Renesas has launched its most advanced microcontroller yet, the RA8P1, designed for engineers building the next generation of AI-powered embedded systems. With a 1GHz Arm Cortex-M85 core, a 250MHz Cortex-M33, and an on-chip Arm Ethos-U55 NPU, this device delivers over 7300 CoreMarks and up to 256 GOPS of AI compute, offering a significant performance boost in a compact, power-efficient form factor.
Built using TSMC’s 22nm ultra-low-leakage process, the RA8P1 balances processing power with energy efficiency. The embedded NPU is optimised to accelerate neural network tasks like CNNs and RNNs, delivering up to 256 MACs per cycle. Models such as DS-CNN, ResNet and Mobilenet are supported natively, and in the right setup, the NPU can deliver inference speeds up to 35 times faster than using the CPU alone.
The move to embedded MRAM is also notable. Compared to traditional Flash, MRAM offers faster writes, better endurance and improved retention, key benefits for applications needing consistent performance over a long lifetime.
For edge AI applications that involve real-time data, Renesas has equipped the RA8P1 with a comprehensive set of interfaces. It supports camera inputs via a 16-bit CEU or MIPI CSI-2, audio via I2S and PDM, and offers 2MB of ECC-protected SRAM for buffer-intensive tasks. There’s also 1MB of MRAM on-chip and support for high-speed external memory or SIP packages with up to 8MB of flash, depending on application requirements.
To simplify deployment, Renesas has introduced a new AI software framework called RUHMI. It’s designed to make it easier for engineers to integrate models built in TensorFlow Lite, PyTorch or ONNX. The framework handles model conversion, quantisation and optimisation, and is fully integrated with Renesas’s e2Studio IDE, streamlining the path from AI model to embedded product.
Security hasn’t been overlooked. The RA8P1 includes the RSIP-E50D cryptographic engine and supports modern standards including ECC, RSA and SHA3. With Arm TrustZone, secure boot from immutable storage, and real-time decryption from external memory, the device is built with system-level security in mind.
Renesas backs the RA8P1 with strong ecosystem support. Developers can take advantage of its Flexible Software Package (FSP), FreeRTOS, Azure RTOS and Zephyr OS support, plus a range of evaluation kits and application examples. Pre-validated “Winning Combination” designs, including AI-enabled cameras, robotic arms and smart conferencing devices, help reduce development time and project risk.
1GHz Arm Cortex-M85
500MHz Arm Ethos-U55 NPU
Optional 250MHz Arm Cortex-M33
1MB / 512KB embedded MRAM
2MB ECC-protected SRAM
SIP options: 4MB / 8MB external flash
32KB I/D caches per core
WXGA (1280×800) LCD controller
Parallel RGB and MIPI-DSI display interfaces
2D drawing engine
Camera input via CEU (16-bit) and MIPI CSI-2
32-bit external memory interface (SDRAM, CSC)
Gigabit Ethernet with TSN
USB FS/HS, CAN-FD, I2C/I3C, SPI, SDHI
Audio: PDM, SSI
Analog: 16-bit ADC with S/H, DAC, comparators, temperature sensor
Timers, XSPI with execute-in-place and decryption-on-the-fly
RSIP-E50D crypto engine
Arm TrustZone
Secure boot from immutable storage
DPA/SPA protection, secure debug and factory programming
Device Lifecycle management
With its high-performance CPU and NPU, embedded MRAM, and robust security features, the RA8P1 sets a new benchmark for edge AI microcontrollers. Its combination of compute power, connectivity, and software support enables engineers to design more intelligent, responsive embedded systems without sacrificing efficiency or flexibility.
Related article: Renesas Launches Gen IV Plus 650V GaN FETs
Learn more and read the original article on www.renesas.com
