Parts/Interconnect/NVIDIA InfiniBand NDR (Quantum-2)
Interconnect · NDR · 2022
NVIDIA InfiniBand NDR (Quantum-2)
400 Gb/s InfiniBand. The default back-end fabric for large commercial GPU clusters.
Why this part
The default fabric for large GPU clusters, and the reason Meta's two published clusters are so useful: they built one on NDR InfiniBand and one on RoCE Ethernet at the same scale, with the same GPUs, and reported that both trained the same model. That is the clearest public evidence anyone has offered that InfiniBand's position is contestable, which is why an operator large enough to run the experiment did it.
Systems using this part
- Systems
- 149
- Rmax underneath
- 5.03 EFlop/s
- Units recorded
- -
0 of 149 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Nexus (Georgia Tech) 2026-10 | Interconnect | USA | - | - | NVIDIA | - |
| HoreKa 2 2026-08 | Interconnect | Germany | - | - | NVIDIA | - |
| Sofia (VSC Tier-1) 2026-07 | Interconnect | Belgium | - | 8.19 PFlop/s | NVIDIA | - |
| Cassava AI Factory 2026-06 | Interconnect | South Africa | - | 77.79 PFlop/s | NVIDIA | #36 → |
| Sunrise 2026-06 | Interconnect | United Kingdom | - | 34.77 PFlop/s | NVIDIA | - |
| NIPA(NHN Cloud)-CL-1 2026-04 | Interconnect | South Korea | - | 137.4 PFlop/s | NVIDIA | #20 → |
| NIPA(NHN Cloud)-CL-2 2026-04 | Interconnect | South Korea | - | 68.42 PFlop/s | NVIDIA | #40 → |
| Jairu 2026-02 | Interconnect | Brazil | - | - | NVIDIA | - |
| NIPA (Kakao) 2026-01 | Interconnect | South Korea | - | 69.10 PFlop/s | NVIDIA | #39 → |
| Maximus-02 2026 | Interconnect | USA | - | 156.1 PFlop/s | NVIDIA | #18 → |
| Neo 2026 | Interconnect | USA | - | - | NVIDIA | - |
| Sirius (Tsukuba) 2026 | Interconnect | Japan | - | 5.63 PFlop/s | NVIDIA | - |
| Nanyang Titan 2026 | Interconnect | Singapore | - | 8.80 PFlop/s | NVIDIA | - |
| Tigris (RIT) 2026 | Interconnect | USA | - | 5.08 PFlop/s | NVIDIA | - |
| Viettel H200 AI Platform 2026 | Interconnect | Vietnam | - | 40.21 PFlop/s | NVIDIA | - |
| Airbus HPC6 2025-12 | Interconnect | France | - | - | NVIDIA | - |
| Shakti Cloud 2025-11 | Interconnect | India | - | 84.31 PFlop/s | NVIDIA | #32 → |
| Nano 4 2025-11 | Interconnect | Taiwan | - | 81.55 PFlop/s | NVIDIA | #33 → |
| Otus 2025-11 | Interconnect | Germany | - | 4.66 PFlop/s | NVIDIA | #5 → |
| PERUN 2025-11 | Interconnect | Slovakia | - | 10.21 PFlop/s | NVIDIA | #162 → |
| Harpia 2025-10 | Interconnect | Brazil | - | 75.20 PFlop/s | NVIDIA | #37 → |
| SEIRAN 2025-10 | Interconnect | Japan | - | 3.32 PFlop/s | NVIDIA | #374 → |
| OpenCanvas GPUaaS SP1 2025-10 | Interconnect | Japan | - | 17.53 PFlop/s | NVIDIA | #105 → |
| TELUS Sovereign AI Factory (Rimouski) 2025-09-24 | Interconnect | Canada | - | 22.74 PFlop/s | NVIDIA | #78 → |
| JUPITER 2025-09 | Interconnect | Germany | - | 1.00 EFlop/s | NVIDIA | #5 → |
| The F.I.N 2025-09 | Interconnect | Poland | - | 7.57 PFlop/s | NVIDIA | #211 → |
| Trillium 2025-08-07 | Interconnect | Canada | - | 9.49 PFlop/s | NVIDIA | #175 → |
| Cannon2-H200 2025-08 | Interconnect | USA | - | 4.24 PFlop/s | NVIDIA | #302 → |
| CHIE-4 2025-07 | Interconnect | Japan | - | 135.4 PFlop/s | NVIDIA | #21 → |
| Plasma Simulator Sosei 2025-07 | Interconnect | Japan | - | - | NVIDIA | - |
