Parts/CPUs/IBM POWER3-II
Why this part
Copper interconnect and silicon-on-insulator, both of which IBM brought to production before anyone else, and both aimed at the same problem: pushing clock speed without the power going vertical. ASCI White used 8,192 of these in 512 sixteen-way SMP nodes, which is a very different shape from the thousands of single-socket nodes its contemporaries used. The reasoning was software: weapons codes written for shared memory ported to a fat SMP node far more easily than to a distributed-memory mesh, and buying fewer, larger nodes bought time for the application rewrite.
Systems using this part
- Systems
- 3
- Rmax underneath
- 8.92 TFlop/s
- Units recorded
- 3,328
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Seaborg 2001 | CPU | USA | 3,328 | 3.05 TFlop/s | IBM | #5 → |
| ASCI White 2000-06 | CPU | USA | - | 4.94 TFlop/s | IBM | #1 → |
| Blue Horizon 2000 | CPU | USA | - | 929.0 GFlop/s | IBM | #8 → |
“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.