Nvidia’s press conference Sunday was like a tale of two missions: Announce its new Tegra X1 “mobile super chip,” a processor so powerful it could put Xbox One-worthy graphics on a smartphone. And then reveal where the company wants to put it first…in mobiles, yes, but of the auto variety.
According to CEO Jen-Hsun Huang, the Tegra X1—built on the Maxwell architecture it unveiled last year—is twice as fast as its lauded predecessor, the Tegra K1. The tiny chip is also energy efficient, which should make it a natural fit for mobile devices.
Too bad it’s not heading to smartphones or tablets. But mobile’s loss could be automotive’s gain. Because the chip could power Nvidia’s vision of self-driving cars, using a system of sensors and cameras.
Chips Ahoy
The Tegra X1, Nvidia claims, can handle a teraflop of computing power. For comparison, that would give the world’s fastest supercomputer from 2000 could a run for its money. Although that may not seem smoking fast by today’s standards, the company claims the eight-core, 64-bit chip does not lack for performance.
To illustrate the X1’s chops, Nvidia showed off a demo of a smartphone running a video built with Unreal Engine 4, a tool used to build graphic-intensive games. The demo worked well, which presumably speaks to the chip’s capabilities.
Unfortunately, Nvidia doesn’t think smartphones can handle the X1’s computational power yet. So instead, it’s taking the X1 to carmakers.
The company announced the Drive CX, a “digital cockpit computer” that brings simulated graphics, realistic finishes—like bamboo or aluminum finishes—and contextual data to gauges, maps and in-dashboard displays.
Nvidia also unveiled Drive PX, a new platform powered by a couple of X1 chips, for 2.3 teraflops of computing power that can use high-powered graphics from sensor- and camera-festooned cars to enable autonomous driving.
Driving The Future Of Smarter Cars
Automobiles will boast more processing power “than anything you currently own today,” Huang said.
To explain what he meant, the exec veered into somewhat academic territory, over-explaining the nature and merits of computer learning, neural networks and specifically “GPU-accelerated learning”—a fancy way of describing processor-intensive image recognition technology that can interpret results and make decisions.
But his enthusiasm, and his company’s vision, were plain: Nvidia sees X1-powered cars that can park themselves, drive on their own and not only stop for animals, but can even tell you what breed of dog has skipped into your path. The chipmaker believes it has the super-fast processor capable of the sort of detailed graphics necessary for split nano-second decisions.
Audi appears to agree. The carmaker joined Nvidia on stage to wax poetic about autonomous cars and graphics-festooned vehicle interiors—hinting that our rides may be on the verge of accelerating into the future.
Photos by Adriana Lee for ReadWrite