@suse_rasputin depends what it needs to compile, like I indicated, install the driver and test all is good. Then review to see what you need as per the whatever, I suspect it will. I use lc0 here (chess engine).
1 Like
Additional question: Is a new NVIDIA driver necessary only for a new major kernel version (e.g. 5.6 ā 5.7 transition) or for each and every minor kernel update?
Nvidia driver versions and kernel versions are completely independent.
malcolmlewis Administrator 2d
@suse_rasputin no, you can download/install/update a new Nvidia driver while the system is running, it will then be available and in use on next reboot.
For a new kernel, yes a manual re-install is required of the driver.
For example I useā¦
confused
@suse_rasputin why? If there is a new kernel, you need to rebuild the driver for the kernel. If there is a new Nvidia driver you can install the new driver and it will rebuild and be available at next boot with the running kernel.
Itās all about where you are with the ārunningā kernel, after the kernel is updated, itās still not running, so you reboot into the running kernel and install the driver for that kernelā¦
OK, do I have to ārebuiltā the driver for each (minor) new kernel update or only for major kernel updates (e.g. 5.6 ā 5.7)? If I get you correct, malcolm, each minor kernel update requires a ārebuiltā (execution of the .run file, correct?).
Better to update a TW with āhardwayā NVIDIA driver only once a month or so, maybe?
@suse_rasputin I do it for every kernel update, major or minor, but it should drop to weak updates and keep workingā¦ when it does stop working, then could rebuildā¦ Once you have done it a few times, use the history and !<n> to run or have a scriptā¦
@suse_rasputin You can set the DKMS to start at boot if youāre worried about building the driver every new kernel. This way you avoid building the .run driver manually. DKMS was unreliable for sometime but the new nvidia .run driver works well now with DKMS. Downside of this is DKMS is a bit slower to rebuild than reinstalling the driver manually.
1 Like
You apparently misunderstand how kernel drivers work.
Driver must be present for each installed kernel. There is no such thing as āmajorā or āminorā kernel updates. Each kernel version is equal.
If you use NVIDIA installer or DKMS you need to compile driver for each installed kernel. If you use SUSE RPM, it checks whether the existing driver is compatible with new kernel and tries to reuse the existing driver without rebuilding. But it still places reference to this driver in the new kernel version directory, otherwise it will not be found by new kernel.
2 Likes
ehhmmm, nope, I have no under standing at all. Came to Linux because so many things ājust workāā¦ 
@suse_rasputin The nvidia 550.54.14.run failed to build on kernel 6.7.7-1 after the snapshot today on my side. I post it here as a precaution.
I am using the beta 550.40.07.run for now.
The Nvidia rpm nvidia-drivers-G06-550.54.14-20.1 works fine with kernel-default-6.7.7-1.1
@conram what @hui said, my test setup with a Quadro K620 (offload) is using the rpm and rebuilt fineā¦
My desktop with the run file also built fineā¦
switcherooctl -g 2 pinxi -SGz
System:
Kernel: 6.7.7-1-default arch: x86_64 bits: 64
Desktop: GNOME v: 45.3 Distro: openSUSE Tumbleweed 20240302
Graphics:
Device-1: Intel Xeon E3-1200 v2/3rd Gen Core processor Graphics driver: i915
v: kernel
Device-2: Intel DG2 [Arc A380] driver: i915 v: kernel
Device-3: NVIDIA GM107GL [Quadro K620] driver: nvidia v: 550.54.14
Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.4 driver: X:
loaded: modesetting,nvidia unloaded: fbdev,vesa dri: crocus,iris gpu: i915
resolution: 1920x1080~60Hz
API: EGL v: 1.5 drivers: crocus,iris,nvidia,swrast
platforms: gbm,x11,surfaceless,device
API: OpenGL v: 4.6.0 compat-v: 4.2 vendor: nvidia mesa v: 550.54.14
renderer: Quadro K620/PCIe/SSE2
switcherooctl pinxi -SGz
System:
Kernel: 6.7.7-1-default arch: x86_64 bits: 64
Desktop: GNOME v: 45.3 Distro: openSUSE Tumbleweed 20240302
Graphics:
Device-1: NVIDIA GP104GL [Tesla P4] driver: nvidia v: 550.54.14
Device-2: NVIDIA TU117GLM [Quadro T400 Mobile] driver: nvidia v: 550.54.14
Device-3: NVIDIA GK208B [GeForce GT 710] driver: vfio-pci v: N/A
Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.4 driver: X:
loaded: nvidia gpu: nvidia,nvidia-nvswitch resolution: 1: 1920x1080~60Hz
2: 1920x1080~60Hz 3: 1920x1080~60Hz
API: EGL v: 1.5 drivers: nvidia,swrast
platforms: gbm,x11,surfaceless,device
API: OpenGL v: 4.6.0 compat-v: 4.5 vendor: nvidia mesa v: 550.54.14
renderer: Tesla P4/PCIe/SSE2
API: Vulkan v: 1.3.275 drivers: N/A surfaces: xcb,xlib
@malcolmlewis Hmmnn. there must be a hidden problem here. When I run the .run driver as I mentioned on my previous post the screen just turns black with no warning from the nvidia installer. I will try it on another tumbleweed once I do the dup if it will work.
Quoting my own post, there is really a problem, it build on another tumbleweed after zypper dup a while ago. Iām running it now.
Is it an option to keep the NVS 310 for the machine itself and have the Geforce only forwarded to a VM (Virtualbox?) for the machine learning?
@suse_rasputin sure thatās always an option (vfio-pci), never used virtualbox, but should work. Or maybe distrobox(?).
Finally and in the endā¦
Swapped cards and now have in PCI-Info Centre
03:00.0 VGA compatible controller: NVIDIA Corporation GA104 [GeForce RTX 3060] (rev a1) (prog-if 00 [VGA controller])
Kernel driver in use: nvidia
Kernel modules: nouveau, nvidia_drm, nvidia
03:00.1 Audio device: NVIDIA Corporation GA104 High Definition Audio Controller (rev a1)
OK? I installed from repos for the time being. Not in adventure mode on this machine. Will have to find a newer Dell Precision for playing around with in the futureā¦
