2017 TOP500 Supercomputer Rankings and Energy‑Efficiency Insights

In the 2017 TOP500 supercomputer ranking, China’s Sunway TaihuLight reclaimed the top spot with a peak performance of 125,435.9 TFlop/s, followed by Tianhe‑2 at 54,902.4 TFlop/s, while Switzerland’s Piz Daint took third with 25,326.3 TFlop/s.

High‑performance‑computing (HPC) statistics are crucial for manufacturers, users, and prospective buyers; since 1986, Hans Meuer has published annual data based on the Mannheim HPC summit, reflecting the diversity and performance gaps among supercomputers.

The TOP500 list, formally maintained since 1993 by U.S. and German experts, is updated twice yearly and ranks systems using the Linpack benchmark, aiming to list general‑purpose high‑end machines while excluding specialized or invalid configurations.

The current top six includes three Cray‑provided systems; NERSC (U.S. Department of Energy’s National Energy Research Scientific Computing Center) exemplifies this with its Cori system that employs Cray’s DataWarp‑based Burst Buffer technology to boost I/O performance.

NER​SC’s mission is to accelerate scientific discovery across domains such as photosynthesis modeling, climate simulation, combustion, magnetohydrodynamics, astrophysics, and computational biology.

In 2017 the United States contributed 169 entries (still the highest total) and China 159, with vendors like HP, Lenovo, Cray, Sugon, IBM, Inspur, and Huawei occupying the leading positions.

Energy efficiency is measured by GFlops per watt; the Green500 list, also based on Linpack, ranks supercomputers by this metric and is published twice a year.

The 2017 Green500 top five contained no Chinese systems; Sunway TaihuLight achieved 6.051 GFlops/Watt (rank 17) and Tianhe‑2 fell to rank 117 with 2.078 GFlops/Watt.

Overall, China’s HPC capabilities and technology reserves are noteworthy, especially the contributions of Lenovo, Sugon, Inspur, and Huawei, which supply several systems within the TOP 13, 14, and 95.

Compared with TOP500, Green500 places greater emphasis on power consumption; improving efficiency can involve better cooling, power‑monitoring systems, and other measures.

Practices from SGI, NEC, NVIDIA, and ZettaScaler offer valuable lessons for domestic vendors as they prepare for emerging HPDA, cloud, and HPC‑HyperScale trends that demand both high performance and green energy efficiency.