ROHM Unveils Ultra-Low-Power On-Device Learning Edge AI Chip
ROHM's new AI chip enables real-time failure prediction in IoT devices with ultra-low power consumption, eliminating the need for cloud servers. The chip, featuring tinyMicon MatisseCORE™, paves the way for efficient edge computing.
ROHM Develops Ultra-Low-Power On-Device Learning Edge AI Chip
ROHM has introduced an innovative AI chip designed for edge computing in IoT applications. This on-device learning AI chip allows real-time failure prediction in devices equipped with motors and sensors while maintaining ultra-low power consumption, all without relying on cloud servers.
The chip incorporates an AI accelerator based on a proprietary algorithm developed by Professor Matsutani of Keio University, combined with ROHM's high-efficiency 8-bit CPU, tinyMicon MatisseCORE™. This technology enables AI learning and inference on the same chip, drastically reducing power consumption by up to 1000× compared to conventional AI chips.
Features
- Ultra-low power consumption suitable for edge computing
- On-device learning enables real-time failure prediction without cloud server dependency
- AI chip integrates a 20,000-gate AI accelerator with ROHM’s tinyMicon MatisseCORE™ CPU
- Ideal for IoT devices with motors and sensors
Applications
- IoT devices, particularly those equipped with motors and sensors for real-time failure prediction and monitoring.
- Edge computing for predictive maintenance in smart factories, vehicles, and other connected devices.
Block diagram
