Re:When It Comes to China, Google’s Experience Still Says It...
by zerohp on Wednesday August 03, 2016 @01:54PM
(#12220001)
As someone who spends a lot of time studying computer architecture, I'm skeptical about this machine because of its interconnect design and memory bandwidth. It seems to be optimized specifically for LINPACK. Many people are critical of that benchmark and some labs refuse to participate in Top 500 because of it.
From [1]:
>However, as we know LINPACK does not tell the whole story. On the HPCG benchmark, Sunway TaihuLight reported only .371 petaflops, which is .3 percent of peak. Compare this with 0.580 petaflops on Tianhe-2 (1.1 percent of peak) and .322 petaflops on Titan (1.2 percent of peak). RIKEN’s K computer reports 0.460 HPCG performance, 4.1 percent theoretical peak.
From [1]:
>However, as we know LINPACK does not tell the whole story. On the HPCG benchmark, Sunway TaihuLight reported only .371 petaflops, which is .3 percent of peak. Compare this with 0.580 petaflops on Tianhe-2 (1.1 percent of peak) and .322 petaflops on Titan (1.2 percent of peak). RIKEN’s K computer reports 0.460 HPCG performance, 4.1 percent theoretical peak.
1: https://www.hpcwire.com/2016/06/19/china-125-petaflops-sunwa...
EDIT: To be more blunt, this machine might have more PR value than scientific value.