DCGM · Data Center GPU Manager
NVIDIA's tool for monitoring, diagnosing and health-checking GPUs across a fleet.
Current numbers
30 s (30,000 ms)NVIDIA DCGM exporter's default GPU metric collection interval; tighten it for second-scale detection
under 45 min (4 GPU) / under 2.25 hr (8 GPU)DCGM level-4 runtime estimate on NVIDIA's 4-GPU and 8-GPU Hopper systems; measure your own