AMD SMI documentation

The AMD System Management Interface (AMD SMI) library offers a unified tool for managing and monitoring GPUs, particularly in high-performance computing environments. It provides a user-space interface that allows applications to control GPU operations, monitor performance, and retrieve information about the system’s drivers and GPUs.

Find the source code at ROCm/rocm-systems.

Important

This AMD SMI project supports Linux bare metal and Linux virtual machine guest environments. For documentation regarding virtualization on SR-IOV Linux hosts, refer to the AMD SMI for Virtualization documentation.

Note

AMD SMI is the successor to ROCm SMI.

Install

• Library and CLI tool installation

• Build from source

How to

• C++ library usage

• Python library usage

• Go library usage

• CLI tool usage

• AI NIC integration

• Use AMD SMI in a Docker container

Reference

• C++ API

  • Modules

  • Files

  • Globals

  • Data structures

  • Data fields

• Python API

• Go API

Conceptual

• GPU partitioning

• GPU violations

• Performance determinism

• Reliability, availability, serviceability

Tutorials

• AMD SMI examples (GitHub)

• AMD SMI CLI walkthrough

To learn about contributing to AMD SMI, see Contributing to ROCm Systems. To contribute to the documentation, see Contributing to ROCm documentation.

Find ROCm licensing information on the Licensing page.