The nvidia drivers installed ok and lsmod shows nvidia drivers.
When I try to install cuda I get: error: Package/capability ‘akmod-nvidia’ is already requested
nvcc and nvidia-smi show nothing. The Nvidia X Server GUI show the Nvidia M40 24GB
What am I missing?
I don’t know if it matters but I have an AMD Radeon RX 7600 installed to use with LM Studio and this application does not require that I install ROCm so ROCM is not installed.
I hope I formatted this post right. Thanks in advance.
It appears those instructions are intended to be used as
sudo rpm-ostree install akmod-nvidia xorg-x11-drv-nvidia
OR
sudo rpm-ostree install akmod-nvidia xorg-x11-drv-nvidia-cuda #optional if using nvidia-smi or cuda
I would note also that dnf simply reports one package is already installed and continues to install the second. It seems rpm-ostree is not that forgiving and cannot handle an attempt to install a package that is already installed.
To me that seems a problem with rpm-ostree coding.
I agree that rpm-ostree is not as feature-rich as dnf. The syntax of the error message can be somewhat misleading too (error: Package/capability '<packageName>' is already requested).
On the other hand, the rpm-ostree install command doesn’t offer a confirmation dialog in the way dnf install does, in which case it’s probably better for the command not to be executed, since the user has no option to review a possibly wrong command syntax.
The process is also more time-consuming, since installing a package via rpm-ostree means overlaying that package and its missing dependencies on top of the current deployment and creating a new local image.
Having said that, installing packages on atomic desktops by overlaying them with rpm-ostree install is the exception, so a possibly less user-friendly interface has lower impact than dnf install on traditional systems. For this reason, I also expect a slower pace of adding new features to rpm-ostree in comparison to dnf.