Parts/CPUs/IBM PowerPC A2 (Power BQC)
Why this part
Eighteen cores where only sixteen do user work: one runs the operating system and one is a spare, held in reserve so that a manufacturing defect in a single core does not scrap the chip. At Sequoia's scale, 98,304 of these, yield economics stop being the vendor's private problem and start determining whether the machine can be built at all. Each core ran four hardware threads to hide memory latency, and the network logic sat on the same die as the cores, so BlueGene/Q has no separate interconnect vendor anywhere in its bill of materials. That is why fabric-share charts that read only the interconnect field understate IBM's historical position.
Systems using this part
- Systems
- 14
- Rmax underneath
- 43.09 PFlop/s
- Units recorded
- -
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Blue Brain IV 2013 | CPU | Switzerland | - | 715.55 TFlop/s | IBM | #48 → |
| Fermi 2012-08 | CPU | Italy | - | 1.79 PFlop/s | IBM | #7 → |
| Mira 2012-06 | CPU | USA | - | 8.59 PFlop/s | IBM | #3 → |
| Sequoia 2012-06 | CPU | USA | - | 17.17 PFlop/s | IBM | #1 → |
| Avoca 2012-06 | CPU | Australia | - | 715.55 TFlop/s | IBM | #101 → |
| JUQUEEN 2012 | CPU | Germany | - | 4.14 PFlop/s | IBM | #5 → |
| Vulcan 2012 | CPU | USA | - | 4.29 PFlop/s | IBM | #9 → |
| Blue Joule 2012 | CPU | United Kingdom | - | 1.21 PFlop/s | IBM | #16 → |
| Turing (IDRIS) 2012 | CPU | France | - | 1.07 PFlop/s | IBM | #29 → |
| DiRAC BlueGene/Q 2012 | CPU | United Kingdom | - | 1.07 PFlop/s | IBM | #20 → |
| KEK SAKURA 2012 | CPU | Japan | - | 536.66 TFlop/s | IBM | #36 → |
| Zumbrota 2012 | CPU | France | - | 715.55 TFlop/s | IBM | #30 → |
| Cetus 2012 | CPU | USA | - | 715.55 TFlop/s | IBM | #55 → |
| Vesta (Argonne) 2012 | CPU | USA | - | 357.78 TFlop/s | IBM | #100 → |
“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.