During its presentation at this week's Consumer Electronics Show, NVIDIA spent nearly 90 minutes talking about cars. The company quickly introduced its new mobile processor, the Tegra X1, and it showed a demo of the Unreal game engine running on it, but that was the last we'd hear about anything other than the automobile market.
Continue Reading Below
NVIDIA has been focused on the automobile market for quite some time, and it introduced two new platforms, both powered by the X1, aimed at making cars smarter. Drive CX, what NVIDIA calls a digital cockpit computer, has enough horsepower to run the equivalent of two 4K screens, allowing for not only infotainment systems but also a myriad of other in-car displays, with realistic graphics meant to mimic real materials.
The more exciting of the two platforms, though, is Drive PX. Drive PX is powered by two Tegra X1 chips, and its job is to identify objects around the vehicle based on inputs from up to 12 cameras. Having detailed information about its environment will be critical for the self-driving car of future, and NVIDIA wants its chips to be the brains behind the operation. If NVIDIA can get out ahead of the competition and push its product as the standard technology, it could have a billion-dollar opportunity in the automotive market.
How Drive PX works
There are two main ways for a computer to detect objects from visual data. The first is for engineers to code specific feature detectors into the software. There would be one for stop signs, one for speed limit signs, and so on. Not only does this approach not scale very well given the enormous number of potential objects the system would need to detect, but variations, such as stop signs partially hidden by trees, or pedestrians holding objects, could prove problematic.
Continue Reading Below
The second way is through deep learning, essentially training the system to detect features by feeding it an enormous amount of data. A supercomputer can be fed images tagged with what the images represent, and after churning through the data it generates software that can then classify objects on the fly, in this case running on the Drive PX platform. When Drive PX can't identify an object, it has the capability to send the data to a data center so the system can be retrained, deploying a software update to all the cars in the network. In this way, the Drive PX system will become more accurate over time.
To see an example of deep learning in action, the University of California, Berkeley, has a demo of its Caffe deep learning system, which runs on NVIDIA GPUs, at this website. Feed in any image, and the system will usually give a fairly accurate classification.
An enormous opportunity
NVIDIA has been working with Audi for years, and at CES the car company announced that it would use the Tegra X1 in its efforts to bring a self-driving car to the market. Previous iterations of the Tegra processor power the infotainment systems in roughly 6 million cars on the road today, including the Tesla Model S and the BMW i8.
The big opportunity for NVIDIA is not only in increasing the number of cars containing its processors, but also boosting the number of processors per car. Fellow Fool Ashraf Eassa estimated last year that NVIDIA gets about $50 in revenue per car using its processor, but so far Tegra has only been used to drive displays. A future car containing a Tegra chip to drive the displays, along with additional chips for driver-assistance features, could raise this revenue per car substantially.
About 83 million vehicles were sold around the world in 2013, and the trend in the automotive industry is clearly toward smart, connected cars. Plenty of other companies are vying for a piece of the automotive pie, including Intel and Qualcomm. While the graphics produced by NVIDIA's Drive CX platform are impressive, the battle for the center console will be intense.
However, NVIDIA is right to focus on the driver-assistance and autonomous driving features that could be made possible by the Drive PX platform. This is, I think, the company's biggest opportunity in the automotive market, and it's an area where NVIDIA has a distinct advantage. Dominance in this market isn't going to be about simply supplying chips, but about providing entire solutions, software and all.
NVIDIA has done exactly that in other markets. For example, NVIDIA sells its line of Tesla GPUs aimed at supercomputing, high-performance computing, and other operations. But it also provides myriad tools and code libraries, and it ensures a vast catalog of applications, from AutoCAD to computational finance simulations, are ported to its proprietary GPU compute language. The ecosystem around the hardware is what ultimately drives sales.
The main challenge in bringing a self-driving car to market is software, not hardware. You can stuff a tremendous amount of computing power into a car, but it's the software that makes the car smart, and it's the software that will determine which company dominates the market. I suspect we're still many years away from a mass-market self-driving car, but when that day comes, NVIDIA is pushing hard for Tegra to be behind the wheel.
The article NVIDIA Corporation Bets Big On Self-Driving Cars originally appeared on Fool.com.
Timothy Green owns shares of Nvidia. The Motley Fool recommends BMW, Intel, Nvidia, and Tesla Motors. The Motley Fool owns shares of Intel, Qualcomm, and Tesla Motors. Try any of our Foolish newsletter services free for 30 days. We Fools may not all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.
Copyright 1995 - 2015 The Motley Fool, LLC. All rights reserved. The Motley Fool has a disclosure policy.