Skip to main content

Nvidia wants to power the self-driving car of the future (and every display inside that car)

Nvidia wants to power the self-driving car of the future (and every display inside that car)


There's more than one way to skin a car

Share this story

There'll be more computing horsepower inside a car than anything you currently own today." That's the assertion Nvidia CEO Jen-Hsun Huang made in announcing Drive CX, "the most advanced digital cockpit computer in the world," and Drive PX, an auto-pilot supercomputer.

Drive CX, the display one, is powered by Nvidia's Maxwell architecture and supports "every major OS in the world" — which for the world of cars includes QNX, Linux, and Android. Basically, it's a small computer inside your car capable of processing 16.6 megapixels across multiple screens. In Huang's automotive vision, everything can be a display. Smart mirrors? Check. Displays in the car's pillars? Check. Head rests? Of course — put a screen on it!

What does that mean in reality, though? So far we've seen it add different skins to your speedometer (think: Winamp). Huang is also showing a vertical screen being split into two to show dynamic menus and navigation maps, but mostly it's crazy neon and faux-metallic speedometer skins so far. At one point, Nvidia starts picking up physical slabs of material and showing off new skins.

  • Aluminum
  • Bamboo
  • More bamboo
  • Back to aluminum, I think
  • Neon lights

As Huang says, "These cannot be recreated simply by pictures. They have to be simulated, they have to be computed." And he's right — Nvidia is trying to show off how powerful its automotive computer is. A superpowered speedometer graphics is certainly one way to do.

The other part of Nvidia's car story is detection — sensors being gradually replaced by cameras. That's where Drive PX comes into play — a "supercomputer" with two Tegra X1s that can process 12 cameras in real time. Huang is heavily emphasizing "GPU-accelerated learning" here — which is to say, processor-intensive image recognition. "The beautiful thing about the neural network is that, once you train it, if you ask it 'what's a car?' it'll already know the answer for 'what's an Audi?'"

Remember: Nvidia's announcement isn't for you or me. It's ultimately for the car makers. Nvidia needs to pitch a powerful car processor to the Audis, Mercedes, and Toyotas of the world. Which, yes, includes really crazy looking, faux-neon-and-bamboo gauges and "computer vision" that can detect different breeds of dogs.

Audi has, in fact, committed to Nvidia's vision, sharing stage time with Nvidia and talking up fast-driving autonomous cars and display-laden interiors. To which Huang said, as you might've predicted already, "Anybody that needs more pixels is just music to my ear. And don't forget the backseat!"