Parts/CPUs/Intel Xeon Platinum 8480C
CPUs · Sapphire Rapids · 2023
Intel Xeon Platinum 8480C
48-core host CPU paired with H100 in the Azure NDv5 node.
Why this part
The default host under an H100 in 2023, and the part beneath both Microsoft's Eagle and NVIDIA's own Eos. Those two are worth reading together: an operator that rents capacity and a vendor that sells the accelerators, running near-identical hardware, neither of them a national laboratory. That is the largest structural change in this dataset over its thirty-year span.
Systems using this part
- Systems
- 27
- Rmax underneath
- 1.53 EFlop/s
- Units recorded
- 270
2 of 27 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Shakti Cloud 2025-11 | CPU | India | - | 84.31 PFlop/s | Intel | #32 → |
| Nano 4 2025-11 | CPU | Taiwan | - | 81.55 PFlop/s | Intel | #33 → |
| Hopper NUS 2025-06 | CPU | Singapore | - | 15.05 PFlop/s | Intel | #120 → |
| VISION (TAMUS) 2025 | CPU | USA | 190 | 34.82 PFlop/s | Intel | - |
| SuperPOD 2024-11 | CPU | United Arab Emirates | - | 55.81 PFlop/s | Intel | #46 → |
| Gefion 2024-10 | CPU | Denmark | - | 66.59 PFlop/s | Intel | #43 → |
| CHIE-2 2024-10 | CPU | Japan | - | 89.78 PFlop/s | Intel | #30 → |
| ASPIRE 2A+ 2024-10 | CPU | Singapore | 80 | 14.20 PFlop/s | Intel | #90 → |
| Arka AI/ML 2024-09 | CPU | India | - | 2.70 PFlop/s | Intel | #400 → |
| Forerunner 1 2024-07 | CPU | Taiwan | - | 3.53 PFlop/s | Intel | #359 → |
| BioHive-2 2024-05 | CPU | USA | - | 23.32 PFlop/s | Intel | #35 → |
| CHIE-3 2024 | CPU | Japan | - | 91.94 PFlop/s | Intel | #28 → |
| Opera Iceland KEF-1 SuperPOD 2024 | CPU | Iceland | - | 10.53 PFlop/s | Intel | #88 → |
| Gautschi AI 2024 | CPU | USA | - | 7.26 PFlop/s | Intel | #157 → |
| Reindeer 2024 | CPU | USA | - | 45.59 PFlop/s | Intel | #32 → |
| Eagle 2023-11 | CPU | USA | - | 561.2 PFlop/s | Intel | #3 → |
| Eos 2023-11 | CPU | USA | - | 121.4 PFlop/s | Intel | #23 → |
| Israel-1 2023-11 | CPU | Israel | - | 41.50 PFlop/s | Intel | #34 → |
| MareNostrum 5 GPP 2023 | CPU | Spain | - | 40.10 PFlop/s | Intel | #61 → |
| DeepL Mercury 2023 | CPU | Sweden | - | 21.85 PFlop/s | Intel | #34 → |
| Yep1 2023 | CPU | USA | - | 23.32 PFlop/s | Intel | #34 → |
| Pre-Eos 2023 | CPU | USA | - | 40.66 PFlop/s | Intel | #14 → |
| Berzelius 2021 | CPU | Sweden | - | 5.25 PFlop/s | Intel | #82 → |
| UltraViolet 2018 | CPU | USA | - | 5.56 PFlop/s | Intel | #225 → |
| Nano 5 | CPU | Taiwan | - | 13.06 PFlop/s | Intel | #135 → |
| Taipei-1 | CPU | Taiwan | - | 22.30 PFlop/s | Intel | #38 → |
| DeepL Ceres | CPU | Sweden | - | 11.30 PFlop/s | Intel | #107 → |
“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.