I’ve already opened a case with NVIDIA and also with Citrix for this. But they don’t have an idea what causes the problems yet. So I thought maybe someone here has an idea or recommendation.
Since we updated to Citrix 1903 and NVIDIA GRID drivers 9.0, we randomly get graphic errors inside XenDesktop Sessions (Please see the screenshot).
This issue can be reproduced when moving windows around inside a XenDesktop session. The graphical errors mostly persist until I resize the Citrix Session Windows or switch from fulscreen to window mode and back.
Please see the screenshot in the link for configuration settings which are applied.
I’ve also tested with Citrix VDA 1811 and 1906, same problem. With VDA 7.15 the issue does not occur, but we recognized a way better performance, quality and also latency with VDA 1903. That’s why we want to upgrade to 1903, it’s a big improvement for our virtual AutoCAD workspaces.
NVIDIA Support mentioned the following thing:
"As I mentioned, starting with Citrix Virtual Apps and Desktops 7 1903 version, Citrix remoting implementation has changed from using NVFBC to using DDAPI."
If anybody has an idea I would be happy to hear some of them.
Have you tried running a newer version of Windows 10? The version you’re running is over a year old now, try a test build on 1903 and see if that makes a difference.
First of all I would suggest to check the Framebuffer usage. Sounds like FB exhausting effects. Our support is right in terms of NVFBC / DDAPI so this can have a huge impact. DDAPI is meant for at least RS5 so I would also suggest to move to latest Win10 build first.
I can confirm these issues also exist on our setup:
Hypervisor: VMware ESXi 6.7 U2
Guest OS: Windows Server 2019
NVIDIA Host driver: 430.46
NVIDIA Guest Driver: 431.79
Oh this is also an a Tesla T4 setup.
You can quickly test for these artifacts by visiting the site https://www.mumfordandsons.com/ where the background video behaves like a severly scratched DivX movie CD (long trails behind objects, tearing all over the place.)
There is no solution atm. This is a Citrix issue with hardware decode YUV444 on their current Citrix receivers. Please open a support ticket with Citrix. They are well aware of this and are working on a fix.
Simply create a screenshot in the session when the issue occurs and you will see, that you won’t see any corruption in the screenshot later…
Do you have a ticket ID we can reference at Citrix?
We’re seeing similar issues, but we are not using YUV444/AlwaysLossLess.
Do you know with what Citrix Workspace App version the issues started? Perhaps a workaround could be to use one of the older versions.
Hi, we have similar issues with M60 YUV444/AlwaysLossLess Settings. Case is opened and they confirmed that they have a known issue regarding T4 but not M60 will test the new Workspace App and report back here.
I can guarantee that you won’t be affected by this issue with Tesla M60! The new workspace app added support for multiframe encode which was introduced with Turing (T4) architecture.
You should test with a bigger framebuffer profile first to see if your issue is related to framebuffer exhaust, which is the issue I see most.
