Just a quick one that’s been keeping me awake lately…
We have 4 XenApp 7.6 servers running on a vSphere 5.1 host with 2x K2 cards installed and configured for pass through. I can see both the VMware sVGA adapter and NVIDIA GRID card in device manager and the NVIDIA GRID driver has been installed.
If a user logs on to an ICA session or via VMware console and runs a DX11 or OpenGL benchmark, the NVIDIA GRID is utilized automatically. We can see this using GPU-Z. However, I cannot start the NVIDIA Control Panel as there’s no monitor attached to the NVIDIA PCI device and if I run DXDIAG, or view the screen resolution settings within the Windows Server 2008R2 OS I can see the sVGA adapter is set to primary, despite the GPU workloads ‘offloading’ to the NVIDIA GRID card automatically. This is causing us an issue with the unreal engine application as it requires a dedicated GPU to install correctly but it detects the sVGA as the primary adapter.
I’ve read that by adding a PCI pass through device to a virtual machine the console would stop as it only functions using the sVGA and not the GIRD card. However, console session through VMware continues to work.
I have tried connecting to the XenApp servers via UltraVNC and made the NVIDIA GRID card the primary display. Once I have done this the console stops.
I guess what I’m trying to figure out is what is the correct setup and user experience, should we continue with both the sVGA and K2 cards available or disable the sVGA adapter and break the console functionality.
Thanks in advance.