That processing work employs machine learning and neural network techniques also being explored by Google and Facebook. They'll allow car computers to identify traffic lights, pedestrians and other upcoming objects, and can improve their image recognition with practice over time.
"There will be more computing power inside your car than in any device you have today," said Huang, who was joined by a representative from Audi, which will use the Tegra X1 in its vehicles next year.
The demonstration seemed labored to some in the audience, but Nvidia is betting on a technology with a good future, McGregor said. "For better or for worse, we're turning our cars into computers," he said. The advances will almost certainly lead to greater road safety, but they could also require more maintenance and upkeep. "Electronic systems are just more fragile than mechanical systems," he said.
Patrick Moorhead, industry analyst at Moor Insights and Strategy, was also surprised to see Nvidia focus exclusively on auto tech, but he doesn't think the company is backing away from portable devices.
The Tegra X1 performs well enough at around 4 watts that it can be used in high-end tablets in the future, Moorhead says, and he expects Nvidia to use the X1 in its own Shield gaming system.
The decision to focus on auto tech at CES was a strategic one, he said, even if it gave some the impression that Nvidia had lost sight of the big picture.
Both he and McGregor think Nvidia will make mobile announcements at the Mobile World Congress show in Barcelona in March. "There's nothing that led me to believe they're not still looking at mobile as well," Moorhead said.
Still, the Tegra hasn't been as successful in smartphones and tablets as Nvidia would like, which also helps explain why it initiated its first ever lawsuits this year over patents, accusing Qualcomm and Samsung of infringing its GPU technology in their mobile processors.
Nvidia probably wouldn't have launched that action if it was making more headway in smartphones and tablets, Moorhead said. And it might not have spent 90 minutes Sunday night discussing car computing.
Sign up for CIO Asia eNewsletters.