What is Jetson

The availability of embedded technology capable of doing machine learning tasks on the edge has exploded in the last year or two, perhaps signaling the beginnings of a paradigm shift in how we think about machine learning and how the Internet of Things may be developed.

However, when we talk about edge hardware, we’re increasingly referring to two distinct sorts of hardware. Machine learning inferences are being performed on incredibly little hardware, all the way down to new battery-free edge computing gear that works on solar power.

Nevertheless, we’ve seen bespoke silicon that accelerates machine learning inferencing at the edge. These boards, which are built on bespoke ASIC and consume more power than the small boards, are blazingly fast in contrast.

NVIDIA’s offering in this arena, which is based on their GPU-based hardware, has traditionally been more powerful and more costly.

  • Nvidia Jetson is a line of embedded computer boards produced by the company. It is a low-power system that helps machine learning applications run faster.

Jetson types

Each NVIDIA Jetson is a full System on Module (SOM), with RAM, CPU, high-speed connections, power management, and other features. Jetson modules are available in a variety of performance, power efficiency, and form factor combinations, allowing them to be used by clients in a variety of sectors. Jetson ecosystem partners help you go to market quicker with AI embedded and edge devices by providing software, hardware design services, and off-the-shelf compatible solutions ranging from carrier boards to entire systems.

  • Xavier Series – The Jetson AGX Xavier is the world’s first computer developed particularly for self-driving cars. Customers can deliver their newest AI applications to the edge with this tiny, power-efficient module, which features hardware acceleration for the whole AI pipeline and plenty of high-speed I/O. Jetson AGX Xavier Industrial extends the temperature range, shock, and vibration parameters, as well as new functional safety features, for clients who wish to manufacture industrial-grade and/or safety-certified devices.
  • Xavier NX – In a tiny form factor module, Jetson Xavier NX offers up to 21 TOPs of accelerated AI computation to the edge. It can handle input from several high-resolution sensors and operate multiple current neural networks simultaneously, which is a necessity for complete AI systems. Jetson Xavier NX is ready for production and works with all major AI frameworks.
  • AGX Orin – With the world’s most important AI computer for energy-efficient autonomous machines, you can bring your next-generation goods to existence. This is the best option for applications ranging from supply chain management to commerce and medical.
  • Orin NX – In the tiniest Jetson size factor, experience the world’s most powerful AI computer for autonomous power-efficient devices. It has 5X the performance of the NVIDIA Jetson XavierTM NX and double the CUDA cores, as well as high-speed interface capability for many sensors. Jetson Orin NX delivers enormous performance in a small size, with 100 TOPS for many concurrent AI inference pipelines.
  • TX2 – In as low as 7.5 W, the Jetson TX2 series of embedded modules delivers up to 2.5X the performance of Jetson Nano. The Jetson TX2 NX has pin and form-factor compatibility with the Jetson Nano, although the Jetson TX2 all have the same form-factor as the original Jetson TX2. The Jetson TX2i is suitable for a variety of applications, including industrial robots and medical devices.
  • Nano –  The Jetson Nano module is a compact AI computer with the power and performance to run contemporary AI workloads, multiple neural networks in parallel, and analyze input from numerous high-resolution sensors at the same time. As a result, it’s an excellent starting point for adding sophisticated AI to embedded devices. The Jetson Nano is unique in that it is not a scaled-down version of NVIDIA’s other boards, and it supports a variety of machine learning frameworks such as Keras, MxNet, Caffe, TensorFlow, and PyTorch.

The board’s power consumption is anticipated to be between 5 and 10W, and it is powered by a single micro USB connector adjacent to the Ethernet socket. However, when reviewing the kit’s specifications, one aspect of the new board that immediately stands out is the possibly problematically large power supply specification needs.