Promhub Logo Prom hub

NVIDIA DCGM Dashboard for Kubernetes (MIG & Non-MIG GPUs)
1,441

Created 5/5/2025
Updated 5/6/2025
Revision 1
Grafana Version >=11.2.0
Datasources
Prometheus

Description

This Grafana dashboard uses the NVIDIA DCGM exporter to monitor GPU performance in Kubernetes clusters. Supporting both MIG and non-MIG GPUs, it provides real-time metrics on utilization, memory, temperature, power, and more, enabling efficient management of NVIDIA GPU resources.

Screenshots

Dashboard screenshot 1
Source Grafana.com

Used Metrics 14

  • DCGM_FI_DEV_FB_FREE

  • DCGM_FI_DEV_FB_USED

  • DCGM_FI_DEV_GPU_TEMP

  • DCGM_FI_DEV_MEMORY_TEMP

  • DCGM_FI_DEV_MEM_CLOCK

  • DCGM_FI_DEV_NVLINK_BANDWIDTH_TOTAL

  • DCGM_FI_DEV_PCIE_REPLAY_COUNTER

  • DCGM_FI_DEV_POWER_USAGE

  • DCGM_FI_DEV_SM_CLOCK

  • DCGM_FI_DEV_TOTAL_ENERGY_CONSUMPTION

  • DCGM_FI_DEV_XID_ERRORS

  • DCGM_FI_PROF_DRAM_ACTIVE

  • DCGM_FI_PROF_GR_ENGINE_ACTIVE

  • DCGM_FI_PROF_PIPE_TENSOR_ACTIVE

Get Dashboard✕
Download
Copy to Clipboard

Dashboard Preview - NVIDIA DCGM Dashboard for Kubernetes (MIG & Non-MIG GPUs)

Most Grafana dashboards are not optimized for small screens. For the best experience, view this preview on a larger screen.
grafana.shipit.dev/nvidia-dcgm-dashboard-for-kubernetes-mig-non-mig-gpus