NVIDIA Drive PX 2: a super-computer with 12 CPU cores and four Nvidia Pascal GPU chips for self-driving cars. Credits: nvdia.com
LAS VEGAS — Like the tradition at the annual CES2016 (Consumer Electronics Show), Nvidia opened by hinting at the major hot topics of the week.
In the past the top topics like Mobile Gaming, Wearable Tech and Internet of Things were dominant, but this year Nvidia opened with a revolution on a global scale.
“We’re going to talk about cars.” said Nvidia CEO Jen-Hsun Huang on the stage at the Four Seasons Hotel on Monday the 4th evening.
Huang didn’t waste time in unveiling the details of the new NVIDIA Drive PX 2:
a super-computer with 12 CPU cores and four Nvidia Pascal GPU chips for self-driving cars.
The hardware offers the power of 150 MacBook Pro laptops but the entire super-computer fits in the trunk in the size of a standard school lunchbox.
This is a revolution for the society and the potential of self-driving cars to new mobility services.
“Humans are the least reliable of the car, we represent almost all of the fatalities around the car worldwide.” said Huang.
Nvidia is tying its self-driving car tech back to another initiative at the top of the company’s agenda: deep learning.
In November they unveiled a new end-to-end hyperscale data center platform built especially for deep learning as well as Jetson TX1, the world’s leading visual computing platform for GPU-accelerated parallel processing, a credit card-sized module promising to power millions of new smart devices to come.
Huang explained how these developments will connect an ecosystem rooted in an end-to-end platform for deep learning that will link the in-car AI super-computer to deep neural networks to the car itself.
“New world will be a cloud-connected Internet of Cars ” in which cars exit the line with “superhuman” features.
Over the last several months Nvidia has been testing its own end-to-end system Nvidia DriveNet, for multiple and single object detection on the road.
“This network, because we want to put it into a car, has to recognize things in real-time” Huang noted.
Actually, Nvidia is working to train its system to recognize both objects and circumstances, from undistinguished road lanes in snowy territories or emergency vehicles.
With a powerful enough computer and deep learning algorithms working on one software stack, it’s possible to imagine navigating yourself down the road in a self-driving car, safely.