| ABCI 3.0 2025-06 | Interconnect | Japan | - | 145.1 PFlop/s | NVIDIA | #19 → |
| ABCI-Q 2025-06 | Interconnect | Japan | - | 74.58 PFlop/s | NVIDIA | #38 → |
| SSC-24 2025-06 | Interconnect | South Korea | - | 106.2 PFlop/s | NVIDIA | #25 → |
| Hopper NUS 2025-06 | Interconnect | Singapore | - | 15.05 PFlop/s | NVIDIA | #120 → |
| Betty 2025-05 | Interconnect | USA | - | 8.49 PFlop/s | NVIDIA | #189 → |
| CRESCO8 2025-04 | Interconnect | Italy | - | 5.43 PFlop/s | NVIDIA | #228 → |
| bwUniCluster 3.0 2025-04 | Interconnect | Germany | - | - | NVIDIA | - |
| Viper-GPU 2025-02 | Interconnect | Germany | - | 31.10 PFlop/s | NVIDIA | #55 → |
| Miyabi 2025-01 | Interconnect | Japan | - | 46.80 PFlop/s | NVIDIA | #28 → |
| Together AI / Hypertec Cloud GB200 Cluster 2025-01 | Interconnect | USA | - | - | NVIDIA | - |
| FPT AI Factory Japan 2025 | Interconnect | Japan | - | 49.85 PFlop/s | NVIDIA | #50 → |
| ISEG2 2025 | Interconnect | Netherlands | - | 202.4 PFlop/s | NVIDIA | #15 → |
| davinci-2 2025 | Interconnect | Italy | - | 14.21 PFlop/s | NVIDIA | #123 → |
| Ursa 2025 | Interconnect | USA | - | - | NVIDIA | - |
| Torch 2025 | Interconnect | USA | - | 10.79 PFlop/s | NVIDIA | #133 → |
| Space HPC 2025 | Interconnect | Italy | - | 4.47 PFlop/s | NVIDIA | #261 → |
| Tillicum 2025 | Interconnect | USA | - | 8.85 PFlop/s | NVIDIA | #185 → |
| SELENA 2025 | Interconnect | France | - | 5.42 PFlop/s | NVIDIA | #260 → |
| Terabit HPC Bubble 2025 | Interconnect | Italy | - | 2.91 PFlop/s | NVIDIA | #437 → |
| AI-Farabium 2025 | Interconnect | Kazakhstan | - | 17.93 PFlop/s | NVIDIA | #103 → |
| Moira (SSC) 2025 | Interconnect | Canada | - | 12.91 PFlop/s | NVIDIA | #137 → |
| Conrad (SSC) 2025 | Interconnect | Canada | - | 12.91 PFlop/s | NVIDIA | #138 → |
| Matrix 2025 | Interconnect | USA | - | 3.11 PFlop/s | NVIDIA | #406 → |
| TII Oryx 2025 | Interconnect | United Arab Emirates | - | 19.48 PFlop/s | NVIDIA | #90 → |
| VISION (TAMUS) 2025 | Interconnect | USA | - | 34.82 PFlop/s | NVIDIA | - |
| Nano 4 GB200 2025 | Interconnect | Taiwan | - | 5.16 PFlop/s | NVIDIA | - |
| amplitUDE 2024-12 | Interconnect | Germany | - | 1.95 PFlop/s | NVIDIA | #8 → |
| SuperPOD 2024-11 | Interconnect | United Arab Emirates | - | 55.81 PFlop/s | NVIDIA | #46 → |
| Capella 2024-11 | Interconnect | Germany | - | 24.06 PFlop/s | NVIDIA | #88 → |
| FPT AI Factory Vietnam 2024-11 | Interconnect | Vietnam | - | 46.65 PFlop/s | NVIDIA | #38 → |
| ROMEO-2025 2024-11 | Interconnect | France | - | 9.86 PFlop/s | NVIDIA | #2 → |
| JETI 2024-11 | Interconnect | Germany | - | 83.14 PFlop/s | NVIDIA | #6 → |
| CHIE-2 2024-10 | Interconnect | Japan | - | 89.78 PFlop/s | NVIDIA | #30 → |
| ASPIRE 2A+ 2024-10 | Interconnect | Singapore | - | 14.20 PFlop/s | NVIDIA | #90 → |
| Genkai 2024-10 | Interconnect | Japan | - | 6.22 PFlop/s | NVIDIA | #238 → |
| Vista (TACC) 2024-09 | Interconnect | USA | - | - | NVIDIA | - |
| Forerunner 1 2024-07 | Interconnect | Taiwan | - | 3.53 PFlop/s | NVIDIA | #359 → |
| Jean Zay H100 2024-06 | Interconnect | France | - | 52.18 PFlop/s | NVIDIA | #49 → |
| Virga 2024-06 | Interconnect | Australia | - | 14.94 PFlop/s | NVIDIA | #72 → |
