Hi.
Is it possible to recognize individual vGPUs as individual physical GPUs?
I want to do distributed deep learning with multiple vGPUs on a single GPU.
I wish I could use vGPU directly, but it doesn’t seem to be possible.
Hi,
Sorry, but I don’t understand the use case described. What exactly are you going to achieve? Assign multiple vGPU profiles to a single VM?
Hi.
Not a VM. I want to use multiple vGPUs at the same time in a PC equipped with RTX A6000.
It does not connect to other PCs.
I want to use vGPUs for deep learning.
Are vGPUs recognized by nvidia-smi like GPU?
2022年2月15日(火) 17:55 sschaber via NVIDIA Developer Forums <nvidia@discoursemail.com>:
For sure this works. This is exactly the intention for vGPU. You need the use the mode selector tool available from our developer portal to modify the A6000 to make it capable to run vGPU. In addition you need a supported Hypervisor
great! I wanted to use vGPUs for distributed reinforcement learning, but no one seems to use vGPUs for it.
Where is the portal site? I would like to know the specific modification method. However, I don’t know if I can do it
2022年2月15日(火) 18:27 sschaber via NVIDIA Developer Forums <nvidia@discoursemail.com>:
I downloaded the’NVIDIA Display Mode Selector Tool’.
How do I run the ‘displaymodeselector’ on Linux? It seems that the manual only mentions Windows…
NVIDIA Display Mode Selector Tool User Guide.pdf (343.2 KB)
The tool also works on Linux an mentioned in the docs.
Suggestion: As it seems you are not really familiar with vGPU you should look for a skilled partner. As you have seen, vGPU is not a one click solution but a software stack that needs some experience. Here is our partner network where you find a skilled partner in your region:
https://www.nvidia.com/en-us/about-nvidia/partners/partner-locator/
That’s right.
What exactly is the next “necessary support”?
Also, what should I do if I want to restore the firmware?
WARNING: This operation updates the firmware on the board and could make the device unusable if your host system lacks the necessary support.
Please check with a virtualization partner first. If you don’t know what you are doing, you may end up with a displayless machine. Keep in mind that changing mode will disable the display heads so you will need a different onboard GPU or second NV GPU in the machine that serves as primary display adapter!
I should check with virtualization partner, but I can’t contact virtualization partner because my purpose isn’t a VM. If there is no problem with the onboard gpu, there is no problem. There is no problem if the RTX A6000 is not damaged in terms of hardware, but is there a possibility that damage will occur in terms of hardware?
The concept of vGPU always works only with VMs. If you want to use GPU partitions on baremetal then you can only use MIG with A30 or A100 GPUs. No way with A6000.
I tried my best to support. This is just a forum and not a consulting replacement. Please get professional assistance in your region