[SOLVED] Nvidia vGPU

[Esben Viborg]

Hi

We are running a mini-hosting center with 7 Proxmox nodes.
One of our customers are in need of some VDI's with some graphics power.
We would like to provide this service, and we are planning on using Nvidia K2 cards for this task.
Officially only Xenserver and vSphere supports vGPU using the K2 cards, however in late 2016 Nvidia announced plans for KVM to get this feature aswell at some point.
I haven't been able to find any relevant updates on the status of this feature.

Do anyone know if it is possible to pass GRID profiles through to vm's on proxmox?

 

[wolfgang]

Hi,

As far I can see this is not stable yet.
The RedHat and Nvidia roadmap say it should work Q2 2018
https://twitter.com/NVIDIAVirt/status/861716134533275648/photo/1

The qemu 2.11 has experimental support for it.
https://wiki.qemu.org/ChangeLog/2.11#VFIO

When I reviewed the Qemu code it looks like it isn't finished yet.

Do anyone know if it is possible to pass GRID profiles through to vm's on proxmox?

No, it is not possible.

 

[Esben Viborg]

Hi Wolfgang

Thanks for the info.
I guess we will need one Xenserver node until Proxmox gets this feature

Best Regards
Esben

 

[xiaolin0199]

Does the current proxmox support the vgpu function of k2?
Is there a list of supported hardware that is currently supported?

Hi,

As far I can see this is not stable yet.
The RedHat and Nvidia roadmap say it should work Q2 2018
https://twitter.com/NVIDIAVirt/status/861716134533275648/photo/1

The qemu 2.11 has experimental support for it.
https://wiki.qemu.org/ChangeLog/2.11#VFIO

When I reviewed the Qemu code it looks like it isn't finished yet.


No, it is not possible.

 

[avladulescu]

Hello,

Is there any update on this topic concerning the prox 6.1-8/6.2 for nvidia grid k2 passing it to multiple kvm ?

Thanks,
Alex