Is this page helpful?

Updates by Release#

Updates for each release in this release family of NVIDIA vGPU software may include new features, introduction of hardware and software support, and withdrawal of hardware and software support.

Updates in Release 21.0#

New Features in Release 21.0#

  • nvidia-xid-logd support for GPU bind and unbind events on Linux with KVM and XenServer hosts

    The service is enabled by default and automatically releases its GPU reference on unbind and resumes monitoring on rebind. You no longer need to stop or disable the service before switching a GPU between vGPU and pass-through modes.

  • Windows Subsystem for Linux v2 in Windows guest VMs on Red Hat Enterprise Linux with KVM and Ubuntu

    Windows Subsystem for Linux v2 is supported in Windows guest VMs on Red Hat Enterprise Linux with KVM and Ubuntu. No other use case that relies on nested virtualization is supported.

  • Support for vGPU migration when NVIDIA CUDA Toolkit debuggers or profilers are enabled

  • Support for hardware-accelerated GPU scheduling for a vGPU

  • Support for the Microsoft Compute Driver Model (MCDM)

  • Notification to guest VMs of when a GPU reset is required

  • Console VNC support with the Wayland display server protocol on XenServer, Red Hat Enterprise Linux with KVM, and Ubuntu

  • Support for GPU performance monitoring (GPM) metrics through NVWMI

  • Miscellaneous bug fixes as explained in Issues Resolved in Release 21.0

Hardware and Software Support Introduced in Release 21.0#

  • Newly supported hypervisor software:

    • Microsoft Azure Local 12.2609.1003.7

    • Microsoft Azure Local 12.2510

    • Red Hat OpenShift Virtualization 4.22

    • Red Hat OpenShift Virtualization 4.21

    • Red Hat OpenShift Virtualization 4.20

    • Red Hat OpenShift Virtualization 4.19

  • Newly supported guest OS releases:

    • Windows 11 26H2

  • Newly supported remoting solutions:

    • Citrix Virtual Apps and Desktops version 7 2607

    • Omnissa Horizon 8 2606

Features Deprecated in Release 21.0#

The following table lists features that are deprecated in this release of NVIDIA vGPU software. Although the features remain available in this release, they might be withdrawn in a future release.

Deprecated Feature

Preferred Alternative

Additional Information

Tesla Compute Cluster (TCC) driver model

Microsoft Compute Driver Model (MCDM)

Enabling the Microsoft Compute Driver Model (MCDM)

Feature Support Withdrawn in Release 21.0#

  • Hypervisor software no longer supported:

    • Microsoft Azure Local 12.2504

    • Red Hat OpenShift Virtualization 4.16

    • Red Hat OpenShift Virtualization 4.15