The V3U provides 60 TOPS with low power consumption for deep learning processing and up to 96,000 DMIPS raw performance for the architectures driving next-generation autonomous vehicles.
The new SoC is the first to use Renesas’ R-Car Gen 4 architecture in the company’s autonomy platform for ADAS and AD. The introduction of R-Car V3U means the platform can now provide complete scalability from entry-level NCAP applications up to highly automated driving systems.
“We are excited to introduce the newest generation of our popular R-Car SoCs for the next generation of ADAS and AD vehicles,” said Naoki Yoshida, Vice President, Automotive Digital Products Marketing Division at Renesas. “The R-Car V3U leverages assets developed on previous-generation devices, such as ADAS and Level 2 perception stack with the R-Car V3M and R-Car V3H, along with the Renesas autonomy platform, to offer a smooth migration path to single-chip Level 3 automated driving with short development turnaround and safe production launch.”
R-Car V3U SoC is expected to achieve ASIL D metrics for the majority of the SoC processing chain. The device can also cut design complexity, time to market, and system cost.
The R-Car V3U SoC features highly flexible DNN (Deep Neural Network) and AI machine learning functions. The device offers a flexible architecture that can handle any state-of-the-art neural networks for automotive obstacle detection and classification tasks while maintaining 60 TOPS with low power consumption and an air cooling system.
As well as impressive AI performance, the R-Car V3U has a broad range of programmable engines, including DSP for radar processing, a multi-threading computer vision engine for traditional computer vision algorithms, image signal processing for anhanced image quality, and additional hardware accelerators for key algorithms.
Renesas’ IDE allows users to take advantage of the R-Car platform’s built-in hardware features for a faster time to market for computer vision and deep learning-based solutions. Easy-to-use debugging and tuning tools for heterogeneous multi-core hardware allow efficient software