NVIDIA Spawns Jetson AGX Xavier Module As The Brain For Next-Gen AI Machines

NVIDIA launched its first Jetson family products back in 2015 when its Jetson TX1 system aimed to be the brains behind autonomous devices, from drones to robotics and autonomous vehicles. The Silicon Valley GPU giant then added to the Jetson line in 2017 with the Jetson TX2 for commercial drones and smart robots. Today, NVIDIA has now announced the latest addition to the Jetson family called the Jetson AGX Xavier Module, and the company says this new processing engine is targeted at bringing up to 32 TOPS of performance to next-generation autonomous machines.

Despite its big processing power, the AGX Xavier Module is small enough to fit in the palm of your hand (or robotic arm), and NVIDIA envisions these modules fitting inside machines that will have crept straight out of science fiction into reality. Xavier can power manufacturing robots that will collaborate with humans, delivery robots that bring items to your door, and devices like handheld DNA sequencers that help save crops from disease. Developers can use AGX Xavier modules as the brain of an autonomous device with the power of a workstation server in a processing plant no bigger than your fist.

The small size makes the platform appropriate for devices with extreme space constraints. Even with this claim of "workstation server performance," the module consumes as little as 10 watts of power, or about as much power as a clock radio. This can enable devices that operate on limited battery capacity to get as much runtime as possible per charge.

However, the Jetson AGX Xavier module is compromised of an 8-core ARM64-based custom processor with 8MB of L2 and 4MB of L3 cache, along with a 512-core NVIDIA Volta GPU with on board Tensor cores for direct hardware assist of machine learning workloads and machine vision. Also on board is 16GB of LPDDR4x memory and 32GB of eMMC storage. 

NVIDIA says Jetson AGX Xavier also is enabled by its AI platform and includes a complete set of tools and workflows to help the developers train and deploy neural networks with the device.

The platform supports applications developed with JetPack and DeepStream software. JetPack is an NVIDIA SDK for autonomous machines and supports AI, computer vision, multimedia, and more. DeepStream SDK for Jetson AGX Xavier allows streaming analytics and aims to bring AI to IoT and smart city applications. The SDKs are specifically designed to save time and money in development and to make it easy to add new features and functionality over time.

The Jetson AGX Xavier module is available now from distributors globally with pricing for 1000+ units at $1,099.