RZ/V2M Vision-AI Microprocessor Datasheet

Overview

The DRP-AI hardware IP, which integrates DRP and AI-MAC, combines high-speed AI inference with low power consumption, achieving approximately 1.0 TOPS/W. The image signal processor (ISP) provides high robustness and produces stable images that are less affected by the environment, supporting improved AI recognition accuracy. These characteristics enable the RZ/V2M to operate at low power, which eases thermal management for embedded devices. Eliminating the need for heatsinks and cooling fans allows smaller device form factors and lower BOM cost. As a result, the device is suitable for a broad range of vision-AI applications in the embedded market, including surveillance and security, retail, office automation (OA), industrial automation, robotics, and healthcare. The RZ/V2M also includes a range of high-speed communication interfaces such as USB 3.1, PCI Express, Gigabit Ethernet, and many CPU peripheral functions, making it applicable to various use cases.

Attachment: RZ V2M datasheet for real-time human and object detection (PDF)

Features

CPU and DDR memory interfaces
- 2 Cortex-A53 cores, up to 1.0 GHz
- 32-bit LPDDR4-3200
Vision and artificial intelligence
- AI accelerator: DRP-AI, approximately 1.0 TOPS/W
- Multi-stream image signal processor (ISP)
- Camera interfaces: 2 MI-PI CSI
- Face and human detection engine
Video, graphics, and display
- H.265/H.264 multi-codec
  - Encoding: H.265 up to 2160p30, H.264 up to 1080p120
  - Decoding: H.265 up to 2160p30, H.264 up to 1080p120
- 2D graphics engine: 200 MPixels/s
- Display: MIPI-DSI (4 lanes), HDMI 1.4a
High-speed interfaces
- 1 Gigabit Ethernet
- 1x USB 3.1 Gen1 host/device
- 1x PCIe Gen2 (1 lane)
- 2x SDIO 3.0
- 1 NAND flash interface ONFI 1.0
- 1x eMMC 4.5.1
Hardware security engine provided
Package: FCBGA, 15 x 15 mm, 0.5 mm pitch