Parts/CPUs/AMD EPYC 9004 (Genoa)
Why this part
Zen 4 moved to DDR5 and twelve memory channels, which is the number that matters more than the core count for most scientific codes: bandwidth per socket rose by about half against Milan. Shaheen III at KAUST runs it CPU-only in one partition. AMD encodes the series in the last digit of the model number rather than the first two, so 9654 is Genoa and 9655 is the following generation, a detail that silently mislabelled parts in this repository's parser until it was found.
Systems using this part
- Systems
- 47
- Rmax underneath
- 456.9 PFlop/s
- Units recorded
- 2,196
4 of 47 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Fran 2026-10-01 | CPU | USA | - | - | AMD | - |
| Sofia (VSC Tier-1) 2026-07 | CPU | Belgium | - | 8.19 PFlop/s | AMD | - |
| Harpia 2025-10 | CPU | Brazil | - | 75.20 PFlop/s | AMD | #37 → |
| Trillium 2025-08-07 | CPU | Canada | - | 9.49 PFlop/s | AMD | #175 → |
| AI-03 2025-06 | CPU | United Arab Emirates | - | 54.12 PFlop/s | AMD | #47 → |
| Aire (Leeds) 2025-02-06 | CPU | United Kingdom | - | - | AMD | - |
| Ursa 2025 | CPU | USA | - | - | AMD | - |
| Space HPC 2025 | CPU | Italy | - | 4.47 PFlop/s | AMD | #261 → |
| SELENA 2025 | CPU | France | - | 5.42 PFlop/s | AMD | #260 → |
| Terabit HPC Bubble 2025 | CPU | Italy | - | 2.91 PFlop/s | AMD | #437 → |
| Capella 2024-11 | CPU | Germany | - | 24.06 PFlop/s | AMD | #88 → |
| Joule 3.0 2024-06 | CPU | USA | - | 4.86 PFlop/s | AMD | #181 → |
| Helios 2024-04 | CPU | Poland | - | 19.14 PFlop/s | AMD | #96 → |
| TSUBAME4.0 2024 | CPU | Japan | - | 39.62 PFlop/s | AMD | #62 → |
| CATT 2024 | CPU | USA | - | 8.27 PFlop/s | AMD | #140 → |
| Blueback 2024 | CPU | USA | - | 10.30 PFlop/s | AMD | #139 → |
| MUSICA 2024 | CPU | Austria | - | 31.84 PFlop/s | AMD | #50 → |
| Helma 2024 | CPU | Germany | - | 32.22 PFlop/s | AMD | #51 → |
| Kuma 2024 | CPU | Switzerland | - | 12.31 PFlop/s | AMD | #102 → |
| THE CRUST 2.5 2024 | CPU | Thailand | - | 13.85 PFlop/s | AMD | #92 → |
| Kraken-Fregata 2024 | CPU | Poland | - | 5.98 PFlop/s | AMD | #154 → |
| Proxima 2024 | CPU | Poland | - | 13.82 PFlop/s | AMD | #93 → |
| CRIEPI HPE Cray XD225v 2024 | CPU | Japan | - | 2.37 PFlop/s | AMD | #466 → |
| SEAS H100 (Harvard) 2024 | CPU | USA | - | 3.38 PFlop/s | AMD | #298 → |
| Shaheen III 2023-11 | CPU | Saudi Arabia | - | 35.66 PFlop/s | AMD | #64 → |
| Gaea 2023-04 | CPU | USA | - | - | AMD | - |
| Carpenter 2023 | CPU | USA | - | 11.62 PFlop/s | AMD | #84 → |
| Skoda C24 2023 | CPU | Czechia | - | 8.40 PFlop/s | AMD | #112 → |
| Volkswagen LX HPC 224Rk-1 2023 | CPU | Norway | - | 2.78 PFlop/s | AMD | #310 → |
| GSI Virgo Cluster 2020-08 | CPU | Germany | 188 | - | AMD | - |
| Mahuika 2018 | CPU | New Zealand | - | - | AMD | - |
| JASMIN 2012 | CPU | United Kingdom | - | - | AMD | - |
| Santos Dumont | CPU | Brazil | 120 | 14.29 PFlop/s | AMD | #52 → |
| Rudens | CPU | Latvia | - | - | AMD | - |
| Shell Houston HPC Cluster | CPU | USA | 1,728 | - | AMD | - |
| Kestrel | CPU | USA | - | - | AMD | - |
| Iridis 6 | CPU | United Kingdom | - | - | AMD | - |
| Oscar (Brown) | CPU | USA | - | - | AMD | - |
| Afton (UVA) | CPU | USA | - | - | AMD | - |
| Irgetas | CPU | Kazakhstan | - | - | AMD | - |
| CSF3 | CPU | United Kingdom | - | - | AMD | - |
| Vera (Chalmers) | CPU | Sweden | - | - | AMD | - |
| GenomeDK | CPU | Denmark | - | - | AMD | - |
| Ruth | CPU | USA | - | - | AMD | - |
| Barfoot | CPU | USA | - | - | AMD | - |
| Niflheim | CPU | Denmark | 160 | - | AMD | - |
| Marlyn | CPU | USA | - | 2.31 PFlop/s | AMD | #500 → |
“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.