Nvidia H100 Temperature

[Thomas.Dietsch]

Hello! We are building a new system that is running two Nvidia H100 cards with NVL, as well as a Nvidia L4 card, and i was wondering if someone could help with getting them to show up under the temperature logs/testing.

I have included the Debug files i generated from the system.

Also, we noticed that the GPGPU testing in BIT does not seem to hit the cards very hard, and we are trying to Stress and maximize the load on the cards as much as possible. is there anything that can be done to increase the load on the cards via the BIT software?

[David (PassMark)]

I found this in your log.
Perform test: GPGPU at 90%

To increase the load set the duty cycle to 100%.

For temperatures, what options do you have available and selected in the preferences window?

Example below

[Thomas.Dietsch]

David,

I didn't see it in there, but i will double check again. It seems that BIT detects the cards (as the GPGPU test seems to see them/ hit them, albeit at significantly less than their maximum bandwidth), but are not available in the temperature selection. I will send over a photo once I can get the system turned back on.

[David (PassMark)]

albeit at significantly less than their maximum bandwidth

This is expected, as you didn't select 100% load.

[Thomas.Dietsch]

Hello David,

I included a few screenshots showing that the gfx cards dont seem to show up in BIT, as well as a screenshot of Nvidia SMI which shows that the cards are not being hit very hard even when i have the GPGPU test set to 100%.

Is there something else you need from me in order to get their temperatures linked up with BIT?