NVIDIA
Passthrough Tesla P4 from XenServer 7.to XenDesktop 7.18
Hello, everybody, I have received a test installation with servers from HPE. These servers have an NVIDIA Tesla P4 graphics card. I want to test how and if Graphic passthrough works to give Inventor users better performance. I have tested so far: I installed XenServer 7.5 on the server. Then I created a VM and passed the graphics card through there. The VM is Windows 10 Built 1709, then I installed XenDesktop 7.18, Inventor 2017 and NVIDIA graphics card drivers. I have activated the following policies on my delivery controller: Enable lossless -> Allowed HDX3DPro qulity settings -> 6553600 Optimize for 3D graphics workload -> Enabled When I now start a desktop, however, not the NVIDIA graphics card is used but the CPU. Now my questions: Is the Tesla P4 card even supported for XenDesktop? Have I activated the correct policies or do I have to activate "3D HDX" in the setup? Do you have any other ideas what I need to change? Many thanks in advance!
Hello, everybody,
I have received a test installation with servers from HPE. These servers have an NVIDIA Tesla P4 graphics card.

I want to test how and if Graphic passthrough works to give Inventor users better performance.

I have tested so far:
I installed XenServer 7.5 on the server. Then I created a VM and passed the graphics card through there. The VM is Windows 10 Built 1709, then I installed XenDesktop 7.18, Inventor 2017 and NVIDIA graphics card drivers.
I have activated the following policies on my delivery controller:
Enable lossless -> Allowed
HDX3DPro qulity settings -> 6553600
Optimize for 3D graphics workload -> Enabled

When I now start a desktop, however, not the NVIDIA graphics card is used but the CPU.

Now my questions:
Is the Tesla P4 card even supported for XenDesktop?
Have I activated the correct policies or do I have to activate "3D HDX" in the setup?
Do you have any other ideas what I need to change?

Many thanks in advance!

#1
Posted 08/23/2018 05:28 AM   
Hi, for sure Tesla P4 is supported for this use case. Most likely this is an issue with Inventor not detecting the right GPU. You can disable the Citrix WDDM adapter in the VM for a quick test to see if the Tesla GPU is used afterwards. Regards Simon
Hi,

for sure Tesla P4 is supported for this use case. Most likely this is an issue with Inventor not detecting the right GPU. You can disable the Citrix WDDM adapter in the VM for a quick test to see if the Tesla GPU is used afterwards.

Regards

Simon

#2
Posted 08/24/2018 09:27 AM   
Hi If you're using Passthrough, then why not use the 8Q vGPU profile? That way you'll get Hypervisor monitoring of the GPU (which you don't with Passthrough) and all the other benefits of using vGPU over Passthrough. You'll still need the same license which ever one you use, so the costs will be the same, but vGPU will give you more options / flexibility over Passthrough. Did you install the components in the correct order? ... 1) XenTools 2) NVIDIA Drivers 3) VDA 4) Inventor Your Citrix Policy settings are wrong. Create a clean Citrix Policy with [u]only[/u] the following: [b]HDX Adaptive Transport[/b] = Preferred [b]Optimize for 3D graphics workload[/b] = Enabled [b]Target frame rate[/b] = 60 fps [b]Use hardware encoding for video codec[/b] = Enabled [b]Use video codec for compression[/b] = Use video codec when preferred If you use 3D SpaceMouse, then don't forget to allow USB Passthrough [b]Client device USB redirection[/b] = Allowed [b]Client USB Plug and Play device redirection[/b] = Allowed Make sure you don't have any other settings in this policy initially, and ensure that no other Citrix Policies are being applied to the VDA. Once the VDA is working, then you can start modifying the Policy settings. The VDA installer will automatically install the correct VDA version, so if you have the NVIDIA vGPU drivers installed and vGPU attached, it will install the HDX 3D Pro VDA automatically. (Hence why I asked whether you installed the components in the correct order ...) Regards Ben
Hi

If you're using Passthrough, then why not use the 8Q vGPU profile? That way you'll get Hypervisor monitoring of the GPU (which you don't with Passthrough) and all the other benefits of using vGPU over Passthrough. You'll still need the same license which ever one you use, so the costs will be the same, but vGPU will give you more options / flexibility over Passthrough.

Did you install the components in the correct order? ...

1) XenTools
2) NVIDIA Drivers
3) VDA
4) Inventor

Your Citrix Policy settings are wrong. Create a clean Citrix Policy with only the following:

HDX Adaptive Transport = Preferred
Optimize for 3D graphics workload = Enabled
Target frame rate = 60 fps
Use hardware encoding for video codec = Enabled
Use video codec for compression = Use video codec when preferred

If you use 3D SpaceMouse, then don't forget to allow USB Passthrough

Client device USB redirection = Allowed
Client USB Plug and Play device redirection = Allowed

Make sure you don't have any other settings in this policy initially, and ensure that no other Citrix Policies are being applied to the VDA. Once the VDA is working, then you can start modifying the Policy settings.

The VDA installer will automatically install the correct VDA version, so if you have the NVIDIA vGPU drivers installed and vGPU attached, it will install the HDX 3D Pro VDA automatically. (Hence why I asked whether you installed the components in the correct order ...)

Regards

Ben

#3
Posted 08/24/2018 06:40 PM   
Scroll To Top

Add Reply