Sunday, August 02, 2015
EXclusive: China develops 270 PFlops supercomputer
100 PFlops China should reach in 2015/6, and 270 PFlops in 2017/8 to compete with US Summit.
99% BAD HARDWARE WEEK: Each Summit node will contain multiple IBM POWER9 CPUs and NVIDIA Volta GPUs all connected together with NVIDIA’s high - speed NVLink and a huge amount of memory . Each node will have over half a terabyte of coherent memory (HBM “high bandwidth memory” + DDR4) addressable by all CPUs and GPUs , plus an additional 800 gigaby tes of NVRAM.
Well, nothing new for China. Beside NVRAM, it will use OWN, EXPORT BANNED FUTURE 8 TFlops accelerators. At the same number of Xeon nodes as at Tianhe 2, that should bring 8 times more average than its LINPACK Rmax od 33,4 PFlops. And that is exactly 270 PFlops. Test version should brake 100 Pflops by November this year or June 2016. Micron's new NVmemory was a God's gift, so China offered hefty $23B on 21st July.
Summit as mentioned before should deliver 10 times more performance than 27 PFlops Titan. That is the same 270 Pflops in 2017/8. Thus, the sumpercomputing race is continuing on.