vSphere host reports less reserved GPU memory than View pool configuration

[mcrampton]

Hello, I have a View pool configured to assign 512 Mb of video memory to all VM's. My Dell R720 hosts have an Nvida Grid K1 card in each.

However, when I run GPUVM on a putty session with my host, I see the following.

Any idea why the mismatch?

[Linjo]

Hi mcrampton

Yes this is by design to increase the number of VM:s that can utilize the GPU.

So half of the VRAM size configured will come from the GPU and the rest from the host memory.

Exactly why that is I don't know, I'm sure one of our developers like krd could it explain it in detail.

// Linjo

[mcrampton]

Would love some more clarification on this. I don't want the video memory assigned to come from host memory, I want it to come from the GPU. We are doing other memory intensive operations with our servers, so I would like to utilize the GPUs as intended and not steal from system RAM.

[Linjo]

Only 256Mb per desktop will be used in worst case, should be negligible compared to all the other overhead.

It does work as intended and will not change anytime soon, another option is to use Directpath IO and dedicate a GPU to a virtual machine, but then you will be limited to the amount of GPU:s on the card.

For example the K1 have 4 GPU:s so then you can have 4 virtual machines with GPU, the K2 have 2 GPU:s.

This is called "vDGA" and is currently in Tech Preview.

// Linjo

[mcrampton]

Thanks. Yes, it's more from a performance standpoint. I have played around with vDGA, but until it supports 2 monitors and can survive a reboot, it's not a feasible solution. Looking forward to when it's released to general availability.