Getting nvidia to work on my old video card with a new kernel seems to be impossible. On advice from the forum, I locked the 6.9 kernel and the old driver gave fine performance. I unlocked the kernel to see if the updates to 6.10 fixed the problems. Obviously, they did not.
Now, I would like to re-install 6.9 so I have useful system. Yast does not offer me that option. Please tell me how to do it manually and tell me which files I need to go with it.
@Prexy You can’t it will be long gone, even from history repositories I suspect https://download.opensuse.org/history/
Fix the driver, replace the hardware, switch back to Leap is probably the best thing.
Yeah, I couldn’t find it there. There are 100 entries on the nvidia driver topic. I didn’t see any clear method on installing the G05 which I believe will work with my GT730 card. One thing I believe is an issue is getting hdmi to work. Analog connections work on either monitor but look terrible. The card no longer sees the second monitor.
@Prexy It should be working with the latest kernel. Did you force the re-install on the nvidia kmp package so it rebuilds for the running kernel?
If everything else fails, there is still an equivalent 6.9.9 kernel by Takashi Iwai at
tiwai:/kernel:/6.9
One, among multiple possible, good moves.
Naive move, as you found out the hard way.
In such situations, instead, either keep the lock and force install another kernel, or make a lock specific to some particular kernel version.
# inxi -S
System:
Host: p5bse Kernel: 6.6.50-1-longterm arch: x86_64 bits: 64
Console: pty pts/5 Distro: openSUSE Tumbleweed-Slowroll 20240901
# zypper ll | grep kern
29 | kernel-de* | package | (any) |
30 | kernel-lo* | package | (any) |
# zypper in kernel-default
…
Problem: 1: conflicting requests
Solution 1: do not ask to install a solvable providing kernel-default.x86_64 = 6.10.8-1.1
Solution 2: remove lock to allow installation of kernel-default-6.10.8-1.1.x86_64[Update]
Choose from above solutions by number or cancel [1/2/c/d/?]
What you see above is my normal configuration and procedure. #29 you see above has existed here for many many moons. #30 has existed here since before the first installation of a longterm kernel. Zypper has never removed either lock. Zypper will not remove a lock, as it claims, when there is the wildcard in the lock. When zypper encounters a wildcard lock, solution 2 is a lie. What it should report is:
Solution 2: disregard lock to allow installation of …
“ignore” rather than disregard would also be appropriate. 
I find it very pointless to not keep Tumbleweed running with the latest updates, if your not prepared to do that, your just going to kick yourself in the foot one day or another as @Prexy discovered. Run Slowroll (lots of updates for that as well lately?) or switch back to Leap.
Thank you for trying to help me. I assume you are referring to nvidia-gfxG05-kmp-default. In my ignorance, “rebuild” means only a forced re-install. I did that but it did not change anything. I also did a forced update to G05 files that from 20240903 to 20240912. That had no effect.
I see so many references to G05 at either “550” or “560” but the only ones listed in yast are “470.” Should I be looking elsewhere for these later drivers?
Thank you for trying to help me. In my situation, it seems like locking the barn after the horse is stolen. I’ve already lost the old, working, kernel. Unless, I can figure out a way to use @OrsoBruno recommendation.
What do I do? Add the that repo and install via yast?
Simple really, even if you cannot manage to boot directly any more, but can manage to setup a chroot after booting installation or other rescue media. Once operational, do:
sudo zypper in https://download.opensuse.org/repositories/home:/tiwai:/kernel:/6.9/standard/x86_64/kernel-default-6.9.9-1.1.g8c0f797.x86_64.rpm
Why it got removed in the first place? Installing kernel does not remove previous versions. It may have been purged, for this you should really get familiar with multiversions handling in zypper which has been discussed countless times on these forums. Open /etc/zypp/zypp.conf, search for multiversion.kernels and read comments.
Just to be sure, do you have nvidia-gfxG05-kmp-default-470.256.02_k6.10.5_1-62.14.x86_64.rpm installed and it doesn’t work with 6.10 kernels? (Please mind the version, there are five of them in the Nvidia repo currently)
@Prexy and just to clarify what @OrsoBruno indicated, the only driver that works with your gpu GT730 is the G05 470 series of packages…
Thank you. Using your suggestion, I successfully installed the kernel from tiwai. Unfortunately, it did not resolve the problem. I also confirmed I had the proper nvidia driver, as specified by @OrsoBruno . The state of the monitor made it difficult to use with my old eyes. There is only 1 resolution: 1024x768. The proper resolution for the monitor is 1920x1080 and the proper resolution for the tv I use as my main monitor is 1360x768. In addition, only one monitor at a time is recognized.
This has been so frustrating that I have gone to my backup pc with an intel graphics setup. The monitors work perfectly. The kernel is 6.9.9.1-default (64 bit). I have locked it down. Don’t know if it makes a difference but it is on wayland while the failed pc is on x11.
Using recommendations from @OrsoBruno and @mrmazda , I reinstalled a 6.9.9.1 kernel and have locked it down. I did not intentionally get rid of it in favor of the 6.10 kernel. I had locked down the kernels, 6.9 and 6.10, but unlocked to update 6.10 to see if my problem had been fixed. When I updated, it DID get rid of 6.9! I actually thought I had unlocked only 6.10 for updates and 6.9 would still be on my pc but that’s not how it worked. I have gone back to an older backup pc with a Mesa Intel HD Graphics 630. It works perfectly for now.
I just remembered. I am now on an older pc with an Intel processor. It’s name is spelled out in the “About this System” info page (Mesa Intel HD Graphics 630). However, before I shut off the problem pc, I noticed that the “Graphics Processor” line did not specify an nvidia card. It gave, instead, a string something like “llvm” and a string of numbers. I didn’t think this was important until I saw the specifics on the working system.
This is a common experience when the required kernel module does not load. For NVidia users who do not use NVidia’s own proprietary drivers, this is the nouveau module shipped with every TW kernel. If NVidia’s proprietary drivers are installed, or have been installed, but have not been perfectly uninstalled, then the nouveau module in the vast majority of configurations will not load, and 1024x768 on only one connected display is the typical result regardless what attached display(s) support.
Does this mean I should uninstall all nvidia and nouveau drivers and re-install?
NVidia proprietary drivers must be eradicated according to the un-installation instructions that should have been included with their installation instructions. It could be as simple as using zypper or yast to remove rpms with nvidia in their names, such as openSUSE-repos-Tumbleweed-NVIDIA. This process sometimes involves reinstitution of OEM file(s), so some form of installation or re-configuration might be required with regard to the various components whose package names include nouveau. Nouveau has multiple meanings in Linux. The nouveau kernel module is but one.