Virtual GPU Software R620 Known Issues

Known issues for NVIDIA virtual GPU software and hardware on all supported hypervisors.

• Switching a Windows VM to MCDM mode unloads the guest driver
• GSP firmware cannot be disabled when NVIDIA Linux open GPU kernel modules are installed
• Task Manager reports less dedicated GPU memory than the vGPU frame buffer
• NovaCore prevents the NVIDIA vGPU Manager service from starting on Ubuntu 26.04
• Migration of multiple vGPU-enabled VMs between XenServer 9 hosts might fail
• Wayland with Ubuntu 24.04 does not work with multi-vGPU VMs
• Wayland sessions on Red Hat Enterprise Linux 10 with vGPUs on Blackwell-architecture GPUs are corrupted and unusable
• Display resolutions can’t be changed in GNOME Settings with Wayland
• Application windows disappear after display resolution change in Wayland
• Remote desktop connections fail for vGPUs with 1 GB of frame buffer when used with Wayland
• With Red Hat Enterprise Linux 10 on VMware vSphere, the VM crashes when graphical.target is set
• The dmesg command reports non-fatal asserts and warnings
• vGPUs created with a vendor-specific VFIO framework on Ubuntu do not support Console VNC
• NVIDIA vGPU software does not support Ubuntu 22.04 with an upgraded HWE kernel
• Linux graphics driver installation from a .run file fails
• GPU device is unavailable on Windows 11 VMs with more than 1 TB of memory
• Default client configuration token folder missing on Windows client VMs
• vGPU VM fails to boot with error vmiop-display unable to reserve vgpu
• NVIDIA Control Panel is not available in multiuser environments
• NVIDIA Control Panel crashes if a user session is disconnected and reconnected
• VM assigned multiple fractional vGPUs from the same GPU hangs
• CUDA profilers cannot gather hardware metrics on NVIDIA vGPU
• NVIDIA vGPU software graphics driver for Windows sends a remote call to ngx.download.nvidia.com
• Multiple RDP session reconnections on Windows Server 2022 can consume all frame buffer
• VM with multiple legacy fractional vGPUs on the same GPU fails to boot
• On NVIDIA H100, creation of multiple compute instances after deletion of existing compute instances fails
• NLS client fails to acquire a license with the error The allowed time to process response has expired
• NVIDIA vGPU software graphics driver fails to load on KVM-based hypervsiors
• With multiple active sessions, NVIDIA Control Panel incorrectly shows that the system is unlicensed
• VP9 and AV1 decoding with web browsers are not supported on Microsoft Windows Server 2019
• Ubuntu guest driver initialization fails with vGPUs and GPUs that support SR-IOV
• nvidia-smi ignores the second NVIDIA vGPU device added to a Microsoft Windows Server 2016 VM
• After an upgrade of the Linux graphics driver from an RPM package in a licensed VM, licensing fails
• After an upgrade of the Linux graphics driver from a Debian package, the driver is not loaded into the VM
• Desktop session freezes when a VM is migrated to or from a host running an NVIDIA vGPU software 14 release
• Application or vGPU VM crashes when multiple application instances are launched
• The reported NVENC frame rate is double the actual frame rate
• VM fails after a second vGPU is assigned to it
• Hypervisor host reboots when multiple cloned VMs are simultaneously powered on or migrated
• NVENC does not work with Teradici Cloud Access Software on Windows
• When a licensed client deployed by using VMware instant clone technology is destroyed, it does not return the license
• A licensed client might fail to acquire a license if a proxy is set
• Session connection fails with four 4K displays and NVENC enabled on a 2Q, 3Q, or 4Q vGPU
• Disconnected sessions cannot be reconnected or might be reconnected very slowly with NVWMI installed
• For some accounted processes, nvidia-smi incorrectly reports maximum memory use of 0 MiB
• Windows VM crashes during Custom (Advanced) driver upgrade
• VMs with vGPUs on GPUs based on the NVIDIA Ampere architecture fail to power on
• Linux VM hangs after vGPU migration to a host running a newer vGPU manager version
• NVIDIA A100 HGX 80GB vGPU names shown as Graphics Device by nvidia-smi
• Idle Teradici Cloud Access Software session disconnects from Linux VM
• GPU Operator doesn’t support vGPU on GPUs based on architectures before NVIDIA Turing
• No virtual GPU types are listed in Citrix XenCenter
• Idle NVIDIA A100, NVIDIA A40, and NVIDIA A10 GPUs show 100% GPU utilization
• NVIDIA vGPU software graphics driver fails to load after upgrade on XenServer
• Windows guest VMs with vGPUs or GPUs with large BAR memory settings fail to boot to the desktop in UEFI mode
• Guest VM frame buffer listed by nvidia-smi for vGPUs on GPUs that support SRIOV is incorrect
• Migrated VM with more than two vGPUs crashes on destination host
• VMs fail to boot on RHV 4.4
• Driver upgrade in a Linux guest VM with multiple vGPUs might fail
• NVIDIA Control Panel fails to start if launched too soon from a VM without licensing information
• VNC client session goes blank and console VNC is corrupted when the guest driver is uninstalled
• Citrix Virtual Apps and Desktops session corruption occurs in the form of residual window borders
• Omnissa Horizon clients cannot connect to a Windows 10 2004 VM with multiple displays
• Suspend and resume between hosts running different versions of the vGPU manager fails
• On Linux, a VMware Horizon 7.12 session freezes after a switch to full screen
