Friday, April 19, 2024

NVIDIA DRIVE AGX Orin, a software-defined platform for autonomous vehicles

Autonomous cars and robots are a future within reach. However, it requires computational power that is still not common. NVIDIA DRIVE AGX Orin is one of the solutions.

On the occasion of the GPU Technology Conference, which was held in Suzhou (China) from December 16 to 19, the co-founder and CEO of NVIDIA, Jensen Huang, gave some characteristics of DRIVE AGX Orin, a software-defined platform which is intended to autonomous vehicles and robots.

The novelty runs on the new SoC (System-on-Chip) called Orin, which consists of 17 billion semiconductor transistors. Orin integrates the NVIDIA’s next-generation GPU architecture and Arm Hercules CPU cores, as well as new deep learning and computer vision accelerators. It is designed to achieve 10% more power and efficiency than those of the previous generation. It will have the capacity to deliver 200 trillion operations per second, which, according to the company itself, is seven times the performance of Xavier SoC.

Orin is designed to work with a large number of applications and a neural network at the same time – a common situation in autonomous cars or robots. The solution also meets systematic safety standards such as ISO 26262 ASIL-D.

NVIDIA DRIVE AGX Orin has acquired the attribute of the software solution by combining and communicating with a wide range of architecturally compatible platforms. In practice, this means it is suitable for autonomous vehicles from a Level 2 to full self-driving Level 5 vehicle, enabling OEMs to develop large-scale and complex groups of software products.

Both Xavier and the new Orin are programmable through the CUDA and TensorRT APIs and libraries, allowing developers to work with solutions across multiple generations of products.

NVIDIA DRIVE AGX Orin will be part of a range of configurations based on a single architecture and should appear in autonomous cars from 2022 onwards.