NVIDIA Driver Installation

Update system

sudo dnf upgrade

Wait for update completion.

Verify driver version

Check the available NVIDIA driver version:

dnf info akmod-nvidia

Install NVIDIA driver

Install the driver with CUDA support:

sudo dnf install akmod-nvidia xorg-x11-drv-nvidia-cuda xorg-x11-drv-nvidia-libs.i686

This installs the driver, CUDA libraries, and build dependencies (about 1 GB).

• akmod-nvidia - NVIDIA driver using akmods for automatic kernel module building and signing
• xorg-x11-drv-nvidia-cuda - CUDA support and driver utilities (includes Wayland support)
• xorg-x11-drv-nvidia-libs.i686 - 32-bit NVIDIA libraries (needed for Steam/Proton)

Build kernel modules

akmods automatically builds and signs kernel modules during installation. To manually trigger a rebuild:

sudo akmods --force
sudo dracut --force

This process may take 5-10 minutes.

Verify kernel modules

Check that kernel modules were built:

ls /lib/modules/$(uname -r)/kernel/drivers/video/

The NVIDIA modules should be present.

Enroll MOK signing key

If Secure Boot is enabled, import the akmods signing key and set a password for the enrollment prompt:

sudo mokutil --import /etc/pki/akmods/certs/public_key.der

MOK enrollment on next boot

sudo reboot

During boot, the MOK Management screen (blue screen) will appear:

1. Select “Enroll MOK”
2. Select “Continue”
3. Select “Yes”
4. Enter the password you set in the previous step
5. Reboot

The system will boot normally after MOK enrollment.

Final reboot

sudo reboot

The NVIDIA driver will now load correctly.

Enable NVIDIA power management services

Enable NVIDIA power services for better suspend/resume behavior and power management:

sudo systemctl enable nvidia-hibernate.service nvidia-suspend.service nvidia-resume.service

What these services do:

• nvidia-hibernate.service - Properly saves GPU state before hibernation
• nvidia-suspend.service - Manages GPU state during system suspend
• nvidia-resume.service - Restores GPU state after resume

These services prevent GPU state issues after suspend/resume cycles.

Important: Do NOT enable nvidia-powerd; mask it permanently

The nvidia-powerd.service manages NVIDIA Dynamic Boost, which shifts extra wattage (~5-15W) from the CPU to the GPU during heavy GPU loads. While useful on Intel-based laptops, it conflicts with AMD ATPX power management on the Zephyrus G16 and causes soft lockups and “GPU has fallen off the bus” errors.

On this laptop, GPU power is managed via ATPX (AMD-driven via ACPI). The NVIDIA suspend/hibernate/resume services handle power states correctly without nvidia-powerd.

What you lose by disabling it: Minimal. Slightly fewer FPS during heavy GPU workloads. The ~5-15W Dynamic Boost is not worth the instability on AMD ATPX hardware.

Disable and mask permanently:

sudo systemctl disable nvidia-powerd.service
sudo systemctl stop nvidia-powerd.service
sudo systemctl mask nvidia-powerd.service

Masking creates a symlink to /dev/null, preventing any process (including NVIDIA driver updates via dnf) from re-enabling the service.

If you want to try re-enabling it later (e.g., after a kernel or driver update that may fix the ATPX conflict):

sudo systemctl unmask nvidia-powerd.service
sudo systemctl enable --now nvidia-powerd.service

Reference:

• NVIDIA Power Management Documentation

Verify NVIDIA driver

After rebooting, check the driver status:

nvidia-smi

You should see the NVIDIA driver and CUDA versions listed.

Check loaded kernel modules

lsmod | grep nvidia

If the modules are listed, the driver is loaded and functional.