ETA: Agh! Mea culpa: one of the components I replaced tracking down my hardware issue was the motherboard. I installed a new BIOS but it seems I either didn’t disable Secure Boot on this new motherboard, or it got reset to Secure. I just checked it - after penning my diatribe, naturally - and disabled it, and now the driver loads and I can use CUDA in Blender.
So evidently there’s a Secure Boot issue. I definitely disabled it on the laptop, but it’s a Dell that comes with Windows and it’s massively locked down - Dell manages to somehow update bits on the machine not under my control. So the laptop is quite possibly not operating correctly re: the 3050Ti on Fedora because of some security issue, which I’ve suspected almost from the off but can’t prove anything. This seems to be corroborating evidence of some sort but I don’t know how useful it is other than to advise people having problems to make sure that Secure Boot is disabled in the BIOS.
Of course it shouldn’t be necessary to disable Secure Boot, but I don’t know how to go about importing certificates into UEFI, etc., and given the problems I’ve had with new machinery lately it’s not something I’m in a rush to attempt (because I figure I’ll make it unusable again, or worse - unbootable: bricked.)
I’ll leave this here as a sort of pour encourager les autres kind of public humiliation a moi.
And now it’s 4:51am, my machine can use an Nvidia GPU on Fedora 37 so I’m going to call it a day!
I’ve built a new 7950X machine that was supposed to be Linux-only (it’s taken a while - I had to trouble-shoot then RMA a part that took a month), but I’m having the familiar problem of not being able to get graphics drivers working properly under any flavor of Linux - I’ve tried K/Ubuntu 22.04.1 and 22.10, Debian 11.6 and Fedora 37 with a variety of discrete GPUs - Radeon 6600 and 6800 and Nvidia RTX 3050.
I’m currently trying to get the 3050 working with Fedora 37 because Nvidia seems to be the only supported combination on Fedora (via rpmfusion - the 6800 was a no-go).
The GPUs provide the basic functionality on all platforms, but I can’t get the full functionality operational.
I’m having the same or worse results than when I tried to get Fedora 37 running with a mobile RTX 3050 Ti on a new Intel i7-12700 laptop (still can’t be found in Blender e.g.)
This is a clean installation of Fedora 37. The only modifications from stock are the Nvidia drivers from the rpmfusion repo.
Based on previous communications here, I’m seeing the following:
dnf list installed *nvidia* && dnf repolist
Installed Packages
akmod-nvidia.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
kmod-nvidia-6.1.9-200.fc37.x86_64.x86_64 3:525.85.05-1.fc37 @@commandline
nvidia-gpu-firmware.noarch 20230117-146.fc37 @updates
nvidia-persistenced.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
nvidia-settings.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-cuda.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-cuda-libs.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-kmodsrc.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-libs.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-power.x86_64 3:525.85.05-1.fc37 @rpmfusion-nonfree-updates
repo id repo name
fedora Fedora 37 - x86_64
fedora-cisco-openh264 Fedora 37 openh264 (From Cisco) - x86_64
fedora-modular Fedora Modular 37 - x86_64
rpmfusion-free RPM Fusion for Fedora 37 - Free
rpmfusion-free-updates RPM Fusion for Fedora 37 - Free - Updates
rpmfusion-nonfree RPM Fusion for Fedora 37 - Nonfree
rpmfusion-nonfree-updates RPM Fusion for Fedora 37 - Nonfree - Updates
updates Fedora 37 - x86_64 - Updates
updates-modular Fedora Modular 37 - x86_64 - Updates
lsmod | grep nvidia
produces an empty list
inxi -G produces:
Graphics:
Device-1: NVIDIA GA106 [Geforce RTX 3050] driver: nouveau v: kernel
Device-2: AMD Raphael driver: amdgpu v: kernel
Display: wayland server: X.org v: 1.20.14 with: Xwayland v: 22.1.7
compositor: kwin_wayland driver: N/A resolution: 3840x2160
API: OpenGL v: 4.5 Mesa 22.3.4 renderer: llvmpipe (LLVM 15.0.7 256 bits)
When I boot there’s a line on the initial screen telling me that the Nvidia kernel module is missing. If I lsinitrd then I notice that there is apparently no firmware for the particular Nvidia GPU, in this case the GA106. It exists on my hard drive however. I’ve tried manually dracut’ing to no avail.
