Nvidia unveiled a new compact artificial intelligence (AI) supercomputer on Tuesday. Dubbed Jetson Orin Nano Super, it is the successor of the Jetson Orin Nano which was launched in March 2023. The tech giant claimed that the new AI hardware offers increased performance, a software upgrade, and lower pricing compared to the predecessor. The AI computer is available as a developer kit, and the company said it can be useful for developers, hobbyists, and students who want to learn and build AI-based tools.
Nvidia Jetson Orin Nano Super Launched
In a blog post, Nvidia detailed its latest AI computer. It was also showcased in a video by CEO Jensen Huang. The Jetson Orin Nano Super is a palm-sized robotics processor that is optimised to run AI inferences, build large language model (LLM) tools, and even for robotics workflows. The AI computer is priced at $249 (roughly Rs. 21,150). Notably, the latest processor costs less than its predecessor which was priced at $499 (roughly Rs. 42,380).
The new AI processor is said to deliver a 1.7x increase in generative AI inference performance. Nvidia claimed it also offers a 70 percent increase in performance to 67 8-bit Integer Tera Operations Per Second (INT8 TOPS) compared to last year’s processor. Additionally, the memory bandwidth is also said to be increased to 102GBps, a 50 percent gain compared to the predecessor.
Apart from hardware improvements, the tech giant is also adding software updates to the AI supercomputer, which is said to further optimise its performance. Nvidia said that the AI computer is efficient at creating retrieval-augmented generation-based chatbots that leverage LLMs, building visual AI agents, or deploying AI-powered robots.
The main advantage of the Nvidia Jetson Orin Nano Super over other AI PCs is that it can handle foundation models and allows developers to build tools and agents that are diverse and can be used for general purposes.
With the developer kit, the company is offering a Jetson Orin Nano 8GB system-on-module (SoM) and a reference carrier board. The SoM features a Nvidia Ampere architecture GPU with tensor cores and a six-core ARM CPU. The processor can also support up to four cameras for working with AI vision models.