Nvidia Corp. has been awarded a contract worth up to $20 million from the Defense Advanced Research Projects Agency (DARPA) to research embedded processor technologies that could lead to dramatic improvements in the ability of autonomous vehicles to collect and process data from on-board sensors.

Existing embedded processors deliver about 1GFLOPS of performance (1 billion floating point operations each second) per watt. The Nvidia program, known as Project Osprey, will research low-power circuits and extremely efficient architectures and programming systems that enable 75GFLOPS per watt, using process technologies as advanced as 7nm compared with today's 28-nm process. Nvidia’s next-generation code-named Maxwell architecture is targeted to deliver 16GFLOPS per watt.

The five-year contract, awarded under DARPA's power efficiency revolution for embedded computing technologies (PERFECT) program, will fund research for processors that are 75-times more energy efficient than current embedded solutions. The goal is to enable surveillance and computer vision systems in ground and airborne vehicles to collect and analyze vastly more data than can be processed today in real time.

"The technologies developed with this program can transform the capabilities of embedded systems, making autonomous vehicles more practical and intelligent. This research will help Nvidia continue to advance mobile computing for both government and consumer applications," said Steve Keckler, senior director of architecture research at Nvidia.

Project Osprey will leverage Nvidia's strengths in heterogeneous computing and parallel processing technology, which enable more efficient processing than traditional CPUs. Nvidia processors are used in a wide variety of embedded applications today, including automobiles made by Audi, BMW, Tesla and Lamborghini, aircraft including the F-22 Raptor, and U.S. Army tanks.

Nvidia researchers will work on the program with academic partners at the University of Utah and the University of Virginia. Project Osprey, which is now underway, could, combined with two optional additional phases, continue over the next five and one-half years.

Tags: Nvidia, DARPA, Osprey, GPGPU, ARM, Maxwell, 7nm, Tesla, Tegra, Boulder, Denver, Quadro, Geforce