According to foreign media reports, semiconductor solutions provider Renesas Electronics Corporation announced the launch of the RA8P1 microcontroller (MCU) series, which is aimed at artificial intelligence (AI) and machine learning (ML) applications and real-time analysis. The new MCUs combine 1GHz Arm® Cortex-M85® and 250MHz Cortex-M33 CPU cores with Arm Ethos-U55™ Neural Processing Units (NPUs) to take MCU performance to the next level. This combination delivers the highest CPU performance of over 7300 CoreMarks, as well as AI performance of 256 GOPS at 500 MHz.
Image source: Renesas
Designed for edge/endpoint AI
Optimized for edge and terminal AI applications, the RA8P1 leverages the Ethos-U55 NPU to reduce CPU burden, enabling it to handle compute-intensive operations in convolutional neural networks (CNNs) and recurrent neural networks (RNNs), delivering up to 256 MACs per cycle and 256 GOPS performance at 500 MHz. The new NPU supports the most commonly used networks, including DS-CNN, ResNet, MobileNet, TinyYolo, and more. According to the neural network used, the Ethos-U55 inference speed is 85 times faster than the Cortex-M35 processor itself.
Advanced technology
The RA8P1 MCUs are manufactured using TSMC’s 22ULL (22nm ultra-low leakage) process, which achieves ultra-high performance and extremely low power consumption. The process also supports the use of embedded reluctance RAM (MRAM) in new MCUs. MRAM offers faster write speeds, increased endurance, and data retention compared to flash memory.
“The demand for high-performance edge AIoT applications is exploding,” said Daryl Khoo, vice president of marketing for embedded processing at Renesas. We are excited to introduce the MCU that we believe is best suited for this trend. The RA8P1 device demonstrates our technical and market expertise and highlights our strong partnerships across the industry. Customers are eager to use these new MCUs in a variety of AI applications. ”