Forlinx releases 20-TOPS M.2 AI accelerator for local LLM inference

Forlinx Embedded has introduced an M.2 AI accelerator card designed for local AI inference on embedded systems. The module is built on Rockchip’s RK1820 and RK1828 processors, delivering 20 TOPS of INT8 computing performance.

Key specifications include:

• Memory: Up to 5GB of integrated DRAM.
• Interface: Standard M.2 2280 form factor.
• Scalability: Supports PCIe cascading to increase compute capacity.

The accelerator targets workloads such as large language models (LLMs), vision-language models, and computer vision tasks. It is compatible with embedded Linux and Android operating systems, offering a hardware solution for edge AI applications.