The GPU State Retention Daemon
When installing proprietary Nvidia graphics drivers on Ubuntu, a background service called nvidia-persistenced.service is often configured to run automatically. The purpose of this daemon is to keep the Nvidia driver’s internal software state (the “persistence mode”) active even when no applications are actively using the GPU to render graphics or perform compute calculations. Normally, when the last program utilizing the GPU closes, the driver tears down its state, powering down aspects of the card to save energy. When a new program requests the GPU, the driver has to spend time reinitializing the hardware.
By keeping the GPU state “persistent,” this daemon eliminates that initialization delay, ensuring that new graphical applications or CUDA compute jobs start instantly. For dedicated AI training servers, rendering nodes, or workstations where maximum responsiveness is required for rapid-fire GPU tasks, this is beneficial. However, for a standard desktop user, a developer compiling code on the CPU, or a laptop user trying to maximize battery life, keeping the Nvidia driver fully initialized 24/7 wastes power and generates unnecessary heat. If you want your GPU to fully spin down and shed its driver state when not in active use, you should disable the persistence daemon.
How to Disable the Nvidia Persistenced Service
You can prevent the persistence daemon from initializing using standard systemctl commands.
- Open your Ubuntu Terminal.
- To stop the service immediately during your current session, run:
sudo systemctl stop nvidia-persistenced.service- To permanently disable the service so it does not start automatically on future boots, run:
sudo systemctl disable nvidia-persistenced.service- To guarantee that no other graphical dependency can accidentally trigger the service to start up again, you should mask it:
sudo systemctl mask nvidia-persistenced.serviceThe Nvidia driver will now properly tear down its state when idle, reducing idle power consumption and heat output.