Renesas has introduced the RZ/V2N AI MPU at Embedded World 2025, addressing the challenge of heat dissipation in AI-powered embedded systems. While AI processing is traditionally associated with power-hungry cloud and data center applications, Renesas is bringing high-performance AI to endpoint devices without the need for bulky heat sinks or cooling fans. This makes the RZ/V2N an attractive solution for space-constrained and cost-sensitive embedded applications. More details can be found at https://www.renesas.com
A key highlight of the RZ/V2N is its ability to handle real-time AI workloads, including image processing and video encoding. A demonstration at the show featured two MIPI-CSI cameras running H.265 video encoding and AI-powered image analysis, streaming in parallel to a monitor—all without requiring active cooling. Thermal imaging revealed an operating temperature of just 40°C, proving its energy efficiency.
This efficiency is achieved through Renesas’ third-generation AI accelerator, DRP-AI (Dynamically Reconfigurable Processor for AI). The RZ/V2N delivers 10 TOPS per watt, reaching a peak of 15 TOPS in sparse mode and 4 TOPS in dense mode. This level of power efficiency is made possible by a unique unstructured pruning technique, which optimizes AI model execution to reduce computational load while maintaining performance.
Unlike general-purpose AI architectures such as ARM Cortex cores, Renesas relies on its proprietary DRP-AI technology, which has evolved significantly since the previous generation. The RZ/V2N marks a tenfold improvement in AI performance per watt, compared to its predecessor. This breakthrough enables completely fanless and heat sink-free designs, making it ideal for embedded applications that demand low power consumption and compact form factors.
The RZ/V2N supports full Linux and Android environments, with software development based on the Yocto Project. To ease AI adoption, Renesas provides an open-source AI toolchain and a library of over 50 ready-to-use AI applications on GitHub. These include people counting, smart building monitoring, and industrial safety solutions. Users can leverage transfer learning to retrain models with their own data and deploy optimized ONNX models using Renesas’ DRP-AI TVM compiler.
Manufactured on a 14nm process node—compared to 28nm in previous generations—the RZ/V2N benefits from improved power efficiency and performance. It also introduces a new low-power mode, further optimizing energy consumption for battery-powered embedded applications. This makes it well-suited for IoT devices, industrial automation, and smart vision systems.
For applications requiring real-time processing, Renesas’ flagship RZ/V2H features an additional dual-core Cortex-R8 and an OpenCV accelerator, making it suitable for robotics and drone control. However, the RZ/V2N fills a crucial gap between its predecessor and the high-end V2H, offering an optimal balance of AI performance and efficiency for a broader range of embedded use cases.
The RZ/V2N is now in mass production, with evaluation boards and samples available starting March 19th. Developers interested in exploring its capabilities can access resources and sample applications on Renesas’ website.
Check out all my Embedded World 2025 videos in this playlist: https://www.youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC using the dual wireless DJI Mic 2 microphones with the DJI lapel microphone https://amzn.to/3XIj3l8 ), watch all my DJI Pocket 3 videos here https://www.youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Join https://www.youtube.com/charbax/join for Early Access to my videos and to support my work, or you can Click the “Super Thanks” button below the video to send a highlighted comment!
