Artificial intelligence and automotive technology have become key focuses of Nvidia executives, who believe the parallel processing power of the company's GPUs offers the best platform for the fast-growing emerging markets.
Those topics were at the center of Nvidia's message at CES 2016 in Las Vegas, where co-founder and CEO Jen-Hsun Huang introduced the company's Drive PX 2, the latest generation of computer designed for self-driving cars. During his talk at CES Jan. 4, Huang stressed that in order for autonomous cars to operate properly—find their way along complex and congested roads full of obstacles and dangers, collect and analyze in real time massive amounts of data from the various control systems and sensors, and learn as they go—they need to possess huge amounts of processing power.
Nvidia is looking to leverage the parallel processing power of its GPUs to bring those capabilities to an automotive industry that is pushing hard to create the self-driving automobiles of the future.
"Modern artificial intelligence and GPU breakthroughs enable us to finally tackle the daunting challenges of self-driving cars," the CEO said in a statement. "Nvidia's GPU is central to advances in deep learning and supercomputing. We are leveraging these to create the brain of future autonomous vehicles that will be continuously alert, and eventually achieve superhuman levels of situational awareness."
Nvidia over the last year has introduced the Drive CX platform, which enables such capabilities in cars as 3D navigation and infotainment, and the Drive PX board, which began shipping last summer and was powered by two Tegra X1 GPUs and offered up to 2.3 teraflops of compute power.
The company has upped the capabilities with the Drive PX 2 development board, which will be offered to early access partners in the second quarter and be generally available in the fourth quarter. It will be armed with two next-generation Tegra processors and two next-generation discrete GPUs that will be based on the Pascal architecture. It will deliver up to 24 trillion deep learning operations per second, more than 10 times the performance of the previous generation, according to Nvidia officials.
It will be capable of up to 8 trillion floating point operations per second (teraflops)—more than four times that of the previous generation board—which they said will enable such capabilities as sensor fusion, localization and path planning. It has the processing power of 150 Apple MacBook Pros, according to Nvidia.
The Drive PX 2 artificial intelligence (AI) platform can process in real time the various inputs of 12 video cameras as well as lidar, radar and ultrasonic sensors to give drivers a complete view of the environment they're driving through, according to Danny Shapiro, senior director of automotive at Nvidia. Drive PX 2 "basically enables automakers to train a deep neural network in the cloud and then bring that rapidly accelerated process of learning into the vehicle," Shapiro said in a video presentation. "So now as we drive, we can sense everything that's going on around the car."
The car can learn to deal with the various challenges on the road—from debris and pedestrians to construction zones and erratic drivers—and find a safe path forward, he said.
Drive PX 2 will work with other Nvidia technologies to offer a broader solution for car makers, according to officials. It includes the company's DriveWorks suite of software—including tools, libraries and modules—to help with the development and testing of self-driving cars, and Digits, which is used for developing and training deep neural networks needed for machine learning and artificial intelligence.
Nvidia's Drivenet technology enables autonomous vehicles to identify objects on the road, such as pedestrians and other vehicles.
Company officials said such top-tier automakers as Audi, BMW, Ford, Mercedes-Benz and Volvo are using the Drive PX platform, as are automotive component makers like Preferred Networks, ZMP and AdasWorks.