I’m trying to avoid creating a larger problem for myself, so I figured I’d call in help. Still fairly novice with Linux (especially Fedora), but I’m determined to finally migrate from Windows.

I’m running the XFCE spin of Fedora with a GeForce RTX 5070 Ti and Samsung Odyssey G9. I kept my Win11 partition during migration and confirmed the monitor setup still works in windows, just not Fedora. I’m currently typing this on my Fedora install, on a second monitor I set up to troubleshoot (HDMI on the 5070).

“rpm -qa | grep -i nvidia | sort” returns the following: akmod-nvidia-615.71.09-3.fc44.x86_64 kmod-nvidia-7.2.8-200.fc44.x86_64-615.71.09-3.fc44.x86_64 nvidia-gpu-firmware-20260916-1.fc44.noarch nvidia-modprobe-615.71.09-1.fc44.x86_64 nvidia-persistenced-615.71.09-1.fc44.x86_64 nvidia-settings-615.71.09-1.fc44.x86_64 xorg-x11-drv-nvidia-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-cuda-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-cuda-libs-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-kmodsrc-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-libs-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-power-615.71.09-3.fc44.x86_64 xorg-x11-drv-nvidia-xorg-libs-615.71.09-3.fc44.x86_64

“nvidia-smi” returns the following: NVIDIA-SMI 615.71.09 KMD Version: 615.71.09 CUDA UMD Version: 13.4

Did I mess something up in my install? I followed Fedora’s official instructions on Nvidia driver installation (sudo dnf install akmod-nvidia -y). As always, I appreciate the help. While frustrating as a start, I’m excited about making the leap.

  • Trucule@lemmy.world
    13·
    2 days ago

    Set the DisplayPort version in the G9’s on-screen menu to 1.2, replug the cable, and check whether xrandr shows that DP output as connected.

    If it still shows disconnected, replug the cable and right after run:

    sudo journalctl -k -b | grep -i -E 'nvidia|drm'

    Look for link training or EDID errors on that connector.

    Also run cat /sys/module/nvidia_drm/parameters/modeset and confirm it returns Y.

    Drafted with AI.

    • non_burglar@lemmy.world
      61·
      2 days ago

      Booooo.

      Prime example of why you don’t ask an LLM for actual information.

      1. Samsung Odyssey G9 cannot drop to below v1.4, not to mention any display port mode over 1.4 should understand mode, so this isn’t necessary.
      2. There is no ‘Y’ in /sys/module/<some graphics driver>/modeset, it is 0 for disabled, 1 for enabled, and 2 for headless.

      I know you think you’re helping, but it’s just muddying the water.

      STOP ASKING LLMS FOR ACCURATE INFORMATION.

      • Trucule@lemmy.world
        1·
        1 day ago

        False. Samsung’s own troubleshooting page for the OLED G9 tells you to go to Menu > System > DisplayPort Ver. > 1.2 or 1.1, then reconnect the cable when the host doesn’t negotiate DP 1.4 cleanly. Users on the Neo G9 report the same OSD option, having dropped the DP setting to 1.2 and 1.1 (OSD) while troubleshooting a no-signal problem. Dropping the version is a standard troubleshooting step. At 1.2 the monitor stops requiring DSC and HBR3, so if link training is the problem the port will often show up as connected even though 240Hz won’t be available.

        And nvidia_drm.modeset is a kernel bool parameter, and sysfs renders bools as Y or N.

        • non_burglar@lemmy.world
          1·
          6 hours ago

          False.

          You said that with your whole chest, and I respect that, but no:

          Samsung’s own troubleshooting page for the OLED G9 tells you to go to Menu > System > DisplayPort Ver. > 1.2 or 1.1, then reconnect the cable when the host doesn’t negotiate DP 1.4 cleanly.

          Source? Please post a link, this menu does not appear on my G9. I’m guessing this is because Display Port 1.2 only supports 3840×2160 (16:9) at 60Hz max, and the G9 Odyssey screen is 5120x1440 (32:9) resolution at 240Hz, so the geometry simply wouldn’t work to force 1.2, the aspect ratio would be insane.

          Also, sysfs output is not what was posted by the LLM, you might be thinking of /proc value returns. For my modeset (kernel config file, not kernel parameter in memory):

          $ sudo cat /sys/module/i915/parameters/modeset 
-1
          • Trucule@lemmy.world
            1·
            37 minutes ago

            https://www.samsung.com/tw/support/computing/about-49-inch-odyssey-oled-g9-display-troubleshooting (OLED G9)

            Under the no-picture item it says to check whether the PC supports standard DP 1.4, and if it doesn’t, to set Menu > System > DisplayPort Ver. to 1.2 or 1.1 and reconnect the cable. The menus are probably different between the Neo and OLED models.

            Your ‘-1’ comes from the Intel driver, where modeset is an integer and -1 means the driver default. OP has an RTX 5070 Ti, and ‘nvidia_drm’ modeset is a boolean, so ‘/sys/module/nvidia_drm/parameters/modeset’ reads ‘Y’ or ‘N’. Files under ‘/sys/module/*/parameters’ show the value the loaded module is running with.

  • [object Object]@lemmy.ca
    1·
    3 days ago

    If you go into the thunderbolt control, have you approved the monitor?

    Most of the issues with thunderbolt on Linux for me have been the insane/annoying approval process.

    I can’t use gnome for example because the gnome login manager uses separate permissions so I can’t approve the display before logging in. (No idea if they fixed this).

    On plasma there is a menu for this, but you may have to go into terminal or ssh into the machine, or use hdmi to enable your display then move your computer back to it.

    ETA: here’s some articles that should work pretty much the same for Fedora

    https://linuxvox.com/blog/thunderbolt-ubuntu/

    https://commandmasters.com/commands/boltctl-linux/

    • LazerDickMcCheese@sh.itjust.worksOP
      6·
      3 days ago

      Isn’t thunderbolt different than displayport? According to the internet (again, no clue what I’m doing here), Fedora should auto detect and use displayport devices without needing intervention.

      ‘xrandr’ recognizes my GPU’s outputs correctly (1 HDMI, 3 DP), but erroneously states all displayports are disconnected.

      • [object Object]@lemmy.ca
        1·
        3 days ago

        It is, usb4 and thunderbolt are (as I understand) extensions of the pcie bus or something, so they do require escalated security permissions in all operating systems.

        macOS desktops auto approve, laptops request permission, and windows does its own approval.

        Linux requires approval until you set your computer’s security settings, but I think it only approves on user level security by default (hence the gnome issue).

        I’m also not an expert, so if your device shows up in boltctl as approved it could very well be a different issue. I’ve only done this on amd.