Nvidia is betting that its new "mobile super chip" will help achieve truly self-driving cars, reflecting a new direction for the company from its traditional market in mobile devices.
The company said the new Tegra X1 will aid in the development of intelligent cars that recognise objects, signs, images, lanes and other things. Cars could also self park, and help improve driver safety.
The Tegra X1 will provide the computational horsepower required to process images based on which vehicles can see and analyze on-road situations. The chip is more of a supercomputer-in-a-car, which could understand the environment around a vehicle, be situationally aware, and take action.
Based on video and image input, the chip has been trained to recognise pedestrians, traffic lights, and cross-walk and speed limit signs to which cars can respond. Images and video go to the Tegra X1 chip, and based on contextual awareness, help the car make a decision on what to do next. For example, a car can detect, recognise, classify and apply traffic lights in relevant situations.
The Tegra X1 was unveiled by Nvidia CEO Jen-Hsun Huang at a press conference on Sunday ahead of the International CES trade show.
Tegra chips have traditionally gone into mobile devices, but Huang brushed away the use of the X1 in mobile devices. That could indicate a new focus for Tegra chips in automobiles and other markets.
Car features such as blind spot detection and parking assistance have evolved over the years with radar and ultrasound technology. But cars are getting higher resolution cameras, which can help gather images, and Tegra X1 will play a key role in processing and understanding those images, Huang said.
The new chip has 256 graphics cores, and eight 64-bit CPU cores. It delivers two times the graphics performance while consuming the same amount of power as its predecessor, the Tegra K1. The Tegra K1 was introduced exactly a year ago, and is used in Google's Nexus 9 tablet, Acer's Chromebook 13 and other products.
The graphics capabilities in Tegra X1 comes from the company's latest Maxwell architecture, which is already in laptops and desktops. The Maxwell graphics architecture can deliver two times the performance and power efficiency of the Kepler graphics architecture, which is used in the Tegra K1.
Nvidia is using two Tegra X1 chips in its new Drive PX hardware platform for the development of intelligent cars. The platform is computationally intensive and can run at 2.3 teraflops and can process information from 12 cameras simultaneously at 1.3 billion pixels per second.
Infotainment systems in cars look like "cartoons" now, Huang said, adding that the new chip will also help deliver improved navigation and entertainment systems.
Nvidia's Tegra X1 will be used in the Drive CX car platform, which Huang called the "world's most advanced cockpit" for vehicles. It has two Tegra X1 chips, and will help make navigation more interactive. The Tegra X1 will also bring 4K video to cars. Car makers can use the Drive CX to develop in-car navigation or infotainment features.
For autonomous cars, the ability to be situationally aware and recognise objects is most important, said Patrick Moorhead, principal analyst at Moor Insights and Strategy.
The Tegra X1 scores in that capability, and Nvidia worked overtime in bringing Maxwell to Tegra chips, Moorhead said.
"It's really important to have that level of object recognition. That's where Nvidia is going to shine," Moorhead said.
One possible partner for X1 could be Tesla, which is always looking for the latest and greatest in cars.
"Tesla wholeheartedly is working on productising an autonomous car," Moorhead said.
The thermal characteristics of the Tegra X1 are the same as K1, so there's nothing to stop the chip from going into mobile devices. But growth in tablets has tapered off, and automotive and other markets are more attractive growth markets for Nvidia, Moorhead said.
Beyond automobiles, Nvidia could use the chips in robots that are highly dependent on object recognition, though the company hasn't talked about it, Moorhead said.
"I could see it playing into that market," Moorhead said.