• On Linux, a VMware Horizon 7.12 session with two 4K displays freezes
• On Linux, the frame rate might drop to 1 after several minutes
• Frame buffer consumption grows with Omnissa Horizon over Blast Extreme
• Citrix XenCenter displays misleading information about vGPU types
• Citrix Virtual Apps and Desktops connection freezes initially
• Microsoft DDA fails with some GPUs
• DWM crashes randomly occur in Windows VMs
• NVIDIA Control Panel fails to launch in a platform layer or published image
• Remote desktop session freezes with assertion failure and XID error 43 after migration
• Citrix Virtual Apps and Desktops session freezes when the desktop is unlocked
• NVIDIA vGPU software graphics driver fails after Linux kernel upgrade with DKMS enabled
• Red Hat Enterprise Linux and CentOS 6 VMs hang during driver installation
• On XenServer, all vGPUs in a VM must be of the same type
• Console VGA cannot be disabled
• Console VNC is unusable with Xorg on multiple vGPUs in a VM
• Blue screen crash occurs or no devices are found after VM reset
• Tesla T4 is enumerated as 32 separate GPUs by VMware vSphere ESXi
• Migrating a VM configured with NVIDIA vGPU software release 21.1 to a host running release 21.0 fails
• Users’ sessions may freeze during vMotion migration of VMs configured with vGPU
• Migration of VMs configured with vGPU stops before the migration is complete
• ECC memory settings for a vGPU cannot be changed by using NVIDIA X Server Settings
• Changes to ECC memory settings for a Linux vGPU VM by nvidia-smi might be ignored
• Incorrect GPU type shown for Quadro RTX 8000 GPUs in Citrix XenCenter
• NVIDIA Notification Icon prevents log off of Citrix Published Application user sessions
• Host core CPU utilization is higher than expected for moderate workloads
• H.264 encoder falls back to software encoding on 1Q vGPUs with a 4K display
• H.264 encoder falls back to software encoding on 2Q vGPUs with 3 or more 4K displays
• Frame capture while the interactive logon message is displayed returns blank screen
• RDS sessions do not use the GPU with Microsoft Windows Server as guest OS
• VMware vMotion fails gracefully under heavy load
• View session freezes intermittently after a Linux VM acquires a license
• When the scheduling policy is fixed share, GPU utilization is reported as higher than expected
• Benign warnings during Virtual GPU Manager installation or uninstallation after hypervisor upgrade
• Benign not in gzip format messages during Virtual GPU Manager installation or uninstallation
• License is not acquired in Windows VMs
• nvidia-smi reports that vGPU migration is supported on all hypervisors
• NVIDIA Control Panel Crashes in a VM connected to two 4K displays
• Hot plugging and unplugging vCPUs causes a blue-screen crash in Windows VMs
• GPU resources not available error during VMware instant clone provisioning
• Module load failed during VIB downgrade from R390 to R384
• vGPU guest VM driver not properly loaded on servers with more than 512 GB or 1 TB or more of system memory
• Luxmark causes a segmentation fault on an unlicensed Linux client
• On Linux, 3D applications run slowly when windows are dragged
• A segmentation fault in DBus code causes nvidia-gridd to exit on Red Hat Enterprise Linux and CentOS
• No Manage License option available in NVIDIA X Server Settings by default
• Licenses remain checked out when VMs are forcibly powered off
• Memory exhaustion can occur with vGPU profiles that have 512 Mbytes or less of frame buffer
• VM bug checks after the guest VM driver for Windows 10 RS2 is installed
• On XenServer 7.0, VMs unexpectedly reboot and XenServer crashes or freezes
• With no NVIDIA driver installed, XenServer misidentifies Tesla M10 cards
• vGPU VM fails to boot in ESXi if the graphics type is Shared
• GNOME Display Manager (GDM) fails to start on Red Hat Enterprise Linux 7.2 and CentOS 7.0
• NVIDIA Control Panel fails to start and reports that “you are not currently using a display that is attached to an Nvidia GPU”
• VM configured with more than one vGPU fails to initialize vGPU when booted
• A VM configured with both a vGPU and a passthrough GPU fails to start the passthrough GPU
• vGPU allocation policy fails when multiple VMs are started simultaneously
• Before Horizon agent is installed inside a VM, the Start menu’s sleep option is available
• vGPU-enabled VMs fail to start, nvidia-smi fails when VMs are configured with too high a proportion of the server’s memory.
• On reset or restart VMs fail to start with the error VMIOP: no graphics device is available for vGPU…
• nvidia-smi shows high GPU utilization for vGPU VMs with active Horizon sessions
• Video goes blank when run in loop in Windows Media Player
• Local VGA console is momentarily unblanked when Citrix Virtual Apps and Desktops changes resolution of the VM desktop
• VM bugchecks on shutdown/restart when Citrix Virtual Apps and Desktops is installed and NVIDIA driver is uninstalled or upgraded.
• Application frame rate may drop when running Citrix Virtual Apps and Desktops at 2560×1600 resolution.
• Windows VM BSOD
• Windows VM BSOD when upgrading NVIDIA drivers over a Citrix Virtual Apps and Desktops session
• XenCenter does not allow vGPUs to be selected as a GPU type for Linux VMs
• If X server is killed on a RHEL7 VM running vGPU, XenCenter console may not automatically switch to text console
• Citrix Virtual Apps and Desktops shows only a black screen when connected to a vGPU VM