With over 7,300 CoreMarks, integrated MRAM, advanced peripherals, and a new AI framework, the RA8P1 sets a milestone in embedded computing for the edge.
Tokyo, Japan – Japanese company Renesas Electronics Corporation this month announced its new RA8P1 microcontroller (MCU) family, establishing a new performance standard for AI applications at the edge. Equipped with a 1 GHz Arm Cortex-M85 core, a 250 MHz Cortex-M33, and an Arm Ethos-U55 neural processing unit, these devices deliver unprecedented performance of more than 7,300 CoreMarks and 256 GOPS for AI tasks.
Designed to meet the growing demand for high-efficiency AI in embedded applications and Internet of Things (IoT), the new MCUs combine processing power, low power consumption, advanced storage, and robust security on a single platform.
Advanced architecture for AI at the edge
The RA8P1 is optimized to run convolutional neural networks (CNN) and recurrent neural networks (RNN) directly on the device, utilizing the Ethos-U55 NPU to offload the most intensive tasks from the main processor. This coprocessor enables models like DS-CNN, ResNet, MobileNet, or TinyYolo to run with up to 35 times more inferences per second compared to the Cortex-M85 alone.
Manufactured with TSMC’s 22ULL process (22 nm ultra-low leakage), the RA8P1 offers a perfect balance of extreme performance and energy efficiency. It also includes embedded MRAM, a non-volatile memory with faster write speeds, greater durability, and retention than traditional flash memory—essential for applications requiring quick and reliable persistence.
Vision and voice peripherals
To enable advanced vision AI capabilities, the RA8P1 features a 16-bit camera interface compatible with sensors up to 5 megapixels, along with a dual-channel MIPI CSI-2 interface (up to 720 Mbps per channel), ideal for surveillance and smart camera applications. In audio, the microcontroller integrates I2S and PDM interfaces, making microphone integration for voice recognition tasks straightforward.
It also includes 2 MB of ECC-protected SRAM for real-time processing and 1 MB of MRAM for storing code or model weights. For larger models, options with 4 MB or 8 MB of external flash via SIP packages are available.
RUHMI: new embedded AI framework
Renesas accompanies this launch with RUHMI (Renesas Unified Heterogeneous Model Integration), a development framework that enables efficient implementation of AI models on both MCUs and MPUs, with native support for TensorFlow Lite, PyTorch, and ONNX. RUHMI offers tools for optimization, quantization, compilation, and model conversion, generating optimized code without relying on specific frameworks.
Integrated into the e²Studio development environment, RUHMI provides a unified platform for embedded AI development, significantly reducing development and time-to-market.
Advanced security for critical applications
The RA8P1 also stands out for its cutting-edge security features, vital for sectors like automotive, healthcare, and industrial. It incorporates the new RSIP-E50D cryptographic engine, supporting algorithms such as CHACHA20, Ed25519, RSA up to 4,096 bits, and SHA3, along with Arm TrustZone® technology. Features include secure boot with immutable storage, tamper protection, and secure debugging.
It also supports XSPI interfaces with real-time decryption (DOTF), enabling secure storage and execution of encrypted images directly from external memory.
Integrated solutions and broad ecosystem
These new MCUs are part of over 400 “Winning Combinations” by Renesas—verified designs combining company’s devices to accelerate product development. They include solutions for AI-powered video conferencing cameras, AI-enabled drawing robotic arms, and intelligent surveillance systems.
The development ecosystem is further strengthened by support for FreeRTOS, Azure RTOS, Zephyr, evaluation kits, integrated tools, application notes, and collaborations with partners like Nota.AI (driver monitoring) and Irida Labs (traffic and pedestrian detection).
MCU leadership and availability
Renesas maintains its leadership in the MCU industry with over 3.5 billion units shipped annually, nearly 50% of which serve the automotive sector. With the broadest portfolio of 8, 16, and 32-bit MCUs on the market, the Japanese company continues to position itself as a key player in developing intelligent, secure, and efficient solutions for the connected world.
With the RA8P1, Renesas not only raises the bar for MCU performance but also reinforces its vision of accessible, efficient, and secure embedded AI at all levels of edge computing.
via: renesas