I saw this issue today when using toolbx fedora 41 with kde F41 where everything works as should, but not dGPU Nvidia…
I was spending some minutes to figure out where is the issue and not sure is there already automated nvidia-container-toolkit on host where it actually enables host Nvidia drivers to containers to be used.
The what I was doing I had to enable rpm-fusion repos and install Nvidia drivers, but on container I can’t install drivers with CUDA all CUDA drivers and packages failed.
Also I had to install nvidia-container-toolkit on container too to get dGPU actually working on graphics demanding apps DaVinci Resolve was on this test and VScode was another where I need to have dGPU for 3D development and rendering
Host system is F41 kinoite with optimus hybrid setup and all apps are flatpaks or inside toolbx as is development environments.
Inside toolbx terminal it still shows nvidia-smi etc when using system terminal, but if I use bscode terminal it is more containerized example is VScode is installed as flatpak and connected to toolbx container where I do all development creating directories and cloning repos shows only on container when using VScode terminal, but if I use system terminal inside toolbx and creating directories and cloning repos those actually shows on host
but i got it running by downloading the nvidia Drivers directly from nvidia and installed it inside the toolbx with this commands: wget https://us.download.nvidia.com/XFree86/Linux-x86_64/565.57.01/NVIDIA-Linux-x86_64-565.57.01.run
Please note: the version downloaded must match the version of the driver installed on the host. if you update the drivers on the host, you must also update them in the Toolbx.