Nvidia says its new mobile processor will help power a particularly mobile machine: the self-driving car.
Over the weekend, the chip maker unveiled its new Tegra K1 mobile processor at the Consumer Electronics Show in Las Vegas. The chip is based on the company's Kepler architecture, which is traditionally used by the chips that power some of the world's most efficient supercomputers, but the K1 consumes even less power than previous Kepler models, making it suitable for phones, tablets, and, yes, autonomous cars.
Autonomous cars may seem like something in the distant future. But Google has spent years testing self-driving cars on public roads, and your own car is at least approaching this brave new world. "You already have some autonomous elements in cars, like cruise control, pedestrian detection, parking assistance and blind spot monitoring," says Nvidia director of automotive marketing Danny Shapiro.
Nvidia already offers chips that can help with such automobile applications, and its new chip, he says, will serve a similar role -- at least initially. But the company harbors even bigger ambitions.
Truly autonomous vehicles require sophisticated machine vision engines that can do stuff like detect a pedestrian in the middle of the street. Your brain handles this sort of visual processing by splitting the task among billions of brain cells, and Nvidia aims to do much the same thing with its new supercomputing chip, which can spread tasks across dozens of processor cores.
The company will also offer programming tools that let automakers build a wide range of applications that run on the new chip. Right now, most parking assistance and blind spot detection tools run on their own specialized hardware and software, Shapiro says, and this complicates things for automakers. With the Tegra K1, the company aims to create a common set of tools for car software, much as Apple has done with the iPhone and the iPad.
The K1 will come in two varieties: a 32-bit chip and a 64-bit chip -- Nvidia's fist mobile 64-bit processor. While Nvidia's existing Kepler supercomputer chips include 2,880 processors cores, the K1 will have only 192. But that's enough, Shapiro says, to provide sophisticated processing on cars as well as smartphones. The K1 should be available for tablets, mobile phones, and the like during the first quarter of this year. But Shapiro says the automotive version will take some more time due to the industry's rigorous certification requirements. The self-driving car is coming, but it's not here yet.
