Renesas Launches 1-GHz MCUs with Advanced AI Acceleration

Renesas Electronics Corporation has unveiled the RA8P1 microcontroller (MCU) group, designed for AI and machine learning applications, as well as real-time analytics. The MCUs set a new performance benchmark by integrating a 1 GHz Arm Cortex-M85 and a 250 MHz Cortex-M33 CPU core with the Arm Ethos-U55 Neural Processing Unit (NPU). This combination delivers over 7300 CoreMarks in CPU performance and 256 GOPS in AI performance at 500 MHz.

The RA8P1 MCUs are optimized for edge and endpoint AI applications. The Ethos-U55 NPU offloads the CPU for compute-intensive operations in Convolutional and Recurrent Neural Networks (CNNs and RNNs), delivering up to 256 MACs per cycle, resulting in 256 GOPS performance at 500 MHz. The NPU supports popular networks like DS-CNN, ResNet, MobileNet, TinyYolo, and more, providing up to 35 times more inferences per second than the Cortex-M85 processor alone.

The MCUs feature extensive memory and security capabilities, with dedicated peripherals for voice and vision AI applications. A 16-bit camera interface (CEU) is included for vision AI, supporting sensors up to 5 megapixels. This enables camera applications and demanding vision AI tasks. A separate MIPI CSI-2 interface provides a low-pin-count interface with two lanes, each capable of up to 720 Mbps. Audio interfaces, including I2S and PDM, support microphone inputs for voice AI applications.

The RA8P1 offers both on-chip and external memory options for efficient, low-latency neural network processing. It includes 2MB SRAM for storing intermediate activations or graphics framebuffers and 1MB of on-chip MRAM for application code and storage of model weights or graphics assets. High-speed external memory interfaces are available for larger models, and SIP options with 4 or 8 MB of external Flash memory are available in a single package for more demanding AI applications.

The RA8P1 MCUs are manufactured using TSMC's 22ULL (22nm ultra-low leakage) process, which enables ultra-high performance with very low power consumption. This process also supports the use of embedded Magnetoresistive RAM (MRAM), offering faster write speeds, higher endurance, and better retention compared to Flash.

Renesas has also introduced RUHMI (Renesas Unified Heterogenous Model Integration), a comprehensive framework for MCUs and MPUs. RUHMI provides efficient AI deployment of the latest neural network models in a framework-agnostic manner, enabling model optimization, quantization, graph compilation, and conversion. It generates efficient source code and supports machine-learning AI frameworks such as TensorFlow Lite, Pytorch, and ONNX. RUHMI is integrated with Renesas's e2 Studio IDE, facilitating seamless AI development.

The RA8P1 MCUs feature the new Renesas Security IP (RSIP-E50D), which includes various cryptographic accelerators, such as CHACHA20, Ed25519, NIST ECC curves up to 521 bits, enhanced RSA up to 4K, SHA-2, and SHA-3. In conjunction with Arm TrustZone, this provides a comprehensive and integrated secure element-like functionality. The MCUs also provide a strong hardware Root-of-Trust and Secure Boot with a First Stage Bootloader (FSBL) in immutable storage. XSPI interfaces with decryption-on-the-fly (DOTF) enable encrypted code images for storage in external Flash, which are then securely transferred to the MCU for execution.

Renesas offers a comprehensive range of user-friendly tools and solutions for the RA8P1 MCUs, including the Flexible Software Package (FSP), evaluation kits, and development tools. FreeRTOS and Azure RTOS are supported, as is Zephyr. Several Renesas software example projects and application notes are also available.