Geared toward AI/ML purposes, Renesas’ RA8P1 MCUs leverage an Arm Ethos-U55 neural processing unit (NPU) delivering 256 GOPS at 500 MHz. The 32-bit units additionally combine twin CPU cores—a 1-GHz Arm Cortex-M85 and a 250-MHz Cortex-M33—that collectively obtain over 7300 CoreMark factors.
The NPU helps mostly used neural networks, together with DS-CNN, ResNet, Mobilenet, and TinyYolo. Relying on the neural community used, the Ethos-U55 gives as much as 35× extra inferences per second than the Cortex-M85 processor by itself.
RA8P1 microcontrollers present as much as 2 MB of SRAM and 1 MB of MRAM, which presents sooner write speeds and better endurance than flash reminiscence. System-in-package choices embody 4 MB or 8 MB of exterior flash reminiscence for extra demanding AI duties.
Devoted peripherals and superior security measures assist voice and imaginative and prescient AI, in addition to real-time analytics. For imaginative and prescient AI, a 16-bit digital camera engine (CEU) handles picture sensors as much as 5 megapixels, whereas a separate two-lane MIPI CSI-2 interface gives a low pin-count connection at as much as 720 Mbps per lane. Audio interfaces together with I²S and PDM allow microphone enter for voice AI. To guard edge AI and IoT methods, the units combine cryptographic IP, implement immutable storage, and monitor for bodily tampering.
The RA8P1 MCUs can be found now in 224-pin and 289-pin BGA packages.
The put up AI-focused MCUs embed neural processor appeared first on EDN.
