nvidia-open-driver-G07-signed-cuda-kmp-default and nvidia-open-driver-G07-signed-kmp-default are built using the same build script and source. So they should be either both working or both broken. But I guess it is worth testing as well.
The proprietary driver rebuilds for whatever kernel-devel was last installed on the system, so as long as the kABI does not change there is no need for a new version.
For the records, just updated to 7.0.3, that driver was rebuilt, loaded and works here (but on hybrid graphics, so did never have boot problems here).
1 Like
Weird, I started writing my post on my tablet, switched to my PC before hitting post. Sorry if somehow the draft got posted (and now is edited) - that draft shouldnāt have shown up here.
This at least appears to be because nvidia_drv_video.so is failing to initialize properly. Terminal output from launching chrome:
$ google-chrome
[124609:124609:0507/162555.855237:ERROR:components/content_settings/core/browser/content_settings_pref.cc:326] Invalid pattern strings: [*.].superfish.com,*
Created TensorFlow Lite XNNPACK delegate for CPU.
[124609:124653:0507/162556.611155:ERROR:crypto/nss_util.cc:377] After loading Root Certs, loaded==false: NSS error code: -8018
Maximum number of clients reached
Maximum number of clients reached
libva error: /usr/lib64/dri/nvidia_drv_video.so init failed
[124609:124634:0507/162601.604137:ERROR:google_apis/gcm/engine/registration_request.cc:291] Registration response error message: DEPRECATED_ENDPOINT
[124609:124609:0507/162608.614508:ERROR:ui/events/platform/wayland/wayland_event_watcher.cc:47] libwayland: warning: queue 0x16c400009640 destroyed while proxies still attached:
[124609:124609:0507/162608.614587:ERROR:ui/events/platform/wayland/wayland_event_watcher.cc:47] libwayland: zwp_tablet_pad_ring_v2#4278190083 still attached
[124609:124609:0507/162608.614628:ERROR:ui/events/platform/wayland/wayland_event_watcher.cc:47] libwayland: zwp_tablet_pad_group_v2#4278190082 still attached
[124609:124609:0507/162608.614665:ERROR:ui/events/platform/wayland/wayland_event_watcher.cc:47] libwayland: zwp_tablet_pad_v2#4278190081 still attached
Thatās a new error. Maybe some packages failed to update properly because of messed up ld?
I suppose thatās possible. The indicated library is actually a link to /usr/lib64/dri/vdpau_drv_video.so, which comes from libva-vdpau-driver, and zypper shows that package as up-to-date.
I was able to work around it by starting google-chrome with --enable-features=UseOzonePlatform --ozone-platform=x11.
ETA: I do have other apps that arenāt displaying either, like slack.
@hendersj Thatās an egl error, try adding NVD_BACKEND=direct or an egl package not updated for nvidia, they should be coming from repo-non-free.
@Lioli7k Tumbleweed is a development platform, there is an expectation that users will be able to test and diagnose the issue with their hardware. If everyone just sits and waits, and is not prepared to add info to relevant bug reports nothing will happenā¦ 
Why is my Turing based GPU not affected, unsureā¦
Where do I add NVD_BACKEND=direct? Environment variable, or just as a command-line parameter somewhere?
I donāt have a non-free repo, but the non-oss repo doesnāt show anything related to egl in it when I search.
@hendersj Try inline, or add to /etc/environment
The āopenSUSE-repos-Tumbleweed-NVIDIAā service is not installed?
zypper se -si -r NVIDIA:repo-non-free
S | Name | Type | Version | Arch | Repository
---+---------------------------+---------+----------------+--------+--------------
i | libnvidia-cfg | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | libnvidia-egl-gbm1 | package | 1.1.3-11.2 | x86_64 | repo-non-free
i | libnvidia-egl-wayland1 | package | 1.1.22-57.4 | x86_64 | repo-non-free
i | libnvidia-egl-x111 | package | 1.0.5-26.2 | x86_64 | repo-non-free
i | libnvidia-gpucomp | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | libnvidia-ml | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | libOpenCL1 | package | 2.3.4-65.4 | x86_64 | repo-non-free
i | nvidia-common-G07 | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | nvidia-compute-G07 | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | nvidia-compute-utils-G07 | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | nvidia-gl-G07 | package | 595.71.05-6.1 | x86_64 | repo-non-free
i | nvidia-modprobe | package | 595.71.05-2.2 | x86_64 | repo-non-free
i | nvidia-persistenced | package | 595.71.05-2.2 | x86_64 | repo-non-free
i | nvidia-userspace-meta-G07 | package | 595.71.05-15.1 | x86_64 | repo-non-free
i | nvidia-video-G07 | package | 595.71.05-6.1 | x86_64 | repo-non-free
I pull the nvidia drivers from the repo at https://download.nvidia.com/opensuse/tumbleweed. I also download the cuda drivers from https://developer.download.nvidia.com/compute/cuda/repos/opensuse15/x86_64 (but I donāt get the video drivers from this repo). While it says itās for release 15, thatās been working for my cuda needs (and I just confirmed with ollama that thatās still working now).
