Tesla M40 Nvenc not initializing

I have a system running windows 10 pro 21H1 with dual E5-2699v3’s, 64GB of 2133Mhz DDR4 and a Tesla M40. My goal for this system was to run VM’s in Hyper-V with GPU-P passthrough and stream the VM’s to clients over Parsec. Whenever a client attempts to connect, they either get a 14003 or 15000 error, meaning the host could not initialize the encoder. I’m assuming this is due to Nvenc not initializing, as parsec hosting also fails on the host OS. I’ve set the card to WDDM mode, and it can run Unigine Heaven very well on the host and VM’s, but Nvenc seems to not have been passed through. Any ideas?


unfortunately M40 was never supported for graphics and should work only in TCC mode. For WDDM support in Windows you would need the vGPU driver for DC boards but M40 is not enabled.


I know this isn’t an intended use case, But the standard Tesla driver has built in support for putting the card into WDDM, you just need another display device to route the rendered frames. I’m just trying to figure out why I’m getting this error in a virtual machine when all other graphics applications are getting full acceleration.