NVIDIA launches a new generation of Jetson AGX modular computers for creating robotics
NVIDIA announced the launch of a new generation of Jetson AGX modular computers for creating robotics. This is reported by Engadget.
The new module, like the production versions of the Thor T5000, is designed for physical artificial intelligence and humanoid robotics. NVIDIA CEO Jensen Huang noted that Jetson Thor is designed for the millions of engineers involved in the development of robotic systems capable of interacting with the physical environment.
Jetson Thor is based on the Blackwell generation graphics processor and is equipped with 128 GB of memory. It provides performance of up to 2,070 teraflops in the FP4 format, which allows you to efficiently run the latest artificial intelligence models. At the same time, the power consumption of the system is about 130 W.
Productivity has increased significantly: the computing power for AI is 7.5 times higher than the previous Jetson Orin model, and the energy efficiency has become 3.5 times higher. Such chips are capable of processing large language and visual models of artificial intelligence, which helps robots better navigate in the surrounding world.
Thor systems run on a full-fledged Jetson software platform, specially developed for the needs of robotics. NVIDIA customers already using Jetson chips include Amazon, Meta, Agility Robotics and Boston Dynamics.
The Jetson AGX Thor development kit is available for $3,499. Thor T5000 production modules for serial robots are priced at $2,999 per unit with a minimum order of 1,000.
