Parts/CPUs/Intel Xeon E5-2690 v3
CPUs · Haswell-EP · 2014
Intel Xeon E5-2690 v3
12-core host CPU in the Cray XC50 nodes of Piz Daint.
Why this part
Haswell-EP was the first Xeon generation on DDR4 and the first with AVX2, doubling the vector width for fused multiply-add. Piz Daint paired it one-to-one with a P100, which is the configuration that became standard for the following five years: a mid-range CPU per GPU, not a large one, because the arithmetic has already left the socket.
Systems using this part
- Systems
- 16
- Rmax underneath
- 38.75 PFlop/s
- Units recorded
- 2,480
3 of 16 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Piz Daint 2017-06 | CPU | Switzerland | - | 21.23 PFlop/s | Intel | #68 → |
| Okeanos 2016-07 | CPU | Poland | 2,168 | 909.61 TFlop/s | Intel | #109 → |
| Lengau 2016-06 | CPU | South Africa | - | 1.03 PFlop/s | Intel | #122 → |
| ASPIRE 1 2016 | CPU | Singapore | - | 1.01 PFlop/s | Intel | #94 → |
| Lonestar 5 2016 | CPU | USA | - | 1.02 PFlop/s | Intel | #79 → |
| NOAA Luna 2015-12 | CPU | USA | - | 1.64 PFlop/s | Intel | #53 → |
| NOAA Surge 2015-12 | CPU | USA | - | 1.64 PFlop/s | Intel | #42 → |
| Nuri 2015 | CPU | South Korea | - | 2.40 PFlop/s | Intel | #29 → |
| Miri 2015 | CPU | South Korea | - | 2.40 PFlop/s | Intel | #30 → |
| BURA 2015 | CPU | Croatia | 288 | 233.56 TFlop/s | Intel | #441 → |
| Piz Kesch 2015 | CPU | Switzerland | 24 | 282.50 TFlop/s | Intel | #353 → |
| Uri 2014-12 | CPU | South Korea | - | 345.60 TFlop/s | Intel | #148 → |
| Magnus 2014 | CPU | Australia | - | 1.10 PFlop/s | Intel | #42 → |
| Occigen 2014 | CPU | France | - | 1.63 PFlop/s | Intel | #27 → |
| Cartesius 2013-06 | CPU | Netherlands | - | 1.09 PFlop/s | Intel | #46 → |
| ATERUI I 2013-04-01 | CPU | Japan | - | 801.40 TFlop/s | Intel | #64 → |
“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.