First wave of Nvidia Jetson Orin boards and systems
Over 30 embedded system makers are using the latest Jetson AGX Orin modules from Nvidia for edge AI and robotics.
The Jetson AGX Orin chip provides up to 275 TOPS, 8x the processing power of the previous Xavier device, in the same pin-compatible form factor. It is aimed at edge AI, AIoT, robotics and embedded applications with a series of modules available later this year.
The Orin chip uses the Nvidia Ampere architecture GPU with 2048 cores, up to 12 ARM Cortex-A78AE CPU cores, deep learning and vision accelerators, high-speed interfaces, faster memory bandwidth, and multimodal sensor support capable of feeding multiple, concurrent AI application pipelines.
- Nvidia boosts safety critical design with Jetson Orin AGX module
- Nvidia robotics AI perception and simulation SDK
The Orin modules in the AGX format vary from a $899 module with eight A78AE cores at 200TOPS, out in July, to the top end at 275TOPS with 12 cores in October for $1599.
The smaller Orin NX modules, the same format as the low cost Jetson TX2 modules, vary from 100TOPS with eight cores at $599, out in September, to a lower cost $399 version providing 70 TOPS from six cores, out in December.
Nvidia says it uses its first silicon for production, presumably with redundancy in the design to provide the different versions from the same die to avoid problems with yield. The chip is built on an 8nm process at Samsung.
Nvidia says it now has over 1m developers using Jetson at over 6000 companies.
Board and system makers at the Computex 2022 in Taipei using the modules include AAEON, Adlink, Advantech, Aetina, AIMobile, Appropho, AverMedia, Axiomtek, EverFocus, Neousys, Onyx and Vecow.
Designs include servers, edge appliances and industrial PCs in fan and fanless configurations with multiple connectivity and interface options, also including specifications for commercial or ruggedized applications. There will also be carrier boards for the modules for custom system development.
“The new Jetson AGX Orin is supercharging the next generation of robotics and edge AI applications,” said Deepu Talla, vice president of Embedded and Edge Computing at Nvidia. “This momentum continues to accelerate as our ecosystem partners release Jetson Orin-based production systems in various form factors tailored towards specific industries and use cases.”
Other partners launching Jetson Orin-based systemsinclude Auvidea, Basler AG, Connect Tech, D3 Engineering, Diamond Systems, e-Con Systems, Forecr, Framos, Infineon, Leetop, Leopard Imaging, MiiVii, Quectel, RidgeRun, Sequitur, Silex, SmartCow, Stereolabs, Syslogic, Realtimes, Telit and TZTEK.
Pretrained models from Nvidia’s NGC catalogue are optimized and ready for fine-tuning with the TAO toolkit and customer datasets to run on the modules. This reduces time and cost for production-quality AI deployments, while cloud-native technologies allow seamless updates throughout a product’s lifetime.
For specific use cases, Nvidia provides the Isaac Sim on Omniverse for robotics; Riva, a GPU-accelerated SDK for building speech AI applications; the DeepStream streaming analytics toolkit for AI-based multi-sensor processing, video, audio and image understanding; and 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.
Related articles
Other articles on eeNews Europe
- World’s largest commercial drone deployment uses German tech
- Micron to ship 232-layer 3D-NAND memory
- Apple looks to e-ink displays for foldable phones
- Flexible printed battery drives $2 Bluetooth IoT tag
