Japanese chipmaker Renesas Electronics Corp. has introduced a microcontroller unit (MCU) series for AI and machine learning (ML) applications as well as real-time analytics.
Called the RA8P1, the MCU series combines a 1 GHz Arm Cortex-M85 and 250 MHz Cortex-M33 CPU cores with the Arm Ethos-U55 neural processing unit (NPU). Renesas said this delivers more than 7300 CoreMarks and AI performance of 256 giga operations per second (GOPS) at 500 MHz.
Microcontrollers play a role in AI and ML in running light models at the edge directly-on-device without sending data to the cloud. This reduces latency, improves privacy and saves bandwidth, thus making it applicable for wearables, smart home devices and sensors.
Additionally, MCUs with integrated NPUs speed up the neural networks for faster ML capabilities.
Renesas said the RA8P1 is aimed at edge and endpoint AI applications to offload the CPU. The NPU works with networks like:
- DS-CNN
- ResNet
- Mobilenet TinyYolo
Depending on the neural network, the Ethos-U55 NPU provides up to 35 times more inferences per second than the Cortex-M85 on its own, Renesas said.
Taiwan Semiconductor Manufacturing Company (TSMC) is producing the RA8P1 MCUs manufactured on its 22 nm ultra-low leakage process technology for low power consumption with high performance. The process also includes 1 MB/512 KB embedded magnetoresistive RAM (MRAM), which enables faster write speeds, higher endurance and higher retention compared to flash, Renesas said.
“The pace of innovation in the age of AI is faster than ever, and new edge use cases demand ever-improving performance and machine learning on-device,” said Paul Williamson, senior VP and GM, IoT line of Business at Arm. “By building on the advanced AI capabilities of the Arm compute platform, Renesas’ RA8P1 MCUs meet the demands of next generation voice and vision applications, helping to scale intelligent, context-aware AI experiences.”
Other features of the series include:
- 4 MB/8 MB external Flash SIP options
- 2 MB SRAM fully ECC protected
- 32 KB I/D caches per core
- Graphics LCD controller support
- Parallel RGB and MIPI-DSI display interfaces
- 2D drawing engine
- Parallel 16 bit CEU
- MIPI CSI-2 camera interfaces
- 32 bit external memory bus (SDRAM and CSC) interface
- Gigabit Ethernet and TSN Switch
- RSIP-E50D cryptographic engine
