The NVIDIA GB200 Grace Blackwell Superchip represents the pinnacle of AI computing, integrating one Grace CPU with two B200 Tensor Core GPUs through a high-bandwidth NVLink 5.0 interconnect. With 384GB of HBM3e memory and unprecedented memory bandwidth, the GB200 delivers up to 30x performance over previous generations for large language model training and real-time inference. Purpose-built for trillion-parameter AI models, it combines NVIDIA's most advanced CPU and GPU architectures in a single module for supercomputing-scale deployments.
Complete technical specifications for the GB200.