NVIDIA Blackwell datasheet, first-party copy (GB200 NVL72 / GB200 NVL4 / HGX B200)

VERBATIM VERIFICATION 2026-10-04, retrieved via curl -L -A 'Mozilla/5.0 (Windows NT 10.0; Win64; x64)' from the 'Read Datasheet' link on https://www.nvidia.com/en-us/data-center/gb200-nvl72.md . This is NVIDIA's OWN hosted copy, resolving the audit flag on the existing source nv-blackwell-datasheet (which cites a byte-identical file on reseller host nor-tech.com). Document footer '4204213. Oct 25'. Technical Specifications table, three per-GPU columns GB200 NVL72 | GB200 NVL4 | HGX B200, verbatim: 'Total NVFP4 Tensor Core2 | 1,440 | 720 PFLOPS | 80 | 40 PFLOPS | 144 | 72 PFLOPS' 'Total FP8/FP6 Tensor Core2 | 720 PFLOPS | 40 PFLOPS | 72 PFLOPS' 'Total Fast Memory | 31 TB | 1.8 TB | 1.4 TB' 'Total Memory Bandwidth | 576 TB/s | 32 TB/s | 62 TB/s' 'Total NVLink Bandwidth | 130 TB/s' and '7.2 TB/s | 14.4 TB/s' 'FP4 Tensor Core2 | 20 PFLOPS | 20 PFLOPS | 18 PFLOPS' 'FP8/FP6 Tensor Core2 | 10 PFLOPS | 10 PFLOPS | 9 PFLOPS' 'INT8 Tensor Core2 | 10 POPS | 10 POPS | 9 POPS' 'FP16/BF16 Tensor Core2 | 5 PFLOPS | 5 PFLOPS | 4.5 PFLOPS' 'TF32 Tensor Core2 | 2.5 PFLOPS | 2.5 PFLOPS | 2.2 PFLOPS' 'FP32 | 80 TFLOPS | 80 TFLOPS | 75 TFLOPS' 'FP64/FP64 Tensor Core | 40 TFLOPS | 40 TFLOPS | 37 TFLOPS' 'GPU Memory | Bandwidth | 186 GB HBM3E | 8 TB/s | 186 GB HBM3E | 8 TB/s | 180 GB HBM3E | 7.7 TB/s' 'Max Thermal Design Power (TDP) | Configurable up to 1,200 W | Configurable up to 1,200 W | Configurable up to 1,000 W' 'Interconnect | Fifth-generation NVLink: 1.8 TB/s PCIe Gen5: 128 GB/s' FOOTNOTES VERBATIM: '1. Specifications in sparse | dense. 2. Specifications in sparse. Dense is one-half of the sparse spec shown.' CONFLICT WITH THE RECORD'S PROSE: the gb200-nvl72 record's notes assert that 'an earlier revision of NVIDIA's Blackwell datasheet carries an explicit exclusion clause in its sparsity footnote, footnote 1, which reads in full: All Tensor Core numbers except FP64 with sparsity.' That wording does NOT appear in this first-party October-2025 edition, whose footnote 1 is the bare 'Specifications in sparse | dense.' with no FP64 exclusion. See vb-nvidia-gb300-nvl72-spec-table for a DIFFERENT footnote convention ('All Tensor Core specifications are with sparsity unless otherwise noted.'). The FP64 dense:true flag on nvidia-gb200-nvl72 is therefore NOT supported by the wording quoted in its own notes.