Skip to content

vGPUs

Roles with access

Only administrators have access to all these features.

IsardVDI supports the use of virtual GPUs (vGPUs) for desktops and deployments. To learn in detail about the types of cards, the different profiles, and the operating modes, you can consult the following guide.

The section for configuring GPU cards and profiles is located in Infrastructure → Hypervisors, where you can find the following box:

  • To add a new graphics card, click the button and the following form will appear:

The fields to be filled out are:

  • Name: Identifier name for the GPU.
  • Description: A short description of the GPU.
  • Select a vGPU Profile: A list of available profiles based on the GPU model. The corresponding profile must be selected.
  • The Reassign physical devices button allows the system to resynchronize physical devices. This is useful for detecting newly added GPUs or updating devices that are out of sync.

  • For each graphics card, the following information columns are displayed:

    • Name: Displays the name of the GPU type.
    • Description: Description of the GPU.

    • Physical Dev: Shows the mapping path to the physical device.

      If it appears as it means the device is not being detected and, therefore, this GPU will not be functional.

    • Active Profile: Shows the profile currently active on this GPU. If none appears, it is because it has not been activated or the GPU is not operational.

      You can see the following icon next to the profile:

      • Force GPU Active Profile: This option forces the GPU to the desired profile. It is useful when changing the profile several times in a short period, which could cause the GPU to become stuck.

    • Plans: Shows whether the selected and activated plan is working correctly. If the profile appears in green, it means it is operational; if it shows a warning sign and the profile in red, it indicates the profile did not start correctly. If "None" appears in the box, it means no plan is reserved for this card.

    • Desktops started: Shows the maximum number of desktops that can be started with GPU and the number of users occupying the available slots.

    • Brand, Model, Architecture, Memory: Information about the GPU: vendor, model, architecture, and available memory.
    • Action: Used to delete the GPU from the list registered on the platform.
Delete GPU

A GPU is a bookable resource that can be assigned to desktops. Deleting a GPU will unassign it from all user desktops and remove all related bookings.

GPU table with a notification showing "Delete GPU. Do you really want to delete this gpu?"

The system will show a warning detailing the elements that will be deleted or unassigned, as well as the affected users:

GPU deletion modal informing users about changes to their plans, bookings, desktops, and deployments

Disabling an individual GPU profile by unchecking its box will generate a similar warning.

Warning

If the GPU profile to be deleted is also present on another active GPU, its bookings and desktop assignments will remain intact. To check the profiles of each GPU, click the Detail button on the corresponding row.

Warning

If a desktop has an active booking at the time of deletion, the GPU cannot be deleted. You must stop the desktop and remove the booking, or wait for the booking time to end.

Notify users

If the Notify users that their bookings will be deleted box is checked, users will be notified by email about the deletion of their GPU bookings and desktop assignments.

The default email notification looks like this:

Email content warning the user about GPU deletion

The email content can be modified in Administration > Config > Notifications > Notification templates, listed as "Deleted GPU Profile".

Warning

Users without a valid email address linked to their profile will not be notified.

For each graphics card, you can select which profiles you want to enable so that they are available when splitting its resources.

Depending on the selected profile, the number of users who can use the GPU will vary: the higher the profile, the fewer users will be able to share it, but they will have greater performance capacity.

The profiles configured here will be the ones that, in the Planning section, can be selected to create a specific plan for the graphics card.