Beginner question- Tesla K80 (or similar $) for 2D/3D acceleration on Dell 720xd

I need some purchasing advice regarding Nvidia cards for VM graphics acceleration. I’m building a home VM server as a multimedia content creation workhorse. My intention is that anyone in the family can log into the appropriate VM from any machine in the house and do whatever project they are working on.

I’m building 3 dedicated VMs right now (all Windows based)
2D graphics VM (photoshop, graphics, etc)
3D CAD/CAM & 3D printing (fusion360, artcam, vectric aspire, slic3r, blender, etc)
Video editing (adobe premier, after effects)

  • I’m looking into some of the open source options for similar functionality, in the future I might build similar VMs based on Ubuntu & gnu licensed products.

This is a home project, so I’m attempting to keep it as budget friendly as possible. I’ve been seeing the NVidia K80 24GB Keppler Cuda cards for $350. On paper, these cards seem to be the biggest bang for the buck I could put in my system. (Dell 720xd). I haven’t been able to figure out if these cards will do 2D/3D graphics acceleration & video compression acceleration on virtual machines. I found something like a white paper from a group that was using them for something similar, but otherwise I get the impression that these cards are used more for stuff like AI and super data crunching. Nvidia marketing materials seem to recommend different cards too, but I’m not sure if that’s because I can’t pull up the same info from several years ago or not.

Ultimately, my question is: is the K80 card a good option for server 2D/3D acceleration of virtual machines?
Is there a better option (something under $400 that’ll work in a Dell 720xd server)?

(I’m aware of Dell’s official stance on graphics cards in this chassis. I can keep it cool. I’m looking for options that will work whether or not they are officially supported.)


The K80 is Compute only and won’t do 3D. Also, it’s an extremely old architecture. Basically, don’t bother.

The easiest way for you to do it is with a Quadro (or multiple Quadros) running in Passthrough (1 Per VM).

There are plenty of Quadro options available depending on your budget. Here are some from the Pascal architecture that you can search for: If you want something current, then you’re looking at the Turing architecture. (I’ve skipped Volta because you’ll be looking at GV100s which are expensive).