NVIDIA's 'mobile supercomputer' can make machines smarter

Hardware developers looking for a quick way to add machine learning to their devices just got some help from Nvidia. The GPU-centric company announced the Jetson TX1 developer module with an emphasis on machine learning. The device could lower the lead time needed to create machines that rely the on computations needed for dynamic input and computations. This includes: autonomous drones, facial recognition and behavioral analysis within a crowd and computer vision. The TX1 development kit replaces the Nvidia TK1 launched last year. The tiny supercomputer pushes one terraflop of performance, networks via ethernet and 802.11 ac, is under 10 watts and ships with Linux installed. Oh, and you can play Doom on it. The TX1 developer kit will be available for preorder on November 12 and will ship on November 16. The retail price is $599 with educators and students able to grab one for $299. The stand-alone module will be $299 and will ship in the first quarter of 2016.

At an event in San Francisco, Nvidia CEO Jen-Hsun Huang said, "the architecture of the computer is fundamentally changing and it's changing right in front of us." He pointed out how GPUs are enabling the computing power needed for machine learning features like those found being utilized by Facebook, Google and YouTube for image recognition and stabilization. By offering the TX1 to the masses, companies without the resources of the Silicon Valley giants can build their own small-scale devices without additional overhead. "We're trying to democratize machine learning," added Huang.