Today’s update broke the graphical interface. My system boots into a character terminal. I believe it might have to do with an update of nvidia-open-driver-G06-signed-kmp-default to 550.76 for kernel 6.8.6_1-1.1. When I boot into kernel 6.8.5_1-1.5 which uses version 550.67, I get to the graphical interface with no problem. I’m running a NVIDIA GeForce RTX 3060. This is a guess on my part. However, my boot messages for the 6.8.6 version of the system ends at:
In my case, display resolution fell down to 1024x768, and got stuck. “Displays” setting withing gnome-setting does not allow any change. Someone needs to fix nvidia-open-driver-G06… immediately. I cannot do anything in the display resolution of 1024x768.
Thanks.
FWIW on my test system I run the proprietary version via rpm, this is still at 550.67, on this system I use proprietary version via the run file 550.76 (still use a Tesla P4) and it’s all fine.
First we misuse developers for the first line help desk and then we complain that developers are too slow to do what they are responsible for - to develop …
@Android_Gynous The OP is using the open Nvidia driver from the oss repository, I suspect a transient issue in your locale as no issues here (or yesterday) with the nvidia repo.
What I see is that there are multiple versions of the kernel firmware. However, the drivers are all 550.67. 550.76 is only available for the kernel-firmware. I’m guessing that the .76 firmware is used with 6.8.6 but the .67 version is used with the previous kernel 6.8.5 and that might be why graphics run with 6.8.5 but not with 6.8.6.
I’ve always used the proprietary drivers as free drivers were not available - hence the Nvidia repository. The utility files were on the system with the 6.8.5 kernel and it works fine. They don’t seem to be available on any other repository. The only Nvidia flie that was updated in the zypper dup that hosed the system was nvidia-open-driver-G06-signed-kmp-default 550.67_k6.86_1. First it was deleted and then reinstalled. Right now I have version 1_1. The fact of that one change was, I think, the problem. Before that update I had 6.8.6 and it ran the graphic interface with no problem. I’d go back, but there seems to be no file available.
@bearymore Ahh, ok, well you need to remove the open drivers then add a lock for nvidia-open-driver-G06-signed-kmp-default. And you have the double up on repos…
nvidia’s open driver only contains a kernel module so the proprietary user space driver packages are actually required to be used together…
imo the real issue is that the nvidia repo isn t always update in time.
so the version of nvidia-open-driver-G06-signed-kmp-default could be out of sync with other driver components. i ve noticed this happening on my system few times before.
idk why zypper does not block this invalid upgrade, thus i had to manually check this before every dup. somewhat annoying.
Malcom, I think you may have solved the issue. I just eliminated all of the nvidia kernel firmware not from the nvidia directory, and added nvidia-driver-G06-kmp-default from the nvidia repository and eliminated the open one from the main repository. 6.86 came up with the graphical interface. However, I think lictex is correct, too. The open-driver-G06-signed-kmp-default is version 550.76. Everything from the nvidia repository is version 550.67.
I am nervous that I eliminated too much, namely all the kernel-firmware-nvidia-gspx-G06 files and the kernel-firmware-nvidia files. Here is what’s left. I hope this is a correct configuartion.
@bearymore AFAIK the firmware is part of the proprietary install so those rpms are only needed to supplement the open driver. It also needs manual activation…
So I have two gpu’s one old and one new (well I actually have three but that’s for pci-vfio)
So I install manually via the run file and can see;
cat /proc/driver/nvidia/gpus/0000\:02\:00.0/information
Model: Unknown
IRQ: 61
GPU UUID: GPU-0f2a639c-3f7b-acc7-9d05-115ef6dcb51f
Video BIOS: 86.04.8c.00.10
Bus Type: PCIe
DMA Size: 47 bits
DMA Mask: 0x7fffffffffff
Bus Location: 0000:02:00.0
Device Minor: 0
GPU Excluded: No
cat /proc/driver/nvidia/gpus/0000\:03\:00.0/information
Model: NVIDIA T400
IRQ: 62
GPU UUID: GPU-41dbedb2-a46d-4f41-4ac3-8a2dbdfea5b6
Video BIOS: 90.17.76.00.0b
Bus Type: PCIe
DMA Size: 47 bits
DMA Mask: 0x7fffffffffff
Bus Location: 0000:03:00.0
Device Minor: 1
GPU Firmware: 550.76 <====
GPU Excluded: No
But have manually added to a conf file options nvidia NVreg_EnableGpuFirmware=1
Thank you Malcolm. I have two questions left. First, is there any reason to prefer the open version of the drivers over the proprietary version or vice versa? Second, what conf file does options nvidia NVreg_EnableGpuFirmware=1 belong in and does it noticeably improve performance?
@bearymore If I had a later offload gpu (I have a Tesla P4) I would give it a whirl to see. I ran it with just the Quadro T400 when if first came out and had power management issues.
My T400 is just for driving displays as opposed to a hard workout, so it probably does improve performance using the gsp firmware. You would need to do some benchmarking to see, eg blender-benchmark, maybe memtest vulkan.
I use a `/etc/modprobe.d/50-nvidia.conf file containing;
blacklist nouveau
options nouveau modeset=0
##Enable NVIDIA GSP Firmware
options nvidia NVreg_EnableGpuFirmware=1
##Power Management
options nvidia NVreg_DynamicPowerManagement=0x02
I also run persistent power with a systemd service…
If it’s an issue, the only way to reach the correct folks is via a bug report