So I'm running into a bit of a strange issue that I can't seem to find an answer for anywhere. I currently have two Windows 10 VMs that I pass a Nvidia Tesla P100 through to using the vgpu unlock to create two 8GB profiles for each VM. The end goal is to have these two vms set up for cloud gaming machines.
For the most part, this setup works fine and gives me about the performance that I would expect. For instance, I can have Unigine's heaven benchmark running on both VMS and have the frame rate locked at 60. The issue I'm running into however is that almost 20 minutes exactly of the VM being online, performance throttles and that locked 60 fps will drop down to a locked 15fps until I restart the VM. I thought I may have been due to thermals at first however the GPU never got over 45C and the CPU never went over 50C. It also doesn't look like it's a case of the Tesla card switching performance profiles due to some kind of error either. It is also interesting to note that if I spin up one VM and let it run until it starts to throttle and then spin up the second VM, that second machine will run at full power for about 20 minutes before also throttling.
I am brand new to all of this and am still trying to figure things out so I don't know if this is due to a setting somewhere I need to change or if this is due to some other sort of issue with the hardware.
Server specs include:
Xeon E5-2640 v4
64GB DDR4 ECC
2x 512GB SSD in Raid 0
nVidia Tesla P100 16GB
VM Resources:
6 cores
16GB Ram (Ballooning Disabled)
100GB virtIO storage.
For the most part, this setup works fine and gives me about the performance that I would expect. For instance, I can have Unigine's heaven benchmark running on both VMS and have the frame rate locked at 60. The issue I'm running into however is that almost 20 minutes exactly of the VM being online, performance throttles and that locked 60 fps will drop down to a locked 15fps until I restart the VM. I thought I may have been due to thermals at first however the GPU never got over 45C and the CPU never went over 50C. It also doesn't look like it's a case of the Tesla card switching performance profiles due to some kind of error either. It is also interesting to note that if I spin up one VM and let it run until it starts to throttle and then spin up the second VM, that second machine will run at full power for about 20 minutes before also throttling.
I am brand new to all of this and am still trying to figure things out so I don't know if this is due to a setting somewhere I need to change or if this is due to some other sort of issue with the hardware.
Server specs include:
Xeon E5-2640 v4
64GB DDR4 ECC
2x 512GB SSD in Raid 0
nVidia Tesla P100 16GB
VM Resources:
6 cores
16GB Ram (Ballooning Disabled)
100GB virtIO storage.
