LOS ANGELES — The nerdy war over who has the better self-driving car chip got a little personal at the auto show here last week, when one company started naming names and calling out the competition.
During a keynote speech, Intel CEO Brian Krzanich stood in front of a slide comparing the Mobileye EyeQ5 chip and Nvidia's Xavier automated driving platform.
"On the chart behind me is a comparison of deep-learning performance efficiency," Krza-nich said. Intel's chip for Mobileye is two and a half times more efficient than the competition, he said.
As cars become more connected and autonomous, computing demands have changed drastically, requiring more processing power with less of a drain on vehicle energy. This dynamic has given rise to an increasingly competitive race among chipmakers to develop the technology that can capably power autonomous vehicles, and Intel is investing heavily to be at the forefront.
Since its $15 billion acquisition of camera-sensor and computer-vision supplier Mobileye in March, Krzanich said, Intel has been working to marry the "eyes" of the Israeli company with the "brains" of the Silicon Valley giant. The result, according to the executive, is a perceptive, powerful and efficient component for autonomous vehicles.
To enable autonomous driving, vehicle electronic control units must be able to power deep learning, or the complex algorithms that allow a car to recognize and react to driving situations without human input.
"These are powerful computing modules," said Walter Sullivan, head of the Silicon Valley lab for Elektrobit, an autonomous driving software supplier, noting the size and heat output of such processors. "It's a physics problem."
Doug Davis, senior vice president of Intel's automated driving group, told Automotive News that Intel achieved efficiency by developing processors for specific functions.
"The way in which you build that hardware and software to do that processing is really where you get that optimization," Davis said. "That focus is on deep learning tera operations per second, really looking at how you manage that type of data and do that computing in a way you can optimize it the best."
Companies such as Intel, Nvidia and NXP have been racing to develop smaller and more efficient processors and cement a place in the vehicles of the future.
In October, Nvidia introduced its Pegasus automated driving platform, capable of powering driverless vehicles with 10 times the processing power of its predecessor, Xavier.
When introducing the platform, Danny Shapiro, Nvidia's head of automotive, said the hardware was no larger than a license plate and would save "thousands of watts" of energy.