3 brand new dell r730 with an m60 in each (factory shipped) each of them throwing the same error with nvidia-smi:
Unable to determine the device handle for GPU 0000:05:00.0: Unable to communicate with GPU because it is insufficiently powered.
This may be because not all required external power cables are attached, or the attached cables are not seated properly.
installed latest grid driver: 361.45.09-1OEM.600.0.0.2494585 from licensing portal
ran gpumodeswitch on all hosts to switch all gpu’s to graphics mode and confirmed with lspci -n | grep 10de: 0000:05:00.0 Class 0300: 10de:13f2 [vmgfx0]
0000:06:00.0 Class 0300: 10de:13f2 [vmgfx1]
A VM with vGPU starts on one host but not on another > Failed to start the virtual machine.
Module DevicePowerOn power on failed.
Could not initialize plugin ‘/usr/lib64/vmware/plugin/libnvidia-vgx.so’ for vGPU ‘grid_m60-0b’.
No graphics device is available for vGPU ‘grid_m60-0b’.
maybe it’s just a wiring issue, I will have to check that tomorrow when going on site
The M60’s should have a 300W power connector, and not a standard 8 pin PCIe cable. I’m not sure if Dell has a specific cable, or whether they use the adapter cable that takes the feed from 2x PCIe cables to deliver 300W.
I seem to recall it was something called a "8pin to 8pin+6pin" but this is one you need to go back to Dell on and check you have the right power supply and cables as per the GPU enablement kit.
It was a wiring issue, after connecting them as you instructed we don’t see the nvidia-smi issues anymore and we are now able to use vGPU on all hosts.
They came from dell direct, so strange how one host was wired correctly and others not.
Hi @Technicalmt have you got any feedback on the specifics or a reference SR with Dell. I think I have one with a similar issue at the moment and am going back and forth with Dell support trying to resolve. Can you please provide and explanation or a photo of how the GPUs are cabled on a working config?
I have the same issue with brand new DELL R730 server and vSphere 6.0 U2.
Each Server has 2x Nvidia M60 factory installed. But only one of the server is able to see and use both gpu boards. The others show the message "Unable to communicate with GPU because it is insufficiently powered.".
I checked the BIOS and software components. I couldn’t find any difference. Please help. It’s for me actually not possible to check the cables inside the servers.
Both power supplies need to be working together (not redundant) wish I knew this before buying my r730xd. I’m running a K80 on esx6.7 2012vm. I would get purple screen of death about 30 min after starting the VM and my vendor explained this power caveat.