New World

Nvidia releases Jetson Thor, the most powerful “brain” for robots: performance increases 7.5 times, and Yushu Technology is the first to use it

August 26th news, NVIDIA announced today that the NVIDIA Jetson AGX Thor™ developer kit and production-grade module are now available. “Jetson Thor has unparalleled performance and energy efficiency, and is able to run multiple generative AI models simultaneously on edge devices. As an excellent supercomputer, Jetson Thor is working to push the realization of the era of physical AI and universal robots.”

It is understood that Lianying Medical, Wanji Technology, UBL, Galaxy General Motors, Yushu Technology, Zhongqing Robot and Zhiyuan Robot have taken the lead in using Jetson Thor.

Nvidia mentioned that based on the NVIDIA Blackwell GPU and equipped with 128GB of memory, Jetson Thor provides AI computing power up to 2,070 FP4 TFLOPS, allowing easy running of the latest AI models, all within the 130-watt power range.

Compared with the previous generation of NVIDIA Jetson AGX Orin™, Jetson Thor’s AI computing performance has been improved to 7.5 times and its energy efficiency has been improved to 3.5 times. It can run a variety of generative AI models—from visual language action models such as NVIDIA Isaac™ GR00T N1.5 to mainstream large language models and visual language models.

This new system module solves one of the most important challenges of robotics, namely running multiple AI workflows, allowing robots to interact in real-time, intelligently with the human and physical worlds. Jetson Thor is able to implement real-time reasoning, which is crucial for high-performance physical AI applications such as humanoid robots, agriculture, and surgical assistance.

Based on the NVIDIA Jetson™ software platform, Jetson Thor is designed for physical AI and humanoid robots, supporting all mainstream AI frameworks and generative AI models. At the same time, it is fully compatible with NVIDIA’s cloud-to-edge software stack, including NVIDIA Isaac for robot simulation and development, Isaac GR00T, NVIDIA Metropolis for visual AI, and NVIDIA Holoscan for real-time sensor processing.

Since its launch in 2014, the NVIDIA Jetson platform and robotics stack have attracted more than 2 million developers and have built an expanding ecosystem covering more than 150 hardware systems, software and sensor partners, of which Jetson Orin has helped more than 7,000 customers deploy edge AI in various industries. Jetson Thor will further promote visual AI agents and complex robotic systems

Leave a Reply

Your email address will not be published. Required fields are marked *