I donāt believe I have any zypper services defined on this system.
@hendersj that repo is not needed for things like ollma to runā¦
time=2026-05-07T20:03:48.892-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-05-07T20:03:48.893-05:00 level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 44459"
time=2026-05-07T20:03:49.094-05:00 level=INFO source=model_recommendations.go:179 msg="model recommendations cache sleep scheduled" wait=3h28m56.738459717s consecutive_failures=0
time=2026-05-07T20:03:49.230-05:00 level=INFO source=types.go:42 msg="inference compute" id=286d62bb-12fa-6321-9522-bb4e512bba53 filter_id="" library=Vulkan compute=0.0 name=Vulkan0 description="Quadro RTX 4000" libdirs="" driver=0.0 pci_id=0000:65:00.0 type=discrete total="8.0 GiB" available="7.4 GiB"
time=2026-05-07T20:03:49.230-05:00 level=INFO source=routes.go:1897 msg="vram-based default context" total_vram="8.0 GiB" default_num_ctx=4096
Good to know. I think it might have been when I was starting out with itā¦But if everything I need is in that nvidia service repo, then Iāll look at switching it over.
Something I noticed with slack (starting that flatpak with the NVD_BACKEND setting got it running) is that I lose drag-and-drop of images into it. not a huge deal (that was broken previously for me), but definitely something Iāll have to dig into further.
@hendersj it uses vulkan for compute, so as long as it supports, will work 
Intel ARC A380;
time=2026-05-07T20:20:54.987-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-05-07T20:20:54.988-05:00 level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 46455"
time=2026-05-07T20:20:55.079-05:00 level=INFO source=types.go:42 msg="inference compute" id=8680a556-0500-0000-0600-000000000000 filter_id="" library=Vulkan compute=0.0 name=Vulkan0 description="Intel(R) Arc(tm) A380 Graphics (DG2)" libdirs="" driver=0.0 pci_id=0000:06:00.0 type=discrete total="5.9 GiB" available="4.9 GiB"
time=2026-05-07T20:20:55.079-05:00 level=INFO source=routes.go:1897 msg="vram-based default context" total_vram="5.9 GiB" default_num_ctx=4096
Iām using the nvidia container runtime IIRC, as I run ollama in a docker container - I start the docker container with --gpus=all:
time=2026-05-07T11:01:42.395Z level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-05-07T11:01:42.395Z level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 34309"
time=2026-05-07T11:01:42.649Z level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 39599"
time=2026-05-07T11:01:42.754Z level=INFO source=model_recommendations.go:179 msg="model recommendations cache sleep scheduled" wait=4h2m30.787719945s consecutive_failures=0
time=2026-05-07T11:01:42.862Z level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1"
time=2026-05-07T11:01:42.862Z level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 37869"
time=2026-05-07T11:01:42.862Z level=INFO source=server.go:433 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 42293"
time=2026-05-07T11:01:42.998Z level=INFO source=types.go:42 msg="inference compute" id=GPU-a564fdb4-58d7-6652-c0e2-8ee8f045edc2 filter_id="" library=CUDA compute=8.6 name=CUDA0 description="NVIDIA GeForce RTX 3090 Ti" libdirs=ollama,cuda_v13 driver=13.0 pci_id=0000:c1:00.0 type=discrete total="24.0 GiB" available="21.9 GiB"
@hendersj then as long as the driver (kmp) is installed and running on the system, the container will find and use, thatās all that I had installed on Leap 16.0, aside for the GPU being a Tesla P4, Pascal based compute nodeā¦ plus I used k3sā¦
1 Like
I use openSUSE Tumbleweed for gaming. My PC has Nvidia RTX 2070 Super. After upgrading to kernel 7.0.*, I saw green screen. I didnāt have snapshots because I disabled that feature.
