Unable to power-on a VM that uses a RTX A6000 vGPU

Unable to power-on a VM that uses a RTX A6000 vGPU

book

Article ID: CTX297175

calendar_today

Updated On:

Description

When you configure your VM to use an NVIDIA RTX A6000 card for its vGPU, the VM can fail to power on and XenCenter displays the error: "Internal Error: Can't determine the number of VFs for PCI".

Resolution

When using an RTX A6000 card for vGPU with Citrix Hypervisor, the RTX A6000 card must be in physical_display_disabled mode. By default, these cards are not shipped in this mode.

To set your RTX A6000 card to physical_display_disabled mode:

1.Contact NVIDIA to request the displaymodeselector tool
2.Use the displaymodeselector tool to change the mode of your RTX A6000 card:
    # chmod +x displaymodeselector
    # ./displaymodeselector --gpumode
    image.png

For more information, see Citrix Hypervisor :: NVIDIA Virtual GPU Software Documentation.


Problem Cause

To be used with vGPU virtualization, RTX A6000 cards must have their physical video-output ports turned off. This behavior is called displayless mode. By default,  RTX A6000 cards are provided in display-enabled mode. These cards do not work correctly with Citrix Hypervisor until their display mode is changed to displayless.

In addition to the RTX A6000 cards, there are other types of NVIDIA graphics card that might require you to change their mode to interact correctly with Citrix Hypervisor. If your graphics card is in the following list, review the NVIDIA Virtual GPU Software Documentation to check what mode the card must be in and how to change the card mode:

  • RTX A6000
  • A40
  • Tesla M60
  • M6

Additional Information