NVIDIA DCGM Documentation

NVIDIA Data Center GPU Manager (DCGM) provides libraries and tools for observing, managing, validating, and integrating NVIDIA data-center accelerators.

Choose a documentation area according to your task. Learn introduces concepts and complete workflows. Reference provides exact syntax, accepted values, state, output, and API details.

Documentation areas

About

Understand DCGM's scope, focus areas, and intended audiences.

Installation

Review supported platforms and prerequisites, install the packages, start the service, and verify the host.

• Install DCGM
• Install DCGM Exporter

Learn

Learn terminology, components, operating modes, core services, modules, and operational workflows.

• System Administrator Getting Started
• Application Developer Getting Started
• Core Services
• Modules
• Configure Prometheus for DCGM Exporter
• Customizing the systemd Service

Reference

Look up exact command behavior, field identifiers, single-node diagnostics, multi-node diagnostics, and public C, C++, and Go interfaces.

• Command Line
• Field Identifiers
• Diagnostics
• Multi-Node Diagnostics
• C/C++ API
• DCGM Exporter Metrics
• Go SDK

Release Notes

Review features, improvements, fixes, and compatibility changes for DCGM and DCGM Exporter.

• DCGM
• DCGM Exporter

Suggested starting points

• New operators: Installation followed by Getting Started for System Administrators.
• Command users: Command Line.
• Diagnostic users: Diagnostics and Multi-Node Diagnostics.
• Telemetry operators: Install DCGM Exporter, then configure Prometheus.
• Library integrators: C/C++ API or Go SDK.