Obsessed with technology?
Subscribe to the latest tech news as well as exciting promotions from us and our partners!
By subscribing, you indicate that you have read & understood the SPH's Privacy Policy and PDPA Statement.
News Categories

The new NVIDIA Jetson TX2 module may be just what edge devices need for AI computing

By Wong Chung Wee - on 8 Mar 2017, 10:30am

The new NVIDIA Jetson TX2 module may be just what edge devices need for AI computing

The NVIDIA Jetson TX2. (Image source: NVIDIA)

NVIDIA has just announced its new Jetson TX2 embedded AI supercomputer. Besides the updated hardware, the company has also tweaked its accompanying SDK to address the need for AI capabilities for edge devices. Edge devices that require artificial intelligence, or more specifically, deep learning computing requirements, process large amounts of data with as little latency as possible. This is where embedded systems like the NVIDIA Jetson TX2, or its predecessor, the Jetson TX1, come into play. For example, traffic cameras that need to monitor busy motorways will benefit from the deployment of AI at the edge. The embedded systems will ensure data is processed with little bandwidth bottlenecks; in the same breadth, they reduce latency and the need for cloud connectivity as data is processed and stored on-premises.

(Image source: NVIDIA)

The NVIDIA Jetson TX2 has improved hardware specifications. For start, its GPU has been upgraded to Pascal, and its onboard processor consists of 64-bit Denver 2 and A57 chips. Though it wasn’t specifically mentioned in its specifications, the Jetson TX2 is powered by what appears to be the Tegra Parker SoC, the same system on chip that powers NVIDIA autonomous vehicle systems. Besides such hardware improvements, NVIDIA claims Jetson TX2 offers twice the improvement for performance and power efficiency. The Jetson TX2 supports two operating modes; Max-Q and Max-P. In Max-Q mode, the Jetson TX2 operates with maximum power efficiency, with a rated power draw of less than 75W. In Max-P mode, Jetson TX2 operates at twice the performance level of the TX1, with a power draw of less than 15W.

The Jetpack SDK version 3.0. (Image source: NVIDIA)

At the same time, the module’s accompanying Jetpack SDK has been updated to version 3.0, with Linux Kernel 4.4. One important addition to Jetpack 3.0 is the libargus camera API.

(Image source: NVIDIA)

With a video capture system built around the new API, the Jetson TX2 is capable of processing two HEVC 4K video streams and perform object tracking. The system is able to support video and graphics compositing, with additional metadata, in a 4K output video stream.

The Jetson TX2 development kit. (Image source: NVIDIA)

The production module of NVIDIA Jetson TX2 will be available from authorized distributors worldwide from the second quarter of this year. The module will cost US$399, in 1K units. The Jetson TX2 development kit is available for pre-order now in the United States and Europe. The kit has a retail price of US$599, and for educational institutions, the kit has a SRP of US$299. For interested parties residing in the Asia-Pacific and other regions, we would have to wait till next month for their availability.

(Source: NVIDIA)