Hi folks. I recently found in the Games forum that Steam stopped launching for me because a recent change to my GPU driver software meant that I needed to install a 32 bit module which was separated off into its own package. Whilst I was looking for the package to install, I realised that during my initial installation of 15.3, I had somehow failed to notice the presence of nvidia-glG06 and associated packages, and had installed the nvidia-glG05 set, instead. Since my GPU is a GTX970 and the nvidia-glG06 packages support 700 series and above, I thought it would be sensible to switch from the nvidia-glG05 to the nvidia-glG06 group of packages at the same time. The television my computer uses as a display provides incorrect EDID information which results in font sizes so small that most characters displayed in the U.I, are barely bigger than 1 pixel by two pixels in size. I did have some notes saved somewhere on how to make the fonts readable again, but I lost them. I’m fairly sure I just modified my X11 config to disable the use of EDID, add a modeline from cvt, and then change to physical measurements of the screen to a different incorrect value which made the the font sizes readable (480mm × 270 mm ≈ 100 dpi)
After installing the the nvidia-glG06 versions of all the software, I restarted my computer and found that the text in the U.I. hadn’t become unreadably small, but everything does look quite fuzzy and low resolution. I had a peek at my x11 configuration file wich I normally edit to correct the font sizes, just to see what it looked like, and it doesn’t seem to have changed at all, so losing my notes on editing the config doesn’t seem to be the problem I expected it to be, but that also seems to mean that the X11 config is being ignored. I had a look at my Display Configuration in the System Settings found in the Suse menu, and the settings I see there seem like they’re probably what’s in effect now, but they’re totally wrong and there are no other options to choose from. The resolution setting is 1024×768, but my television is 1920×1080 (which would explain the stretched and fuzzy appearance) and the refresh rate is set to 76 Hz, but my television is 50 Hz. 1024×768 is the only resolution available in the resolution drop box, and 76 Hz is the only refresh rate available in the refresh rate drop box.
Like I said, the information in my /etc/X11/xorg.conf.d/50-monitor.conf remains correct and unchanged, but it also seems to be completely ignored—they aren’t even offered as options in the Display Configuration drop boxes.
lol, yeah I did. I used YaST to install the nvidia-glG06 packages, so as soon as those packages were selected for installation, YaST automatically selects all of the nvidia-glG05 related packages for removal.
Quick update to add a little bit more information:
I’ve just seen that KDE’s desktop Software Updates notifier has told me there are a few updates available, and that one of them is for “axf86-video-vesa”. This made me think that my computer might be loading a default VESA driver, rather than the nvidia one, so I asked hwinfo about my card, and here’s what it said:
I notice there’s no mention of nvidia in the driver modules, but I don’t know if that’s how it’s supposed to be after the recent restructuring of the drivers, but I have my doubts. I also notice the the driver activation command given, is for [FONT=monospace]nouveau. I remember that,years ago, the [FONT=monospace]nouveau diver had to be blacklisted to allow the Nvidia driver to work properly, but I’m pretty sure that was completely resolved a long time ago.
Maybe it might be better to wait for 15.4 to hit the shelves, and then just do a fresh install.
[/FONT][/FONT]
What have you actually installed? Try “zypper se -is g06”:
**pluto:~ #** LANG=C zypper se -is g04
Loading repository data...
Reading installed packages...
S | Name | Type | Version | Arch | Repository
---+---------------------------+---------+-------------------------------+--------+------------------------
i+ | nvidia-computeG04 | package | 390.151-lp153.12.1 | x86_64 | nVidia Graphics Drivers
i | nvidia-gfxG04-kmp-default | package | 390.151_k5.3.18_57-lp153.12.1 | x86_64 | nVidia Graphics Drivers
i | nvidia-glG04 | package | 390.151-lp153.12.1 | x86_64 | nVidia Graphics Drivers
i | x11-video-nvidiaG04 | package | 390.151-lp153.12.1 | x86_64 | nVidia Graphics Drivers
**pluto:~ #**
(Just think of G04 as G06.)
You should have these for packages. If yes, but not running it may well be that your GPU is just not properly supported. I have the same issue with mine. According to nvidia documentation it should work with G05 but that just won’t work. Then you might ask in the nvidia board or just try the G05 driver (if only as interim solution).
Looks like the output says the Nvidia driver is not loaded. Does this mean the G06 packages just won’t work with my card? I take it the solution is to simply roll back to G05?
For sure the nvidia driver failed to load. I’m not sure, why. I didn’t get that thing about the “32 bit module which was separated off into its own package.” Why would in a modern system the update of the GPU driver require a 32 bit package? Can you provide more information about this?
What if you just try to remove that package in YaST or
zypper rm nvidia-computeG06-32bit
Anything complaining for dependency going amiss? If not, just remove it and restart. Any change?
inxi -Ga does a much better job of providing useful Graphics information, especially after upgrading 15.3’s broken antique version to something recent. Use inxi -U to upgrade to upstream’s current. If and when Leap gets an upgraded inxi, what will actually happen is a downgrade, unless the new rpm arrives within a month or so of latest upstream release. At some point you’ll need to use -U again if you wish to stay current. If you don’t want downgrades to happen, uninstall the inxi rpm, and install the upstream script.
