Nvidia issues with bluefin-dx-asus-nvidia

Hello everyone, I am having issues with installation based on the bluefin-dx-asus-nvidia image. Target of the installation is Asus Zephyrius G14 G401QM (2021 with Nvidia GPU), previously used Fedora Workstation with asus-linux software (supergfxctl, asusctl) without issues, so I believe that hardware problems are unlikely.
I have created iso following the instructions for the latest (39) image, instalator worked without problems, but upon reboot the systems froze, without any apparent message. Later I found out these issues were caused by the missing kernel parameters for Nvidia. Since I suspected problem with the iso, I followed old instructions and installed bluefin-gts from the published iso and after successful installation (including drive encryption) I have rebased to the gts variant of bluefin-dx-asus-nvidia and the problem returned. Trying to add Nvidia parameters (https://universal-blue.org/images/nvidia/) manually I was able to boot.
Then I have use the just nvidia to set kernel parameters. Rebooting I got to the Setting up kargs (or something like that) message followed by an automatic reboot. Sadly this ended with another freeze. Inspecting the deployments from GRUB i noticed that there are two deployments: 0 without nvidia kargs, 1 with nvidia kargs. Anything (adding them via rpm-ostree, repeating just, updating via just) I have tried to get the kargs set up persistently led to the Setting up kargs message, reboot an freeze.
While the systems is working (booting deployment 1), the Nvidia GPU does not work as expected. Launching programs with dedicated GPU via Gnome does not lead to any load on GPU being displayed in Mission Center. Launching Firefox leads to a transparent window being created. Offloading from command line does not work either, see below.
Thank you in advance for any help.

> rpm-ostree status
State: idle
AutomaticUpdates: stage; rpm-ostreed-automatic.timer: no runs since boot
Deployments:
  ostree-image-signed:docker://ghcr.io/ublue-os/bluefin-dx-asus-nvidia:gts
                   Digest: sha256:c47244037ea36b27e8f73e89912cc01d6fc56a93dc23db19651b3e2b77382a73
                  Version: 38.20240224.0 (2024-02-24T17:55:25Z)

â—Ź ostree-image-signed:docker://ghcr.io/ublue-os/bluefin-dx-asus-nvidia:gts
                   Digest: sha256:c47244037ea36b27e8f73e89912cc01d6fc56a93dc23db19651b3e2b77382a73
                  Version: 38.20240224.0 (2024-02-24T17:55:25Z)
> __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears
X Error of failed request:  BadAlloc (insufficient resources for operation)
  Major opcode of failed request:  150 (GLX)
  Minor opcode of failed request:  5 (X_GLXMakeCurrent)
  Serial number of failed request:  0
  Current serial number in output stream:  36

Note: I have just noticed that similar situation with kargs getting lost was previously discussed on ublue-os Discord and was archived on answersowerflow:
Nvidia drivers stop loading after any sort of system update
Nvidia dgpu laptop problems

I’m also running into this issues where the nvidia KARGS I have set from running ujust nvidia get removed when I rebase into bluefin-dx-asus-nvidia:latest from bluefin-dx-nvidia:latest on my ROG Zephyrus G15.

When I’m just sticking to bluefin-dx-nvidia:latest my GPU is loaded just fine.

Looking through the Github I see that there’s a recently open issue (Nouveau blacklist kargs removed after reboot, even after being added manually. · Issue #902 · ublue-os/bluefin · GitHub) which also sounds similar to what we are both hitting.

It’s left me wondering if this if-statement that checks if the IMAGE_FLAVOR is “nvidia” fails and leads to the code into the else where the KARGS are removed. (i.e IMAGE_FLAVOR is set to either “asus” or “asus-nvidia” and thus failing the check)

1 Like

I would like to summarize how I solved my problem:

  • Installed system from bluefin-dx-nvidia image
  • On first boot after installation from medium entered grub and manually added boot parameters:
    rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1
  • Finished setup with GNOME
  • add kargs via ujust configure-nvidia → kargs (or with rpm-ostree kargs --append rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1)
  • The script for updating no longer removes the arguments, so they persist over updates.

This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.