Issue with Dual Nvidia GPU Setup on Fedora

Hardware Hub GPU
, , ,
1

I have a Dell Precision 7820 that came with a Quadro RTX 4000 and i installed a 2nd Nvidia GPU RTX 3060 to get more vram for local AI training.

The RTX 3060 shows up fine but, the Quadro RTX 4000 no longer shows up and i find the following error below when i run sudo dmesg | grep -i nvidia. Is there a way to fix this?

ERROR [nvidia-drm] [GPU ID 0x00000400] Failed to allocate NvKmsKapiDevice

Things I have confirmed:

Both cards show up in the BIOS.

Both cards show up when running lspci | grep NV

04:00.0 VGA compatible controller: NVIDIA Corporation TU104GL [Quadro RTX 4000] (rev a1)
d5:00.0 VGA compatible controller: NVIDIA Corporation GA106 [GeForce RTX 3060 Lite Hash Rate] (rev a1)

Both cards can use the same NVIDIA driver I have installed which is 595.71.05.

Troubleshooting steps I have tried so far:

Reinstalling Nvidia drivers with sudo dnf reinstall akmod-nvidia -y

Adding the open driver to kernel and rebuilding akmod(Based on this answer I found on the Fedora forms)


sudo sh -c 'echo "%_with_kmod_nvidia_open 1" > /etc/rpm/macros.nvidia-kmod'
sudo akmods --kernels $(uname -r) --rebuild
2

I’ve hit similar snags on multi-GPU boxes, and the fix was usually about who gets to be the primary card. I’d try forcing the Quadro as the boot GPU in the BIOS and then setting `nvidia-drm.modeset=1` on the kernel line. If the Quadro still sulks, check `dmesg` for BAR space complaints and try toggling Above 4G decoding or Resizable BAR. Sometimes that alone clears things up.