NVIDIA continues to push the boundaries of AI accelerators both in the data center and at the edge. Last month, the company announced the availability of the latest device in his Jetson family, the Jetson Orin Nano developer kit.
First announced in September 2022, the Jetson Orin Nano System-on-Module (SoM) offers 80x the performance of previous generation Jetson Nano devices. The Developer Kit puts the power of her SoM into the developer’s hands by giving her access to the SoM.
Below is a benchmark snapshot comparing the performance of AI vision models on Jetson Nano and Jetson Orin Nano.
Jetson Orin features a graphics processing unit (GPU) based on the NVIDIA Ampere architecture, an ARM CPU, next-generation accelerators for deep learning and vision, high-speed interfaces, high memory bandwidth, and support for multiple types of sensors. . Performance and flexibility enable customers to build products with built-in advanced intelligence never before thought possible. With this device, developers can inject AI into applications running at the edge, and makers and builders can leverage Robotics Operating Systems (ROS) to build intelligent his machines.
The Jetson Orin Nano 8 GB module features an NVIDIA Ampere architecture GPU with 1024 CUDA Cores, 32 3rd Gen Tensor Cores, and a 6-core ARM CPU. With this device, you can run multiple AI application pipelines simultaneously for high-performance inference. The developer kit’s carrier board has a number of connectors, including two of his MIPI CSI connectors that support up to 4-lane camera modules and allow for higher resolutions and frame rates than before.
The developer kit includes a bootable microSD card and two M.2 Key-M NVMe sockets for additional storage needs. It comes with a Gigabit Ethernet connector and a 40-pin expansion header for connecting sensors and actuators.
With generative AI gaining traction, NVIDIA has enabled transform models to run on our latest SoMs and developer kits. Transformer models such as GPT, BERT, and DALL-E learn context by analyzing relationships between different elements of sequential data, eliminating the need for large labeled datasets.
NVIDIA’s USP is a powerful combination of hardware and software for AI accelerators. The Jetson Orin Nano developer kit runs the latest version of his JetPack-v5.1.1. This includes everything from the operating system to the high-level APIs developers use to build edge AI applications. It features a CUDA-X accelerated computing stack targeting AI use cases for industrial IoT, manufacturing, and smart cities.
NVIDIA now has a portfolio of SoMs covering the full spectrum of edge computing.
Jetson Orin Nano 4GB and 8GB – Priced at $259 and $499, these SoMs are the most affordable and powerful edge computing devices, delivering 40 TOPS or 1 trillion AI performance per second. Note that these modules should be part of your developer kit for developing and debugging AI models and applications. The Jetson Orin Nano 8GB Developer Kit will be available soon for $499.
Jetson Orin NX 8GB and 16GB – These modules deliver AI performance up to 100 TOPS in the smallest Jetson form factor, with power configurable between 10W and 25W. The 16GB version of SoM is $699 and the 8GB version is $479.
Jetson AGX Orin 32GB and 64GB – Offering AI performance up to 275 TOPS, this Jetson Orin variant is the most powerful version. The 32GB SoM is priced at $999.00, while his higher-end 64GB version retails for $1,799.00. A fully assembled and ready to use Jetson AGX Orin 64GB Developer Kit retails for $1,999.00.
NVIDIA has partnered with various OEMs and hardware manufacturers to develop Jetson Orin Nano reference boards and developer kits. Companies like Advantech, Aetina, Auvidea, and Seeed Studio have devices with different form factors and capabilities.
When it comes to edge computing, NVIDIA continues to invest in hardware and software to enable developers to build the next generation of AI-enabled applications. The Jetson Nano Orin Developer Kit brings the power of AI to developers, builders and makers.
Follow me please twitter or LinkedIn. check out my website.