Parts/CPUs/Intel Xeon Gold 6148
CPUs · Skylake-SP · 2017
Intel Xeon Gold 6148
20-core Xeon, two per ABCI node alongside four V100s.
Why this part
Skylake-SP brought AVX-512 down from Xeon Phi into the mainline server part, along with a mesh on-die topology replacing the ring. ABCI paired it with V100s in Japan's first large machine built explicitly for AI rather than simulation, which is why its published figures emphasise half precision and its cooling design assumes sustained rather than bursty load.
Systems using this part
- Systems
- 29
- Rmax underneath
- 79.43 PFlop/s
- Units recorded
- 10,231
8 of 29 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Banting 2020 | CPU | Canada | - | 2.68 PFlop/s | Intel | - |
| Daley 2020 | CPU | Canada | - | 2.60 PFlop/s | Intel | - |
| Béluga 2019-04 | CPU | Canada | - | 2.28 PFlop/s | Intel | #106 → |
| Hera 2019 | CPU | USA | - | 3.08 PFlop/s | Intel | #88 → |
| Orion (MSU HPC2) 2019 | CPU | USA | 1,800 | 3.67 PFlop/s | Intel | #62 → |
| G42 POD3 2019 | CPU | United Arab Emirates | - | 1.76 PFlop/s | Intel | #219 → |
| ABCI 2018-08 | CPU | Japan | - | 19.88 PFlop/s | Intel | #7 → |
| Kay 2018-08 | CPU | Ireland | - | 665.00 TFlop/s | Intel | - |
| AFI-NITY 2018-08 | CPU | Japan | 1,760 | 1.69 PFlop/s | Intel | #126 → |
| ATERUI II 2018-06 | CPU | Japan | - | 2.09 PFlop/s | Intel | #117 → |
| JFRS-1 2018-06 | CPU | Japan | 1,370 | 2.79 PFlop/s | Intel | #209 → |
| Taiwania 2018-05 | CPU | Taiwan | - | 1.33 PFlop/s | Intel | #148 → |
| Niagara 2018-03 | CPU | Canada | - | 3.60 PFlop/s | Intel | #70 → |
| Maui 2018 | CPU | New Zealand | - | 910.00 TFlop/s | Intel | #321 → |
| COBRA 2018 | CPU | Germany | - | 5.61 PFlop/s | Intel | #36 → |
| Electra 2018 | CPU | USA | 2,232 | 5.44 PFlop/s | Intel | #46 → |
| Aleph 2018 | CPU | South Korea | - | 971.00 TFlop/s | Intel | #445 → |
| Joule 2.0 2018 | CPU | USA | - | 3.61 PFlop/s | Intel | #239 → |
| Lemhi 2018 | CPU | USA | - | 1.00 PFlop/s | Intel | #427 → |
| SUSTech TaiYi 2018 | CPU | China | - | 1.69 PFlop/s | Intel | #127 → |
| IMS Molecular Simulator 2017-10-01 | CPU | Japan | - | 1.79 PFlop/s | Intel | #261 → |
| HOKUSAI BigWaterfall 2017 | CPU | Japan | 1,680 | 1.60 PFlop/s | Intel | #83 → |
| Quriosity (2017) 2017 | CPU | Germany | - | 1.75 PFlop/s | Intel | #66 → |
| Zenith (Dell EMC) 2016 | CPU | USA | 844 | 1.01 PFlop/s | Intel | #265 → |
| Emmy | CPU | Germany | 449 | 5.95 PFlop/s | Intel | #47 → |
| PARAM Shakti | CPU | India | - | - | Intel | - |
| Triton | CPU | Finland | 96 | - | Intel | - |
| Hawk (Cardiff) | CPU | United Kingdom | - | - | Intel | - |
| Sunbird | CPU | United Kingdom | - | - | Intel | - |
“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.