SuperPOD · DGX SuperPOD
NVIDIA's reference cluster design wiring many DGX systems into a validated, scalable AI supercomputer.
Current numbers
0.5–5 GB/s/GPUDGX H200 SuperPOD shared-storage read guidelines: 0.5, 1, and 5 GB/s per GPU for Good, Better, and Best performance, with 4 GB/s/GPU recommended for computer-vision workloads
9,216 chipsGoogle Ironwood (TPU v7) scale-up superpod via ICI (9.6 Tb/s/chip) + OCS (48 units, 13,824 ports)
576 GPUs / SUNVIDIA GB200 SuperPOD scalable unit (8 systems); a 16-SU, 9,216-GPU cluster needs a three-tier switch BOM on the order of 1,000 boxes