SiFive and AMD demonstrated a system combining a RISC-V processor with an AMD Radeon AI PRO R9700 graphics card at the AI Infra Summit. The setup utilized ROCm 10.0 software to run the Google Gemma4-E2B model.

The collaboration aims to optimize ROCm for RISC-V server architectures. The companies intend to increase processing speeds and expand support for larger AI models and additional acceleration scenarios in data centers.

SiFive is integrating the ROCm ecosystem into its BigSky development platform. This 2U rack-mountable server is designed for enterprise use, allowing hyperscalers and developers to run AI workloads on RISC-V hardware.

This demonstration marks an initial step in enabling developers to implement ROCm-based AI acceleration on RISC-V host platforms. The companies plan to continue evaluating optimization solutions for data center applications.