Quadro RTX A4000 as vGPU in RDSH-VM

Virtual GPU Forums General Discussion
1

Hello everyone,

i am trying to get a Quadro RTX A4000 working properly as a vGPU for a newly set up RDSH.

The GPU runs inside a Dell R740 Host-System and is passed through via DDA to the RDSH, which runs as a Hyper-V VM.

The GPU is doing it’s job fine until more than 15 Users connect to the RDSH, dwm.exe starts crashing instantly if User 16 tries to connect, giving the User only a blackscreen.

Error Infos:
Name der fehlerhaften Anwendung: dwm.exe, Version: 10.0.17763.831, Zeitstempel: 0xd5c9fdea
Name des fehlerhaften Moduls: udwm.dll, Version: 10.0.17763.1852, Zeitstempel: 0x6f3c3f5d
Ausnahmecode: 0xc00001ad
Fehleroffset: 0x00000000000b15d5
ID des fehlerhaften Prozesses: 0x4344
Startzeit der fehlerhaften Anwendung: 0x01d87db83505be01
Pfad der fehlerhaften Anwendung: C:\Windows\system32\dwm.exe
Pfad des fehlerhaften Moduls: C:\Windows\SYSTEM32\udwm.dll
Berichtskennung: afed10c4-e78d-4753-a9dd-00ae4d92b896
Vollständiger Name des fehlerhaften Pakets:
Anwendungs-ID, die relativ zum fehlerhaften Paket ist:

So i’m wondering if the A4000 is supported for what I am trying to do. Drivers and Windows Updates are all up to date.

Any help would be appreciated!

2

How is your framebuffer saturated when the 16th user tries to connect?
A4000 is not vGPU capable but doesn’t matter as you run just Passthrough. I doubt that Dell supports the A4000 in server systems at all.
What you are seeing always happens sooner or later on every RDSH installation. But CCU count us different. For most customers this happens with almost 50 CCUs instead of that early.

3

Hey, thanks for your reply.

Thats how the workload on the GPU looks with 15 Users signed in on the RDSH and User 16 trying to sign in.

image
image2018×1004 60.2 KB

On the other RDSHs we currently use Quadro P4000 without problems, we had one A4000 laying around so I decided to test it.

4

Hmm, really strange. Could you test an older driver based on R470 branch? 473.47 for example. Looks like a resource issue but there are so many GPU resources left that I don’ t understand why this is happening

5

Unfortunately I already tried these drivers with the same results:

image

Could try even older drivers but I’m pessimistic ^^

6

Hi, I’m in the exact same situation. After installing a brand new RDSH (HPE-DL380G10, 2 x 3.1Ghz Gold Xeons 24c, 1Tb mem, NVME disks, 2 x 1400w PSU) including 2 x RTX A4000, I have GPU crashing after 8 users already. I tried many different setups, baremetal with only 1 gpu and with both, Vsphere 8.1 passthrough, with or without vGpu. All same results. Oddly enough, no issues with the exact same setup and T4 or M4000 installed.

Running fully updated Server 2022 for the RDSH. 537.70 driver. Has there been any resolution to the above case that might help me out here?

PS, some users work with Revit, which obviously is not your typical office product, may that cause the issues to happen sooner?