Skip to content

AMD ROCm GPU support #10

Description

@pmady

Currently the server only supports NVIDIA GPUs via NVML. Adding AMD ROCm support would make gpu-mcp-server vendor-neutral.

Requirements:

  • Implement a new ROCmCollector that satisfies the gpu.Collector interface
  • Use rocm-smi CLI or AMD ROCm SMI library for GPU metrics
  • Auto-detect GPU vendor at startup and choose the right collector
  • Maintain parity with NVML metrics where possible (utilization, memory, temperature, power)
  • Add build tag //go:build cgo && linux (ROCm is Linux-only)

Challenges:

  • ROCm has no official Go bindings — may need to shell out to rocm-smi --json or use CGO with librocm_smi64.so
  • Some metrics may not have 1:1 equivalents between NVML and ROCm

Difficulty: Hard — requires AMD GPU hardware for testing

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requestgpuGPU hardware relatedhelp wantedExtra attention is needed

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions