Nvidia creates a new computer to drive autonomous car future
10 Oct 2017 - 22:54
By Ian King / Bloomberg
Nvidia Corp. has built a powerful new computer that the chipmaker says can turn semi-autonomous cars into fully driverless vehicles, and the stock rallied.
The new mini-computer, code-named Pegasus, is 10 times more powerful than its predecessor, making it capable of handling more than 320 trillion operations per second, Chief Executive Officer Jen Hsun Huang is scheduled to say at a company conference in Germany Tuesday.
That computing muscle is needed to quickly process information on cars’ surroundings and turn that into safe driving instructions and actions. That’s especially true of Level 5 vehicles that don’t require human drivers, steering wheels, pedals or mirrors.
The Santa Clara, California-based company will offer the new product in the second half of 2018, it said. One potential application is delivery vehicles. Nvidia said Tuesday it’s helping Deutsche Post AG and its subsidiary DHL automate their fleet of electric delivery vehicles. Trials will begin next year, and eventually driverless trucks may follow human-controlled vehicles drivers around as they drop off parcels. Or the robovans might take over deliveries entirely.
Nvidia shares rose as much as 4.1 percent to $192.95 in New York. They were trading at $188.73 at 11.11 a.m., up 1.7 percent, while the Nasdaq Composite Index was little changed.
Like rivals Intel Corp. and Qualcomm Inc., the company is trying to parlay its core skill -- computer graphics processing -- into leadership in autonomous vehicles, a potential growth area for the semiconductor industry. Graphics processors are good at small manipulations of visual data in parallel, and Nvidia argues that’s ideal for the kind computing needed to make cars see the world around them.
Nvidia has learned from 225 partners, including Yandex NV, Zoox and NuTonomy, that are using its computing systems to bring autonomous capabilities to vehicles. The trunks of their current self-driving cars resemble small data centers stuffed with racks of personal computers. That’s too expensive, bulky and uses too much power for mass-production vehicles, Nvidia said. Its new Pegasus systems are a better fit, the company added.
"These are going to be the highest performing and most energy efficient for a Level 5 vehicle,” said Danny Shapiro, who heads Nvidia’s automotive chip business.
Nvidia initially expects trials of taxi and delivery vehicles on private land such as corporate campuses, said Shapiro. The company isn’t predicting when they’ll make their way onto public highways.
“The technology is going to be ready before the laws are in the United States,” he said.