Nvidia has now unveiled the latest generation of the Jetson platform for edge-of-network devices, dubbed the TX2, which is capable of doing a handful of things, including powering commercial drones, factory robots, and smart cameras.
The credit card-sized platform comes with the "embedded AI supercomputer" Jetson TX2, which is able to power 4K drones that use up only about 7.5 watts of power. Video drones with the TX2 can even operate two cameras at the same time.
The Jetson TX2 can cough up twice the power of its predecessor, the TX1, and it's able to reroute efficiency to power savings, using significantly less power consumption than the TX1 to achieve the same processing results.
Nvidia Jetson TX2 Specs
The Jetson TX2 packs a Pascal-based GPU and it comes with two 64-bit Nvidia Denver 2 ARM A57 chips, with 8 GB of LPDDR4 RAM and 32 GB eMMC storage. It features built-in 802.11 ac WLAN, Bluetooth connectivity, and 1 GB Ethernet for wired internet connectivity. The spec sheet also features 4K x 2K 60 fps encode and decode, with 12 CSI lanes supporting up to six cameras at 2.5 GB per second per lane.
The Jetson 3.0 platform was developed for AI "at the edge" of the network, instead of in the cloud, or an internet-connected data center. Since camera-equipped drones can capture a lot of data, the platform is able to handle a lot of the processing data at the edge - within the device itself - instead of pushing that data to the cloud.
This makes the Jetson TX2 capable of pushing edge-of-network computing at greater lengths, paving the way for running distributed neural networks directly on edge devices that can more accurately perform AI actions, such as identifying objects in images, recognizing speech, or interpreting environments for unmanned navigation.
"We're seeing a lot of this inference not just in the cloud, but also moving toward the edge, whether it's a robot, or a drone, or a security camera," Deepu Talla, vice president and general manager of Nvidia's Tegra business unit, said. "And Jetson is our platform for doing inference and AI computing at the edge."
Apart from the Jetson TX2, Nvidia also announced JetPack 3.0, the latest version of the company's AI SDK for the Jetson line, which according to TechCrunch includes support for TensorRT 1.0, cuDNN 5.1, VisionWorks 1.6, alongside the latest graphics drivers and APIs.
Nvidia Jetson TX2 Partner Support
Some companies are already looking forward to the possible advantages Nvidia's Jetson TX2 can bring. Cisco says that it can use the platform to add local AI-powered features such as facial and voice recognition to its Spark enterprise, which would spell a handful of advantages with regard to security and authentication.
Nvidia Jetson TX2 Pricing And Availability
The new Jetson developer kit launches March 7, with a $599 preorder price point in the United States and Europe and a shipping date of March 14.
The Jetson TX2 module, on the other hand, will retail for $399 by Q2. Sales of the predecessors TX1 and TK1 Jetson-embedded computing platforms will continue, this time at lower price points.
Thoughts on Nvidia's Jetson TX2 module, or the new Jetson platform in general? Feel free to sound off in the comments section below!