Jump to content
Updated Privacy Statement
  • 0

Nvidia Grid cards and XenApp 7.15 LTSR CU2


Andy White1709154166

Question

Hello,

 

We use XenApp 7.15 LTSR CU2.  We use published Apps and Published desktops to users.

 

I've added the K1 Nvidia Grid cards to our VDA workers servers and users are working just fine.  I can't see much of a difference, but are there any additional settings to be done once added?

 

We also use WEM and I'm not sure if this has any settings to implement too.

 

Thanks

Link to comment

Recommended Posts

I have a feeling from what you said above you may not be actually using Nvidia GRID as a technology and rather the older vDGA pass through. Before I start giving you GRID setup information can you attach a screenshot of one of the VMs with the GPU core currently assigned so I can have a look at its setup?

 

https://www.lewan.com/blog/2015/03/30/vgpu-vsga-vdga-software-why-do-i-care

 

If it turns out you are using vDGA then its definitely time to move to GRID....you use the same GPU cards and its free for the K1 & K2 Nvidia cards! 

 

Some reading on GRID below:

 

https://images.nvidia.com/content/grid/pdf/GRID-vGPU-User-Guide.pdf

 

 

Link to comment

I think we have been using them wrong for years then.  I did have the K1 cards installed on our 2 hosts for years and the company that host them just allowed passthrough.  I would then just shut the Citrix VMs down and edit them in vcenter and add a vGPU from the K1 and turn on. The Nvidia would be on the master image  too.  Nothing else was done.  What is strange we had a Citrix consultant help set this up back then too.

 

We changed vcenter recently so we had to remove the K1 cards of all VMs and now is the time to read them.  Glad I posted here.  I remember them saying they didn’t need to do anything else at the time too.

 

Just some questions from you numbered replies.

 

Regarding number 2 where would they get this VIB driver from for ESXi?  We are on 6.5.

 

Regarding number 3 where do I assign this profile and does it only appear when they add the VIB driver?

 

Regarding number 5 do I just rerun the agent install or uninstall and reinstall on the master image?

 

We are on LTSR CU2 and hope to go to CU3 in a couple of months.

 

thanks

 

 

Link to comment

Of course. Once the GPU cards have been put into the servers a driver (VIB) should be installed onto the ESXI operating system. Once this has been done in vSphere you get the option to add a shared PCI device, after adding this option to each VM it will show up as Nvidia GRID and you can then select whats called a GPU profile for that VM.  On the drop down the user should see a selection of profile to choose from ranging from K100 to K180Q. this is basically what share of the GPU this VM gets and I have attached an image showing this. 

 

Depending if your going for density or performance in 3D apps like AutoDesk & Adobe stuff. For user density a lower profile would be ok but you wont see a lot of improvements. The K140Q is a nice middle of the road giving the VM 1GB of GPU memory allocation. 

 

The order would be:

 

1) install physical GPU's in servers

2) install Nvidia GRID drivers onto server.....VIB format file for ESXI operating systems

3) assign the VM's a GRID profile suitable for workload

4) install nvidia GRID drivers on the VM itself 

5) best to configure the rest of the VM & install any software at this stage before but the final stage installing the VDA in HDX3DPro mode. 

6) test with programs like GPUshark (detailed mode ticked) etc to see if the GPU core given to the VM is being utilised when playing videos and such

7) configure Citrix policies to tune the 3D quality and performance to get the best experience and balance.  

 

Thats pretty much it in a nutshell mate. 

Screenshot_1.jpg

Link to comment

Our VMware is managed for us, can you explain more about what I need to do with this k140q profile so I can explain to them as I guess they will need to do this? Do they download it etc?

 

For the order is is:

 

- add passthru card to VMs

- rerun the VDA install on the master image and check the HDX3D option and push out to the VDA workers

- install the k140q profile

 

thanks

Link to comment

If these VDA's were installed in normal setup and then you have assigned GPU's to them you will no doubt have to re-install and during the setup wizard you will need to make sure HDX3DPRO is ticked. 

 

Also if you have assigned a small GRID profile to the VM's in vSphere like the K120Q profile you will struggle to see a difference. Use at least the k140Q profile for decent (ish) performance on the older K1 cards. 

 

Once you have the correct GRID profile assigned to the VM's in vSphere, got the VDA re-installed in HDX3D mode you will need to tweak the Citrix policies for 3D passthrough to get the very best out of the GPU's. 

 

Once you have done the above pop back and I will share with you the polices we have set here. 

 

GPUShark is a good tool to use as it shows GPU usage in % so you can see in real time when playing a video etc if the GPU is being used or just the CPU. 

Link to comment

There are a number of Citrix policies you should configure for HDX3DPro passthrough for varying performance & graphics quality but first......

 

Have you got the GPU card showing in device manager on the Citrix desktops?

 

Have you got the correct GRID drivers installed on both the host servers & VMs?

 

Have you got the Citrix VDA installed in HDX3DPro mode?

 

What are you using to test the difference? some programs like AutoCAD, Adobe Photoshop/Illustrator need to be told to use hardware 3D sometimes or they will default to software rendering. 

Link to comment

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...