NVIDIA has officially launched the new Jetson Thor chip, specifically designed for robots. The release aims to significantly enhance AI computing power in the robotics field, better supporting various forms of intelligent robots, especially humanoid robots.
Compared to the previous generation Jetson Orin, the computing power of Jetson Thor has increased by an impressive 7.5 times, reaching up to 2070 FP4 TFLOPS, with a power consumption of 130W and energy efficiency improved by 3.5 times. At the same time, Thor's memory has been significantly upgraded to 128GB, with a memory bandwidth of 273GB/s.
The Jetson Thor chip features the latest Blackwell architecture GPU, specifically designed for generative AI model inference, and is capable of supporting the next generation of "physical AI" agents. These agents can run in real-time at the edge, minimizing reliance on the cloud. To meet the demand for low latency and high performance in real-time applications, Jetson Thor comes with a range of powerful software tools that support all major generative AI frameworks and inference models, allowing developers to quickly conduct model experiments and inference locally.
NVIDIA also stated that with optimizations in FP4 precision and speculative decoding, the performance of Jetson Thor is expected to improve further. In addition, the chip supports the complete NVIDIA AI software stack, which can accelerate almost all physical AI workflows, covering areas such as robotics, video analysis, and sensor processing. The newly released Jetson Thor product includes a developer kit and mass-produced modules. The starting price of the development kit is $3,499 (approximately 25,000 RMB), while the module price is $2,999 (approximately 21,400 RMB).
At the recently concluded World Robotics Conference, several well-known domestic robot companies, such as Yuque Technology and Galaxy General, have announced that they will launch robots equipped with the Jetson Thor chip. Galaxy General's Galbot robot demonstrated its application in multiple industrial scenarios at the conference. This new chip not only attracts attention from the industry but also encourages related hardware and sensor companies to actively participate in the system building of Jetson Thor.
Additionally, NVIDIA's dedicated chip for autonomous vehicles, Nvidia Drive AGX Thor, will be available soon and is now open for pre-orders. With the development of artificial intelligence technology, NVIDIA is not only committed to providing strong computing power foundations but is also continuously advancing new research results, demonstrating its core position in the AI field. In the future, NVIDIA will continue to explore huge market opportunities in robotics, autonomous driving, and other fields, helping the industry develop.
Key Points:
🚀 AI Computing Power Improvement: The AI computing power of the Jetson Thor chip is 7.5 times higher than the previous generation, meeting the needs of various intelligent robots.
🤖 Memory and Performance: The Thor chip has a memory capacity of 128GB, with a memory bandwidth of 273GB/s, and has extremely high energy efficiency.
🌍 Industry Collaboration: Several domestic robot companies have announced the adoption of the new chip, promoting the integration and application of AI and robot technology.
