DGX Dashboard metrics

I’m a bit confused about DGX’s dashboard metrics. Can someone explain to me why these two metrics are different?

  • GPU Util Shows how busy the GPU’s Streaming Multiprocessors (SM) are. i.e., how much of the GPU’s compute (CUDA core) capacity is actively executing kernels.
  • system memory: Shows how much of the unified VRAM is currently allocated by applications.

Note, we are aware of an issue with memory reporting issue report on the dashboard which will be addressed in a future release.

1 Like

@ashreaperone How did you add the additional telemetry on the right hand side of your dashboard please, including processes, resources and file systems? I only see the default set-up (shown below)

@pedrorush That is not a part of dgx dashboard. If you go to Show Apps in in the dock then search System Monitor

1 Like

@ashreaperone Ah - I guess you are using the Spark in desktop mode (ie you’re using the DGX GUI)? I am using it in network appliance mode. So I need to figure out how to make the system monitor available via NVIDIA Sync or some other telemetry.

I’m a bit confused by DGX Dashboard, and the prominent placement of DGX Dashboard in the Sync UI, when it does so few things. I was expecting to be able to add more tiles/apps. And a bit more comprehensive system utilization stats, tell me what is using all the ram.

1 Like