NVIDIA delivering Xavier processors, expands DRIVE software to three platforms; partnering with Aurora, Uber and ZF/Baidu
NVIDIA is delivering first samples of its Xavier processors (earlier post) to customers this quarter. Xavier, an all-new AI supercomputer designed for use in self-driving cars, is a complete system-on-chip (SoC), integrating a new 512-core Volta GPU, a custom 8-core CPU architecture, and a new computer vision acceleratora, new deep learning accelerator, and new 8K HDR video processors.
Xavier will power the NVIDIA DRIVE software stack, now expanded to a trio of AI platforms covering every aspect of the experience inside next-generation automobiles. DRIVE Xavier delivers 30 trillion operations per second (TOPS) while consuming just 30 watts. Xavier is 15 times more energy efficient than the previous generation architecture.
With more than 9 billion transistors, Xavier is the most complex system on a chip ever created, representing the work of more than 2,000 NVIDIA engineers over a four-year period, and an investment of $2 billion in research and development.
Complementing the original NVIDIA DRIVE AV autonomous vehicle platform, which uses neural networks to let cars drive themselves, NVIDIA unveiled at CES two new software platforms: DRIVE IX and DRIVE AR.
DRIVE IX, an intelligent experience software development kit, will enable AI assistants for both drivers and passengers, using sensors inside and outside the car.
DRIVE AR is the augmented reality SDK; the software platform for the AI car will fuse computer vision, computer graphics and AI. DRIVE AR will enable next-generation augmented reality interfaces that deliver information points of interest along a drive, create alerts and navigate safely and easily.
Pegasus. Xavier is a key part of the NVIDIA DRIVE Pegasus AI computing platform. (Earlier post.) Delivering the performance of a trunk full of PCs in an auto-grade form factor the size of a license plate, it’s the world’s first AI car supercomputer designed for fully autonomous Level 5 robotaxis.
With an unprecedented 320 TOPS of deep learning calculations and the ability to run numerous deep neural networks at the same time, Pegasus will provide everything needed for safe autonomous driving.
Pegasus is built on two Xavier SoCs and two next-generation NVIDIA GPUs. Customers will get the first samples of Pegasus with its 320 trillion operations per second of processing performance in mid-2018.
More than 25 companies are already using NVIDIA technology to develop fully autonomous robotaxis; Pegasus will be the path to production.
Aurora, Uber, and ZF/Baidu. At CES, NVIDIA also announced partnerships with Aurora, Uber and ZF/Baidu.
Aurora—which itself just announced partnerships with Hyundai (earlier post) and Volkswagen (earlier post)—is working with NVIDIA to create a new Level 4 and Level 5 self-driving hardware platform that will use the NVIDIA DRIVE Xavier processor.
Uber utilizes NVIDIA technology for Uber Advanced Technologies Group’s fleets of self-driving cars and freight trucks, running AI algorithms that enable vehicles to perceive the world, predict what will happen next and quickly choose the best course of action, even in complex environments.
Uber began using NVIDIA GPU computing technology in its first test fleet of Volvo XC90 SUVs, and currently uses high-performance NVIDIA processors to run deep neural networks in both its self-driving ride-hailing cars and self-driving freight trucks. The development pace of the Uber fleet has accelerated dramatically, with the last million autonomous miles being driven in just 100 days.
NVIDIA, ZF and Baidu are creating a production-ready AI autonomous vehicle platform designed for China, the world’s largest automotive market. The collaboration is based on Xavier, ZF’s new ProAI car computer and Baidu’s Apollo Pilot, an autonomous driving product targeted for mass production.
ZF’s new Xavier-based ProAI will process inputs from multiple cameras, plus lidar and radar, paint a 360-degree view around the vehicle, locate it on an HD map, and find a safe path through traffic.
Baidu’s Apollo open autonomous driving platform provides a comprehensive, secure and reliable all-in-one solution that supports all major features and functions of an autonomous vehicle. Apollo Pilot is an autonomous driving product targeted for mass production, which provides a safe, economical and comfortable autonomous driving experience. It’s built on Baidu’s technology and insight from the driving behaviors of Chinese users.
This solution is created not only for Chinese automakers, but any vehicle maker planning to do business in China, and can scale from super Level 2 automated driving capabilities to Level 5 fully autonomous robotaxis. Production vehicles utilizing AI self-driving technologies from NVIDIA, ZF and Baidu are expected on the roads of China starting in 2020.