<xdpyinfo missing>
That will go away if you install package xdpyinfo.
xf86-video-vesa and xf86-video-fbdev are fallback drivers, something capable to provide a crippled X instead of blackscreening when the KMS or NVidia’s drivers aren’t working. Metapackage xorg-x11-driver-video causes installation of xf86-video-. If using NVidia’s proprietary drivers, none of xf86-video-need to be installed, if NVidia is working, so it’s useful to keep the two fallbacks.
Why would in a modern system the update of the GPU driver require a 32 bit package?
For 32-bit applications like steam. The 32-bit driver and libs where included before in the “normal” package.
I am running the G05 package (including the three 32-bit rpms for Steam) on openSUSE Leap 15.3 with a Nvidia RTX3060 without problems.
The G06 package is based on Nvidia driver 515.43.04 which is a beta version. Maybe this is the problem.
Following a similar path to yours I found that the /etc/modprobe.d/50-nvidia-default.conf was no longer present. Restoring it from 50-nvidia-default.conf.rpmsave solved the issue for me if I remember correctly.
Nvidia has apparently decided to make their drivers open source and integrate them into the Linux kernel, but from the tiny bits of information about it that I’ve heard, they’re doing it in a staged process where they’re splitting it up into smaller components and merging them bit by bit. This 32 bit package could be part of this process, but I honestly don’t know. I just know that, like Kieltux said, if I want to run 32 bit software (such as Steam) I need the 32 bit package installed.
I tried uninstalling the 32 bit package anyway, just to see if it might help, but unfortunately it didn’t. My next logical step was to roll back to the G05 packages and see if everything went back to normal after a reboot like I expected it would, but that didn’t help either. As things stand now, I’m pretty sure I have the same packages installed as I did when everything was working fine, but I’m still stuck with this stretched out, low resolution setting (presumably 1024×768@76Hz). Here’s what inixi and zypper say about the state of the system right now:
Thankfully, FrankyU2’s post gives me a bit of reason to be hopeful. I haven’t tried it yet, and I was worried that the conf.rpmsave file might have been overwritten during the process of rolling back to the G05 packages, but there does seem to still be a difference between the files which might dig me out of this hole. I compared the two files with diff, and diff shows that the only difference between the files is a line containing a reference to something that used to break Nvidia drivers years ago:
I remember that, years ago, I used to need to have (think) “nomodeset” specified on the OS selection menu of Grub, otherwise the Nvidia driver couldn’t be loaded. So I’m going to try restoring the rpmsave file and hope for the best. Wish me luck!
Just read mrmaxda’s post again because I didn’t quite understands it the first time, and I noticed that he was suggesting to use inxi -Ga, rather than just -G. I did notice that xdpyinfo was missing because I remember I had that in the notes I wrote for myself about getting the driver installed, but then lost. I just avoided installing it to reduce the number of variables in this problem. It’s installed now, though, so here’s the output of inxi with xdpyinfo installed,
Sorry, I’m not that sophisticated with these things. If you’re back at the same level that has been working before, I’m stuck. nvidia is still not loaded. You may yet make sure that nouveau is blacklisted. Others may be able to give better advice.
That’s okay. I’m thankful to anyone who even tries to offer solutions. My knowledge about hardware configuration is negligible. I tend to write little tools and bits of things for friends and family so, to them, I’m their computer guy, but whenever they ask me about computer hardware I always tell them the same thing: “Hardware is Voodoo”. Even on Windows, there’s just too much that can go wrong with it, and I know that I know far too little about most of it.
I’ve installed xdpyinfo and the 32 bit G05 module again, but yeah, as far as I can tell, all of my installed packages are exactly as they were before I installed the G06 modules which caused the problem, but the problem still persists. The only X11 config file I’ve ever actually changed has remained unchanged throughout the entire process, so I have absolutely no idea where to look for potential problems.
Given that it won’t be long before 15.4 hits the shelves, I’m mostly just interested in getting this solved as more of a learning exercise. It might be helpful to go into the transition the 15.4, armed with a tiny bit more GPU troubleshooting knowledge, especially given that Nvidia is reportedly making significant changes to their drivers.
I also made a backup of 50-nvidia-default.conf before I overwrote it with the rpmsave version as suggested by FrankyU2. I notice that the backup contains references to modeset, but the /etc/modprobe.d/50-nvidia-default.conf file that I’m using now doesn’t contain any references to modeset. Would this be the equivalent of having nomodeset in Grub?
I can’t even remember what modeset is about, it’s been so long since I even had to give any thought to whether or not it is in use.
If xf86-video-nouveau is installed, most likely instead of modesetting as above, you would find nouveau in its place as loaded display driver.
I suggest to try fully purging the proprietary drivers now installed, verifying success by getting the FOSS drivers working similar to above. Once you’ve reached that state, you should be able to install appropriate NVidia drivers successfully.