results of journalctl | grep -i nvidia from the most recent boot:
Feb 07 19:39:32 fedora kernel: Command line: BOOT_IMAGE=(hd1,gpt3)/root/boot/vmlinuz-6.1.9-200.fc37.x86_64 root=UUID=3c897167-c3da-4233-90da-3dbf940be503 ro rootflags=subvol=root rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init
Feb 07 19:39:32 fedora kernel: Kernel command line: BOOT_IMAGE=(hd1,gpt3)/root/boot/vmlinuz-6.1.9-200.fc37.x86_64 root=UUID=3c897167-c3da-4233-90da-3dbf940be503 ro rootflags=subvol=root rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init
Feb 07 19:39:32 fedora dracut-cmdline[503]: Using kernel command line parameters: rd.driver.pre=btrfs BOOT_IMAGE=(hd1,gpt3)/root/boot/vmlinuz-6.1.9-200.fc37.x86_64 root=UUID=3c897167-c3da-4233-90da-3dbf940be503 ro rootflags=subvol=root rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init
Feb 08 03:39:35 fedora kernel: input: HDA NVidia HDMI/DP,pcm=3 as /devices/pci0000:00/0000:00:02.2/0000:16:00.1/sound/card1/input21
Feb 08 03:39:35 fedora kernel: input: HDA NVidia HDMI/DP,pcm=7 as /devices/pci0000:00/0000:00:02.2/0000:16:00.1/sound/card1/input22
Feb 08 03:39:35 fedora kernel: input: HDA NVidia HDMI/DP,pcm=8 as /devices/pci0000:00/0000:00:02.2/0000:16:00.1/sound/card1/input23
Feb 08 03:39:35 fedora kernel: input: HDA NVidia HDMI/DP,pcm=9 as /devices/pci0000:00/0000:00:02.2/0000:16:00.1/sound/card1/input24
Feb 08 03:39:35 fedora systemd[1]: Starting nvidia-powerd.service - nvidia-powerd service...
Feb 08 03:39:35 fedora /usr/bin/nvidia-powerd[1180]: nvidia-powerd version:1.0(build 1)
Feb 08 03:39:36 fedora systemd[1]: Starting nvidia-fallback.service - Fallback to nouveau as nvidia did not load...
Feb 08 03:39:36 fedora kernel: nouveau 0000:16:00.0: NVIDIA GA106 (b76000a1)
Feb 08 03:39:37 fedora /usr/bin/nvidia-powerd[1180]: Allocate client failed 38
Feb 08 03:39:37 fedora /usr/bin/nvidia-powerd[1180]: Failed to initialize RM Client
Feb 08 03:39:37 fedora systemd[1]: nvidia-powerd.service: Main process exited, code=exited, status=1/FAILURE
Feb 08 03:39:37 fedora systemd[1]: nvidia-powerd.service: Failed with result 'exit-code'.
Feb 08 03:39:37 fedora systemd[1]: Failed to start nvidia-powerd.service - nvidia-powerd service.
Feb 08 03:39:37 fedora audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=nvidia-powerd comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=failed'
Feb 08 03:39:37 fedora systemd[1]: Finished nvidia-fallback.service - Fallback to nouveau as nvidia did not load.
Feb 08 03:39:37 fedora audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=nvidia-fallback comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=success'
No matter what graphics card I stick in any machine, Windows finds it and runs it properly - I can use CUDA with an Nvidia GPU and OpenCL with an AMD GPU in Blender, e.g.
I’m about ready to give up on Linux for graphics. Not one of the distros I favor has worked with any of my GPUs.
One interesting thing: I did manage to get Fedora 36 running correctly on an i9-12900 machine with the RTX 3050 following the same procedure I’ve tried since to get working with Fedora 37, so as far as Fedora is concerned it seems my problem is with 37 (i’ve not tried going back to Fedora 36). Obviously others have managed to get their Nvidia GPUs to work with it but I can’t get it running on 3 different machines (2 AMD 16-core desktops and one Intel laptop), and AMD graphics seems to be even more of a lost cause on Fedora (I thought AMD GPUs were better supported on Linux?!)
Has anyone managed to get a modern Nvidia GPU (30X0 series) working with Fedora 37 on the actual hardware (not in a VM)?