Starting from version 7.17, Linux VDA 3D Pro supports vGPU for NVIDIA Pascal GPUs (Tesla P40) so that customers can better benefit from the latest NVIDIA vGPU technologies. It also supports hardware encoding on all NVIDIA supported vGPU profiles in Linux, which removes the legacy limitations on vGPU hardware encoding.

In this article, we provide steps to use NVIDIA Pascal GPUs (Tesla P40) and hardware encoding on Linux HDX 3D Pro VDAs, as well as general troubleshooting tips.

Instructions

How to use NVIDIA Pascal GPUs (Tesla P40) on Linux HDX 3D Pro VDAs

• NVIDIA GRID™ - Tesla P40 cards
• Linux VDA 7.17 and above
• A server capable of hosting your hypervisor (such as, Citrix XenServer) and NVIDIA Tesla P40 cards.
  • Make sure one or more NVIDIA Tesla P40 cards are installed in your server platform.
  • Install your hypervisor with proper license, the minimum supported hypervisor versions are:
    • ESXi 6.0 Update 3 and later, or ESXi 6.5 and later. See Virtual GPU Software R384 for VMware vSphere Release Notes .
    • Citrix XenServer 7.0. See Virtual GPU Software R384 for Citrix XenServer Release Notes.
•  Download NVIDIA virtual GPU software package for your chosen hypervisor, which consists of the following software:
  • NVIDIA Tesla P40 host drivers for your hyperviso
  • NVIDIA Tesla P40 software display drivers for supported Linux guest VMs
  • The minimum support NVIDIA Linux driver version is 384.73
• Prepare XenApp and XenDesktop 7.17 environment

1. Install and configure NVIDIA Virtual GPU Manager on your choosing hypervisor, refer to the following instructions:
   • Installing and Configuring the NVIDIA Virtual GPU Manager for Citrix XenServer
   • Installing and Configuring the NVIDIA Virtual GPU Manager for VMware vSphere
2. Assign Tesla P40 virtual GPUs to the target Linux VM, both “pass-through the whole GPU” and vGPU scenarios are supported. Refer to following instructions:
   • Configuring a Citrix XenServer VM with Virtual GPU
   • Configuring a vSphere VM with Virtual GPU
3. Install the NVIDIA Virtual GPU guest drivers on the target Linux VM, refer to following instructions: Installing the NVIDIA Virtual GPU Software Display Driver on Linux
4. License NVIDIA vGPU so that the target Linux VM is capable of running the full range of DirectX and OpenGL graphics applications. See the instructions at Licensing NVIDIA vGPU
5. Install Linux VDA HDX 3D pro on the target Linux VM, make sure that the VDA can register to the Delivery Controller and the published Linux desktop sessions can be launched successfully.

How to enable/disable hardware encoding

• Only H.264 YUV420p is supported
• Nvidia Tesla M60 card vGPU profiles other than 8Q do not support cuda, as a result, NVENC API and Citrix 3D Pro hardware encoding are not available.

1. On Citrix Studio, enable the Citrix HDX policy Use hardware encoding for video codec. This policy is enabled by default.

2. Edit the HDX policy Use video codec for compression and set the value as “For the entire screen” (default setting: Use when preferred).

 
Note: You need to log off the current user session from either Citrix Studio or inside the Linux VDA and log on again to make the policy settings take effect.

1. Launch Linux VDA session
2. Open the terminal on Linux VDA and run command nvidia-smi, type “C+G” means hardware encoding is active:

Note: The NVIDIA System Management Interface, nvidia-smi, is a command-line tool that reports management information for NVIDIA GPUs. See NVIDIA System Management Interface nvidia-smi for more information.