NVIDIA developer blog 'Upgrading Multi-GPU Interconnectivity with the Third-Generation NVIDIA NVSwitch' (GB200 NVL72)

First-party source for the Hopper-generation switch, which had NO record in this repo. The page title carries NVIDIA's own generation number: with 'fourth-generation NVLink and third-generation NVIDIA NVSwitch, a system with eight NVIDIA H100 Tensor Core GPUs features 3.6 TB/s of bisection bandwidth and 450 GB/s of bandwidth for reduction operations'. Also establishes the NVLink 1 origin point NVIDIA itself uses ('the first generation at 160 GB/s, introduced in 2014') and the fifth-generation per-GPU figure as bidirectional ('1.8 TB/s of bidirectional throughput per GPU'), plus 'Each NVLink switch tray delivers 144 NVLink ports at 100 GB'.