GPU Driver

During the deployment process of a Virtual Cluster a GPU driver is installed automatically. This GPU driver is tested during the release validation process. It is strongly recommended to stay with this driver version, but in some cases it can be useful or necessary to update to another driver. The GPU driver is installed from a software repository provided by the GPU vendor Nvidia and the major version is pinned so that a full system upgrade (dnf update) as well as a module upgrade (dnf update nvidia-driver) will only install a new GPU driver which has the same major version number.

In case you need to switch to a new major version of the GPU driver this can be accomplished by switching to a new version of the nvidia-driver module:

sudo dnf module switch-to nvidia-driver:<VERSION>-dkms

It is important to use the dkms version of the module since this will automatically compile required modules for the currently installed kernel.

Note

A GPU driver is only required on the "session" and "compute" instances, i.e. the command above needs to be applied to the "session" template instance or the "compute" template instance and new versions of the instance templates need to be created to get the new GPU driver for all newly created instances. Please refer to the Template Instances section to find out more about how to manage template instances.