To help developers bring new AI and robotics applications to market, or to support existing projects, Nvidia has released the Jetson AGX Orin 32Byte production module.
Companies in the Nvidia Partner Network are offering commercially available products powered by the new module, which provides up to a six-fold performance improvement compared with the previous generation.
Developers can build and deploy Orin-powered systems with cameras, sensors, software and connectivity for edge AI, robotics, AIoT and embedded applications, proposed Nvidia.
The company envisaged applications in manufacturing, retail and construction to agriculture, logistics, healthcare, smart cities and last-mile delivery vehicles. Production-ready systems have options for peripherals to encourage creativity.
According to Nvidia’s Jetson partner ecosystem manager, Thejas Boggaram Shivashankar, the Jetson AGX Orion allows developers and engineers to handle multiple concurrent data streams for complex application environments without strict latency requirements, energy-efficiency constraints or issues with high-bandwidth wireless connectivity.
The Jetson AGX Orin developer kit is capable of up to 275 trillion operations per second and supports multiple concurrent AI application pipelines with an Nvidia Ampere architecture GPU, together with deep learning and vision accelerators, high-speed I/O and fast memory bandwidth.
Using Jetson AGX Orin, he continued, customers can develop solutions using the largest and most complex AI models to solve problems such as natural language understanding, 3D perception and multi-sensor fusion.
The company announced four Jetson Orin-based production modules, with the Jetson AGX Orin 32Gbyte module available to purchase now, and the 64Gbyte version available in November, with two Orin NX production modules due later this year.
The production systems are supported by the Nvidia Jetson software stack to build and deploy fully accelerated AI.
The JetPack software development kit includes the Nvidia CUDA-X accelerated stack while Jetson Orin supports multiple Nvidia platforms and frameworks such as Isaac for robotics, DeepStream for computer vision, Riva for natural language understanding and TAO Toolkit to accelerate model development with pretrained models. There is also Metropolis, an application framework, set of developer tools and partner ecosystem that brings visual data and AI together to improve operational efficiency and safety across industries.
Carrier boards and full hardware systems options are available from Aaeon, Auvidea, Connect Tech, MiiVii, Plink-AI, Realtimes and TZTEK.
Over 350 camera and sensor options are available from Allied Vision, Appropho, Basler AG, e-Con Systems, Framos, Leopard Imaging, LIPS, Robosense, Shenzhen Sensing, Stereolabs, Thundersoft, Unicorecomm and Velodyne. These can support indoor/outdoor lighting conditions, as well as capabilities like lidars for mapping, localisation and navigation in robotics and autonomous machines.
Software support like device management, operating systems (Yocto and Realtime OS), AI software and toolkits, developers are available from Allxon, Cogniteam, Concurrent Realtime, Deci AI, DriveU, Novauto, RidgeRun, and Sequitur Labs.
Connectivity options, including WiFi 6/6E, LTE and 5G are offered by Telit, Quectel, Infineon and Silex.