| BioHive-2 2024-05 | Interconnect | USA | - | 23.32 PFlop/s | NVIDIA | #35 → |
| JEDI 2024-04 | Interconnect | Germany | - | 4.50 PFlop/s | NVIDIA | #1 → |
| IBM Blue Vela 2024-04 | Interconnect | USA | - | - | NVIDIA | - |
| Meta GenAI cluster (InfiniBand) 2024-03 | Interconnect | USA | - | - | NVIDIA | - |
| Marvin 2024-03 | Interconnect | Germany | - | 2.18 PFlop/s | NVIDIA | #34 → |
| CLAIX-2023 2024-01 | Interconnect | Germany | - | 5.24 PFlop/s | NVIDIA | #229 → |
| Njoerd 2024 | Interconnect | United Kingdom | - | 78.20 PFlop/s | NVIDIA | #35 → |
| CHIE-3 2024 | Interconnect | Japan | - | 91.94 PFlop/s | NVIDIA | #28 → |
| TSUBAME4.0 2024 | Interconnect | Japan | - | 39.62 PFlop/s | NVIDIA | #62 → |
| ARF-ACC 2024 | Interconnect | Turkey | - | 12.09 PFlop/s | NVIDIA | #145 → |
| Opera Iceland KEF-1 SuperPOD 2024 | Interconnect | Iceland | - | 10.53 PFlop/s | NVIDIA | #88 → |
| MUSICA 2024 | Interconnect | Austria | - | 31.84 PFlop/s | NVIDIA | #50 → |
| Helma 2024 | Interconnect | Germany | - | 32.22 PFlop/s | NVIDIA | #51 → |
| PITAGORA 2024 | Interconnect | Italy | - | 27.90 PFlop/s | NVIDIA | #44 → |
| Gautschi AI 2024 | Interconnect | USA | - | 7.26 PFlop/s | NVIDIA | #157 → |
| Cardinal 2024 | Interconnect | USA | - | 5.04 PFlop/s | NVIDIA | #211 → |
| THE CRUST 2.5 2024 | Interconnect | Thailand | - | 13.85 PFlop/s | NVIDIA | #92 → |
| CMCC Cassandra 2024 | Interconnect | Italy | - | - | NVIDIA | - |
| Lem 2024 | Interconnect | Poland | - | 12.80 PFlop/s | NVIDIA | #80 → |
| Kraken-Fregata 2024 | Interconnect | Poland | - | 5.98 PFlop/s | NVIDIA | #154 → |
| Proxima 2024 | Interconnect | Poland | - | 13.82 PFlop/s | NVIDIA | #93 → |
| Minerva 2024 | Interconnect | USA | - | 7.91 PFlop/s | NVIDIA | #147 → |
| Ubilink H100 Cluster 2024 | Interconnect | Taiwan | - | 45.82 PFlop/s | NVIDIA | #44 → |
| fortythree 2024 | Interconnect | Germany | - | 11.27 PFlop/s | NVIDIA | #42 → |
| AMD Ouranos 2024 | Interconnect | France | - | 2.99 PFlop/s | NVIDIA | #9 → |
| Reindeer 2024 | Interconnect | USA | - | 45.59 PFlop/s | NVIDIA | #32 → |
| JLab 24s 2024 | Interconnect | USA | - | - | NVIDIA | - |
| Alpha (Flatiron) 2024 | Interconnect | USA | - | 4.10 PFlop/s | NVIDIA | #243 → |
| Medusa (Dell) 2024 | Interconnect | USA | - | 3.08 PFlop/s | NVIDIA | #387 → |
| HOKUSAI BigWaterfall 2 2023-12-04 | Interconnect | Japan | - | - | NVIDIA | - |
| MareNostrum 5 ACC 2023-12 | Interconnect | Spain | - | 138.2 PFlop/s | NVIDIA | #8 → |
| NASA Cabeus 2023-12 | Interconnect | USA | - | - | NVIDIA | - |
| Eagle 2023-11 | Interconnect | USA | - | 561.2 PFlop/s | NVIDIA | #3 → |
| Eos 2023-11 | Interconnect | USA | - | 121.4 PFlop/s | NVIDIA | #23 → |
| Camphor 3 2023-10 | Interconnect | Japan | - | 6.71 PFlop/s | NVIDIA | #226 → |
| Nabuchodonosor 2023-10 | Interconnect | France | - | - | NVIDIA | - |
| NHN Cloud Gwangju AI 2023-10 | Interconnect | South Korea | - | 12.84 PFlop/s | NVIDIA | #97 → |
| AOBA-S 2023-08 | Interconnect | Japan | - | 17.22 PFlop/s | NVIDIA | #75 → |
| Ngarrgu Tindebeek 2023-08 | Interconnect | Australia | - | - | NVIDIA | - |
| CoreWeave AI cluster for Inflection AI 2023-06 | Interconnect | USA | - | - | NVIDIA | - |
