Fedora – Chris Paquin

First determine what driver is installed. Below we can see that we are running the “nouveau” driver and not the Nvidia proprietary driver.

$ sudo lspci -n -n -k | grep -A 2 -e VGA -e 3D
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GA106 [GeForce RTX 3060 Lite Hash Rate] [10de:2504] (rev a1)
	Subsystem: PNY Device [196e:138f]
	Kernel driver in use: nouveau

Add rpm fusion repos.

$ sudo dnf install https://mirrors.rpmfusion.org/free/fedora/rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm https://mirrors.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm

Dnf update.

$ sudo dnf update -y

Install kernel-devel which provides kernel headers and makefiles required to build kernel modules.

$ sudo dnf -y install kernel-devel

Install akmod-nividia, which checks if there are any missing kernel mods at boot time and will if so, will build them.

$ sudo dnf -y install akmod-nvidia

Now install the CUDA driver for xorg-x11-drv-nvidia.

$ sudo dnf -y install xorg-x11-drv-nvidia-cuda

Enable/start Nvidia nvidia-persistenced.service.

sudo systemctl enable --now nvidia-persistenced.service

Reboot.

Now we check to make sure that the driver is loaded.

$ sudo lspci -n -n -k | grep -A 2 -e VGA -e 3D
[sudo] password for cpaquin: 
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GA106 [GeForce RTX 3060 Lite Hash Rate] [10de:2504] (rev a1)
	Subsystem: PNY Device [196e:138f]
	Kernel driver in use: nvidia

Run “nvidia-smi” to show the exact driver revision. Below we have 565.77. And this driver is compatible with CUDA Version: 12.7.

Nvidia Cuda Toolkit Install

We now need to install the nvidia-cuda-toolkit. We will get this directly from nvidia. According to the driver/toolkit version matrix found here, we should install Cuda Toolkit 12.6.

First install the following pre-req. Nvcc will not compile on gcc14 which is default for Fedora 40.

# sudo dnf install gcc13-c++ -y

Pull down the installer

wget https://developer.download.nvidia.com/compute/cuda/12.6.2/local_installers/cuda_12.6.2_560.35.03_linux.run

Make the file executable

$ sudo chmod +x cuda_12.6.2_560.35.03_linux.run

Now install. Since we already have drivers installed, make sure that you uncheck all install options except for the toolkit itself.

sudo ./cuda_12.6.2_560.35.03_linux.run

Add the following to your .bashrc. And if you intend to run/install anything as root, you may want to add it to root’s .bashrc as well.

export PATH=/usr/local/cuda-12.6/bin:$PATH

Next add “/usr/local/cuda-12.6/lib64” to “/etc/ld.so.conf” and run “ldconfig” as root. See example below

# cat /etc/ld.so.conf
include ld.so.conf.d/*.conf
/usr/local/cuda-12.6/lib64

Now run ldconfig

# ldconfig

Now test nvcc as shown below.

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Thu_Sep_12_02:18:05_PDT_2024
Cuda compilation tools, release 12.6, V12.6.77
Build cuda_12.6.r12.6/compiler.34841621_0

Category: Fedora

Drivers and CUDA Toolkit Setup for Nvidia on Fedora 40

Nvidia Cuda Toolkit Install

Resources

Like this: