For the fourth year in a row China’s Milkyway 2 (also known as Tianhe-2) retained its ranking as the world’s fastest supercomputer on the most recent Top500 list with the same 33.86 petaflop/s (quadrillions of calculations per second) performance record it set in June 2013.
The Top500 list recorded no changes in the top nine systems since the June 2014 edition and similarly low performance growth rates in the lower portion of the list. A new Cray CS-Storm system installed at an undisclosed U.S. government site entered the list at number 10, at 3.57 petaflop/s. The November 2014 list reported the lowest turnover rate in two decades.
The fastest supercomputers on the biannual list are measured by Linpack, the favored benchmark that many have said is losing relevance in determining the true performance measurement of modern high-performance computing systems. The High Performance Conjugate Gradient (HPCG) metric was introduced last year, offering an alternative HPC metric, that better correlates computation and data access patterns found in modern applications. The two metrics could both be used to evaluate systems and judge true performance.
Other statistics noted in the November list showed continued growth in the use of accelerator/co-processor technology (NVIDIA GPU, Intel Xeon Phi). Systems by HP, IBM, and Cray continued to dominate the list. A total of 50 systems now have greater than 1 petaflop/s performance, up from 37 six months ago.