Skip to content

Hardware requirements

Server

The server needs a GPU for the ML pipeline. The two reference configurations:

PlatformGPUAcceleration
LinuxAMD RDNA3 (e.g. RX 7900 XTX)ROCm torch + Vulkan llama.cpp
WindowsNVIDIA (e.g. RTX 5090)CUDA
  • VRAM: TODO
  • RAM: TODO
  • Disk: TODO (model downloads on first launch)

Satellites

Satellites are lightweight — they only capture and play audio: