Hello,
I am experiencing issues running Vulkan applications in my Linux guest VM. Both Vulkan and OpenGL always fall back to LLVMPipe instead of using the NVIDIA vGPU. I am using a Quadro RTX 6000 and currently testing the vGPU during the trial period.
So far, most components seem to work correctly: the NVIDIA vGPU Manager is installed on the host, nvidia-smi reports the vGPU without errors, and the guest system loads the NVIDIA drivers properly. The NVIDIA Settings panel also shows all supported Vulkan and OpenGL versions. However, Vulkan/OpenGL itself does not work.
Another issue I am encountering is that even when the Linux guest is completely idle, the vGPU remains at a constant 90–100% compute load, causing the GPU to run very hot. Because I am running a headless setup, I am also streaming the display using x11vnc, but the desktop is barely usable and runs at very low FPS even while idle. This behavior occurs only in the Linux guest. A Windows guest works perfectly, including Vulkan/OpenGL support and normal GPU usage.
I am looking for recommendations regarding a suitable Linux distribution, kernel version, and NVIDIA driver version, or any advice on how to resolve these issues.
My system setup:
Distro: Ubuntu 24.04 LTS
Kernel: 6.8.
NVIDIA Driver: 550.163.02
GPU: Quadro RTX 6000
CPU: Ryzen 9 9950X3D
RAM: 128 GB DDR5 5600 MHz