PANews, March 19, Nvidia announced today the launch of the NVIDIA Blackwell platform to empower a new era of computing. The platform enables institutions around the world to build and run real-time generative AI on trillion-parameter large language models (LLMs) at up to 25x lower cost and energy consumption than the previous generation. The Blackwell GPU architecture is powered by six transformative accelerated computing technologies that will enable breakthroughs in data processing, engineering simulation, electronic design automation, computer-aided drug design, quantum computing, and generative AI.
The NVIDIA GB200 Grace Blackwell superchip connects two NVIDIA B200 Tensor Core GPUs to NVIDIA Grace CPUs via an ultra-low-power 900GB/s chip-to-chip interconnect. Compared to NVIDIA H100 Tensor Core GPUs, the GB200 superchip delivers up to 30x faster performance for LLM inference workloads and up to 25x lower cost and power consumption.
In addition, NVIDIA also launched a general base model for the field of robotics called GR00T during the GTC conference in San Jose, USA. In addition, Nvidia has introduced a new computer for robotics, Thor, which has been specifically optimized for performance, power consumption and size. According to reports, Nvidia hopes that through this initiative, robots will have smarter brains, so that they can make natural imitation actions by observing human behavior, which will greatly improve the coordination and flexibility of actions.
