MMIO space (2 GPU's per VM) Tesla M10 (4 GPU's in 32GB total)?

Virtual GPU Forums More vGPU Forums Tesla Boards
1

Hi!

I have configured a Tesla M10 (4 GPU’s) in a HP proliant DL380 Gen10.
Also have used DDA and assigned 2 GPU’s per VM (2 Hyper-V RDS 2016 servers).

Can somebody tell me if these are the correct commands to apply to the VM with 2 GPU’s:

#Configure 32 bit MMIO space
Set-VM -LowMemoryMappedIoSpace 2000MB -VMName $vm

#Configure Greater than 32 bit MMIO space
Set-VM -HighMemoryMappedIoSpace 8000MB -VMName $vm

(Set-VM VM_NAME -GuestControlledCacheTypes $True -LowMemoryMappedIoSpace 2000MB -HighMemoryMappedIoSpace 8000MB)

If not what setting do I need to use for a correct optimal config?
I also could not find what the MMIO space will be for this card.

Thanks!

Regards

2

You cannot add 2 GPUs to a single VM. This won’t work properly. You don’t need to modify the MMIO settings for M10. Should work out of the box.

Regards

Simon

3

Hi Simon,

Thanks for the reply.

Should I change it back to some default config settings?
Or can I leave the settings as I described above?

Also on the microsoft site multiple GPU is supported. But not for M10?
Could you explain what would not work propperly when assigning two GPUs for one VM?

Regard Stephan

4

In general you won’t see any benefit from multi GPU unless the application really supports it (CUDA based applications for example). Could you show me where MSFT claims to support multi GPU in DDA mode? Currently I’m not aware of.
KVM for example supports multi-vGPU starting with vGPU7.0. Citrix or VMWare do not support multi GPU yet.

Regards

Simon