Nvidia’s newest AV processor: 30 trillion operations per second on 30 watts

  • 11-Jan-2018 11:11 EST
nvidia xavier.jpg

Nvidia's high-powered, low-consumption Xavier processor (image: Bill Visnic).


It was two “P’s”—processing and partnerships—that headlined the 2018 Consumer Electronics Show for graphics-processing giant Nvidia, as the company introduced its most-powerful system-on-a-chip (SOC) to enable high-function automated driving and announced autonomous-vehicle development partnerships with Volkswagen, ZF, Uber and recent startup Aurora Innovation.

Unveiling the new Xavier processor, Nvidia CEO Jensen Huang said it can handle up to 30 trillion operations per second—yet answers concerns regarding the power consumption of automated-vehicle computing and the numerous sensors and other associated hardware by drawing just 30 watts of power—making it 15 times more efficient than its predecessor, the company claims.

The Xavier chip board, about the size of a license plate, is the processing “engine” for the company’s lineup of DRIVE automated-driving software “stacks,” as well as Nvidia’s in-house artificial-intelligence (AI) and augmented-reality systems. Huang said Nvidia will begin delivering Xavier samples to customers in the first quarter and will start series production sometime later in 2018.

“The complexity of future cars is incredible. It begins with Xavier, which can do deep learning, computer vision and high-performance computing at highly efficient levels,” said Huang.

The Xavier board incorporates some 9 billion transistors, the company said, calling it “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.”

But as seems to be the customary business model for chip developers, there’s something in the wings even more powerful.

Robo-flying with Pegasus

Nvidia also revealed at CES 2018 the Pegasus, which it billed as “the world's first artificial-intelligence computer designed to drive fully-autonomous robotaxis.” It works with the company’s DRIVE PX artificial-intelligence platform to enable SAE Level 5 autonomy—without a driver.

Pegasus, essentially made up of dual Xavier chipsets, has almost 10% more capacity than Xavier, processing more than 320 trillion operations per second, a figure that’s also more than ten times the performance of its predecessor, Nvidia’s DRIVE PX 2.

“Nvidia DRIVE PX Pegasus will help make possible a new class of vehicles that can operate without a driver—fully autonomous vehicles without steering wheels, pedals or mirrors, and interiors that feel like a living room or office. They will arrive on demand to safely whisk passengers to their destinations, bringing mobility to everyone, including the elderly and disabled,” the company said in a release accompanying Pegasus’ CES unveiling.

Nvidia said more than 25 of its 225 partners already using the DRIVE PX platform are developing robotaxis using the company’s CUDA GPUs. But the systems’ trunk-filling size, cost and power requirements make them impractical for production vehicles, the company claims. Instead, Pegasus is a fraction of the size and draws much less power.

An Nvidia engineer told Automotive Engineering the Pegasus processor draws about 500 watts.

"Creating a fully self-driving car is one of society's most important endeavors—and one of the most challenging to deliver," said Jensen Huang, NVIDIA founder and CEO. "The breakthrough AI computing performance and efficiency of Pegasus is crucial for the industry to realize this vision.”

New AV partners

At the opening of the CES gathering, Nvidia announced new development partnerships with Volkswagen, ZF, Uber and Aurora Innovation; VW said at the announcement that it will use the Drive IX platform to “create new cockpit experiences and improve safety,” in coming new models; VW pointed to the I.D. Buzz concept, a reimagining of the classic Microbus, as a vehicle that will draw on AI for new user-interface safety advances.

The companies said AI-driven deep learning will enable future vehicles to assess situations and analyze the behavior of others on the road, allowing driver-assist or autonomy functions to make proper decisions with the required high degree of accuracy.

Aurora, based in Palo Alto, CA and Pittsburgh, PA. It bills itself as a company that “designs and builds self-driving technology, partnering with automakers to integrate, pilot and deploy advanced self-driving platforms into vehicles.”

In a release, VW said the “collaboration with Aurora will provide valuable experience with the world-class engineering team to the ongoing development of software and hardware for driverless vehicles, and, additionally, mobility services for urban and rural areas.”


Share
HTML for Linking to Page
Page URL
Grade
Rate It
4.67 Avg. Rating

Read More Articles On

2017-08-09
A recent Engineering program at the University of Michigan’s Dearborn campus jettisoned lectures and text books and replaced them with a fresh Design, Build, and Test curriculum including new lean-Startup courses addressing customer discovery techniques.
2017-08-18
Delphi positions itself for wider access to emerging autonomous-driving technology with an investment in Israel-based Lidar developer Innoviz Technologies.
2017-05-08
Artificial intelligence has the potential to vastly advance automated driving and other aspects of automotive safety. But developers must determine how to mitigate undesirable side-effects.
2017-06-16
It's possible to thwart most, but not all, cyber incidents.

Related Items

Technical Paper / Journal Article
1990-02-01
Training / Education
2018-03-12
Technical Paper / Journal Article
1990-02-01
Training / Education
2017-09-06