vDGA GRID K2 passthrough multi-gpu support


I have 3 x Dell R730s configured with 2 x NVIDIA GRID K2 boards in each servers. I have dedicated 1 GPU to a VM by activating the passthrough mode on one of my ESXi server. When launching an application Image Generator (Genesis IG), I have noticed that my GPU was actually being used at 99% which explains some not so good performance, it is close to what the GeForce 970GTX on a regular pc provides but not really there yet.

In order to address the performance issue I have decided to add another GPU to this machine in passthough mode and I have now 2 GRID GPUS installed and showing up on my VM. My question is it possible to have the 2 GPUS working together in a multi-gpu support concept or SLI? Right now if I launch the same application I can only see 1 GPU working and the other one is at 0%. A single GPU would probably be enough to run the IG if I didnt have to waste 25% on the GPU for the PCoIP session alone going to my Zero client.



This isn’t supported by the hypervisor vendors.

That is really unfortunate. Thanks for your prompt reply.


what is the performance difference of a GRID 2.0 M60 board compared to the K2s ?

Depending on the workload between 1.4x and 2x

It also has double the amount of GPU RAM.