Vision AI processor delivers low power, simplified cooling

By Gina Roos, editor-in-chief

Renesas Electronics Corp. has released the first product in the RZ/V series of microprocessors (MPUs) that features DRP-AI (Dynamically Reconfigurable Processor), the company’s vision-optimized artificial intelligence accelerator. The RZ/V2M offers real-time AI inference and low power consumption for embedded devices used in industrial, infrastructure, and retail applications.

Renesas said that the new devices solve the challenges faced by embedded developers of higher power consumption and heat generation required by AI processing. This is particularly the case in surveillance cameras for industrial and infrastructure as well as product scanners and POS terminal cameras that require real-time, AI-based person- and object-recognition functionality.

The RZ/V2M leverages the DRP-AI’s power efficiency to achieve power consumption as low as 4 W (typ.). This eliminates the need for heat sinks and cooling fans, which greatly simplifies heat-dissipation measures, said Renesas. It also enables the processors to be used in compact devices or help to minimize equipment sizes and reduce bill-of-materials costs. It also opens up new opportunities to incorporate AI in embedded devices.

The RZ/V2M also features an imaging signal processor (ISP) capable of processing high-resolution 4K pixels at 30 frames per second. The ISP uses high dynamic range (HDR) functionality capable of handling images with large differences between brightness and darkness, noise-reduction functionality, and distortion-correction functionality to substantially increase precision in AI recognition regardless of factors such as the weather, the time of day, and the installation location, said Renesas.

The DRP-AI vision-optimized AI accelerator is an intellectual property (IP) evolved from the DRP  built into the RZ/A2M . The new IP core is capable of AI processing with approximately 10× the power efficiency of the DRP, achieving the 1-TOPS/W class. In addition, because the DRP can dynamically change the configuration of its operation circuits every clock cycle, the DRP-AI adds the ability to support advancing AI algorithms.

Samples are available now, with mass production scheduled to start in December 2020. Renesas also plans to offer the DRP-AI Translator, a tool for DRP-AI based development that simplifies the implementation of users’ learned AI models into embedded devices.