system freeze when a Wine game’s fullscreen window is key.

I guess it’s up to Nvidia, PopOS, or X Server developers to decide if this is a bug.

19

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.