Ambarella Introduces CV3 AI Domain Controller Family To Power Autonomous Vehicles

The AI chip company, Ambarella, made the news at the CES 2022. The chip family is the latest addition to the CVflow family of scalable, power-efficient system-on-chips for the automobile sector.

The chip, according to Ambarella, offers the most fantastic AI processing performance, with up to 500 eTOPS, a 42-fold improvement over Ambarella’s previous automotive family. With up to 16 Arm Cortex-A78AE CPU cores, the CV3 boosts CPU performance by up to 30 times over the previous generation, making it ideal for autonomous vehicle (AV) software applications.

Consequently, robust advanced driver assistance systems (ADAS) and L2+ to Level 4 autonomous driving (AD) systems with higher degrees of environmental awareness for both driver seeing and machine perception in demanding lighting, weather, and driving situations have been developed.

Ambarella’s next-generation CVflow architecture debuts on the CV3, continuing the company’s algorithm-first design philosophy.

The on-chip neural vector processor (NVP), with up to 500 eTOPS of AI computation, industry-leading power efficiency, and support for the latest breakthroughs in neural network (NN) inferences, was developed as a result to the firm.

Advanced radar perception software, such as the Oculii adaptive AI software algorithms, may now be run on the NVP. The business also announced a new floating-point general vector processor (GVP) that would offload traditional computer vision and radar processing from the NVP engines, as well as floating-point-intensive algorithms from the Arm CPUs.

The CV3 family’s inherent hardware scalability, according to Ambarella, allows automakers to integrate their software stacks across their entire retail portfolios while lowering software development costs and complexity. By providing an alternative to rivals’ fragmented ADAS SoC solutions, this scalability directly tackles the increased complexity of automotive software.

Both central and zonal architectures are supported by this new family. The CV3 can execute in-cabin sensor applications, including driver and occupant monitoring while keeping the AD stack.

From basic ADAS to L4 AVs, Ambarella’s AI perception SoCs are ideally positioned to enable the full range of viewing, recording, sensing, and path planning applications. This implies that manufacturers will no longer need to design separate software stacks for their entry-level, mid-range, and luxury vehicles since Ambarella’s unified CVflow platform will be used across all models, reducing engineering costs and allowing for faster market reaction.

Ambarella’s versatile CVflow AI platform, which uses a complete set of mature software tools and a rich SDK, gives companies an unparalleled chance to develop and distinguish their goods in the industry. Ambarella has established a vast ecosystem of software partners for the CVflow platform, working directly with them to port and improve their applications, to further enhance these capabilities, according to the business.

A single CV3 can analyze the whole sensor suite, which for typical L2+ deployments includes 10 cameras, five radar modules, and multiple ultrasonic sensors while concurrently supporting up to 12 physical or virtual cameras. According to Ambarella, the CV3 also has Ambarella’s next-generation ISP, expanding the company’s advantage in image signal processing quality. Higher-performance stereo and dense optical flow engines also improve depth and motion perception.

The CV3 supports applications like 3D surround-view rendering, a hardware security module, speedier connections, and processing headroom for over-the-air upgrades on the automobile graphics processing unit (GPU).



Prathamesh Ingle is a Mechanical Engineer and works as a Data Analyst. He is also an AI practitioner and certified Data Scientist with an interest in applications of AI. He is enthusiastic about exploring new technologies and advancements with their real-life applications

🐝 Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others...