| MareNostrum 5 GPP 2023 | Interconnect | Spain | - | 40.10 PFlop/s | NVIDIA | #61 → |
| DeepL Mercury 2023 | Interconnect | Sweden | - | 21.85 PFlop/s | NVIDIA | #34 → |
| SPR-HBM 2023 | Interconnect | USA | - | 6.46 PFlop/s | NVIDIA | #121 → |
| ISEG 2023 | Interconnect | Netherlands | - | 46.54 PFlop/s | NVIDIA | #15 → |
| Yep1 2023 | Interconnect | USA | - | 23.32 PFlop/s | NVIDIA | #34 → |
| Laurel 3 2023 | Interconnect | Japan | - | - | NVIDIA | - |
| Pre-Eos 2023 | Interconnect | USA | - | 40.66 PFlop/s | NVIDIA | #14 → |
| HiPerGator AI 2021 | Interconnect | USA | - | 17.49 PFlop/s | NVIDIA | #106 → |
| Berzelius 2021 | Interconnect | Sweden | - | 5.25 PFlop/s | NVIDIA | #82 → |
| Horizon | Interconnect | USA | - | - | NVIDIA | - |
| Santos Dumont | Interconnect | Brazil | - | 14.29 PFlop/s | NVIDIA | #52 → |
| Databricks DBRX Training Cluster | Interconnect | USA | - | - | NVIDIA | - |
| Della | Interconnect | USA | - | - | NVIDIA | - |
| AzInTelecom Supercomputer Center | Interconnect | Azerbaijan | - | - | NVIDIA | - |
| Clementina XXI | Interconnect | Argentina | - | 5.39 PFlop/s | NVIDIA | #199 → |
| Grand Chariot 2 | Interconnect | Japan | - | 4.38 PFlop/s | NVIDIA | #294 → |
| Pegasus (Tsukuba) | Interconnect | Japan | - | 4.34 PFlop/s | NVIDIA | #296 → |
| HKUST SuperPOD | Interconnect | China | - | - | NVIDIA | - |
| KAIROS | Interconnect | France | - | 3.05 PFlop/s | NVIDIA | #1 → |
| ACES (Texas A&M) | Interconnect | USA | - | - | NVIDIA | - |
| Marlowe (Stanford) | Interconnect | USA | - | - | NVIDIA | - |
| Triton | Interconnect | Finland | - | - | NVIDIA | - |
| Redtail | Interconnect | USA | - | 12.85 PFlop/s | NVIDIA | - |
| Nano 5 | Interconnect | Taiwan | - | 13.06 PFlop/s | NVIDIA | #135 → |
| Taipei-1 | Interconnect | Taiwan | - | 22.30 PFlop/s | NVIDIA | #38 → |
| Hops | Interconnect | USA | - | 10.03 PFlop/s | NVIDIA | #142 → |
| Lynx (Arizona) | Interconnect | USA | - | - | NVIDIA | - |
| Quest (Northwestern) | Interconnect | USA | - | - | NVIDIA | - |
| CUBOX Cray XD6500 | Interconnect | South Korea | - | 4.96 PFlop/s | NVIDIA | #155 → |
| DeepL Ceres | Interconnect | Sweden | - | 11.30 PFlop/s | NVIDIA | #107 → |
| Vader | Interconnect | USA | - | 3.14 PFlop/s | NVIDIA | #365 → |
| DAIS | Interconnect | Germany | - | 6.03 PFlop/s | NVIDIA | #247 → |
| GenomeDK | Interconnect | Denmark | - | - | NVIDIA | - |
| Voltage Park H100 Cloud | Interconnect | USA | - | - | NVIDIA | - |
| Niflheim | Interconnect | Denmark | - | - | NVIDIA | - |
| Tyche-hsg | Interconnect | USA | - | - | NVIDIA | - |
| LiDO4 | Interconnect | Germany | - | 3.14 PFlop/s | NVIDIA | - |
| Peano | Interconnect | Italy | - | 4.53 PFlop/s | NVIDIA | - |
| BNL Institutional Cluster Gen2 | Interconnect | USA | - | - | NVIDIA | - |
| PARAM Rudra (IIT Bombay) | Interconnect | India | - | - | NVIDIA | - |
“Units recorded” sums only the edges where a public source states a quantity. It is a floor, not a total, and should never be read as installed base.
Related news
- NVIDIA completes its acquisition of Mellanox · NVIDIATogether with HPE's acquisition of Cray seven months earlier, this ended the last two large independent fabric businesses and made accelerator share and interconnect share dependent variables.