The GB200s communicate with each other via NVLink at 2 x 900 GByte/s. This makes the GB200 NVL2 suitable for training large models for generative AI applications. Finally, Jensen Huang gave an ...