Iāve found a perfect solution to this issue. I recommend everybody to follow the following steps:
-
If your system does not even boot, you see a green or black screen and you canāt switch ttys, boot the prev. kernel (6.x.x) from the boot menu. If you donāt have a prev. kernel anymore, hit āeā on the 7.0 kernel on the grub menu, find the line that starts with ālinuxā and add the following to the end:
systemd.unit=multi-user.target nomodeset
and then press Ctrl+X from linux line to boot the kernel with those extra params. -
Your system will boot into a terminal. Delete ALL nvidia packages:
zypper remove cuda-compat* cuda-drivers* libnvidia-* nvidia-* -
Install the LTS kernel:
zypper install kernel-longterm -
Reboot and select the LTS kernel from the grub menu. It is 6.18.*
-
Remove the default kernel (7.0).
zypper rm kernel-default -
Install G06 NVIDIA drivers. DON"T install G07!!! It is too new. It causes issues!!!
zypper in nvidia-open-driver-G06-signed-kmp-longterm -
Reboot.
For my part (with a 2080 Ti) it worked fine just installing nvidia-open-driver-G07-signed-kmp-longterm and booting the longterm kernel, without needing to uninstall anything.
I tried G07, but my games didnāt work for some reason. Not sure why. When I launched them via Wine, there was some strange message that my card or driver was not detected. Maybe vulkan drivers didnāt work or I installed some wrong packages.
I narrowed it down to refresh rate / early modeset behavior.
On my RTX 4090 + KDE Wayland + G06 proprietary:
- 1920x1080@60 boots reliably
- 2560x1440@120/144 often hangs at early display init
- once Plasma is running, 144 Hz works perfectly stable
This suggests the issue is not general GPU instability, but a problem during early DRM/NVIDIA link training or modeset at high refresh rates.
The monitor does not lose signal completely ā it stays lit but displays black.
Happens on both DP and HDMI.
Problem persists through 7.0.1 , 7.0.2, and since today 08.05.26 , 7.0.3.
My solution at this moment is to force grub to switch the disply mode to 1080p:
ā¦rd.driver.blacklist=nouveau video=DP-2:1920x1080@60"
This is a workaround that seems to work with me. a simple shell script switches my screen back to 1440p@144hz via a shell script. basically telling kdoctor to switch to mode1 10s after boot.
not very elegant but keeps the blackscreen issue at bay for now.
2 Likes
I got it to work by switching to long term kernel and driver.
Operating System: openSUSE Tumbleweed 20260506
KDE Plasma Version: 6.6.4
KDE Frameworks Version: 6.25.0
Qt Version: 6.11.0
Kernel Version: 6.18.26-1-longterm (64-bit)
Graphics Platform: Wayland
Processors: 12 Ć AMD Ryzen 5 4600H with Radeon Graphics
Memory: 8 GiB of RAM (7.6 GiB usable)
Graphics Processor 1: NVIDIA GeForce GTX 1650 Ti/PCIe/SSE2
Graphics Processor 2: NVIDIA GeForce GTX 1650 Ti/PCIe/SSE2
Manufacturer: LENOVO
Product Name: 82B5
System Version: Lenovo Legion 5 15ARH05
kernel-longterm-6.18.26-1.1
kernel-longterm-devel-6.18.26-1.1
libnvidia-egl-gbm1-1.1.3-11.2.x86_64
libnvidia-egl-gbm1-32bit-1.1.3-11.1.x86_64
libnvidia-egl-wayland1-1.1.22-1.4.x86_64
libnvidia-egl-wayland1-32bit-1.1.22-57.2.x86_64
libnvidia-egl-x111-1.0.5-26.2.x86_64
libnvidia-egl-x111-32bit-1.0.5-26.1.x86_64
libnvidia-gpucomp-G06-32bit-580.159.03-49.1.x86_64
libnvidia-gpucomp-G06-580.159.03-49.1.x86_64
nvidia-common-G06-580.159.03-49.1.x86_64
nvidia-compute-G06-32bit-580.159.03-49.1.x86_64
nvidia-compute-G06-580.159.03-49.1.x86_64
nvidia-compute-utils-G06-580.159.03-49.1.x86_64
nvidia-gl-G06-32bit-580.159.03-49.1.x86_64
nvidia-gl-G06-580.159.03-49.1.x86_64
nvidia-modprobe-580.159.03-26.2.x86_64
nvidia-open-driver-G06-signed-kmp-longterm-580.159.03_k6.18.26_1-2.1.x86_64
nvidia-open-driver-G06-signed-kmp-meta-580.126.18-50.1.x86_64
nvidia-open-driver-G06-signed-longterm-devel-580.159.03-2.1.x86_64
nvidia-persistenced-580.159.03-2.2.x86_64
nvidia-video-G06-32bit-580.159.03-49.1.x86_64
nvidia-video-G06-580.159.03-49.1.x86_64
nvidia-xconfig-595.71.05-2.2.x86_64
2 Likes