Parts/Accelerators/NVIDIA Tesla M2050
Accelerators · Fermi · 2010
NVIDIA Tesla M2050
3 GB GDDR5 Fermi module. The accelerator behind the first Chinese No. 1, Tianhe-1A.
Why this part
Fermi was the first NVIDIA architecture with ECC memory and a serious double-precision rate, which is what turned the GPU from a visualisation device into something a national laboratory would put in a procurement. Tianhe-1A used 7,168 of them and took the top of the list in November 2010, four years before any American machine did the same thing. The accelerator era begins here, and it began in China.
Systems using this part
- Systems
- 4
- Rmax underneath
- 3.98 PFlop/s
- Units recorded
- 11,804
3 of 4 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| HokieSpeed 2011-12 | Accelerator | USA | - | 120.40 TFlop/s | NVIDIA | #147 → |
| Tianhe-1A 2010-11 | Accelerator | China | 7,168 | 2.57 PFlop/s | NVIDIA | #1 → |
| TSUBAME2.0 2010-11 | Accelerator | Japan | 4,224 | 1.19 PFlop/s | NVIDIA | #4 → |
| Edge (LLNL) 2010 | Accelerator | USA | 412 | 100.50 TFlop/s | NVIDIA | #73 → |
“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.