EArly in 2016, China had announced plans to build three different pre-exascale systems with three very different architectures, according to some Tweets put out by James Lin, vice director for the Center of HPC at Shanghai Jiao Tong University.
The three-way horse race for exascale machines in China will set up a horse race between three different organizations to build pre-exascale clusters based on ARM, Shenwei, and AMD (presumably Opteron) technologies. The first pre-exascale machine is being created by NUDT and will use ARM-based processors and will be deployed at the national supercomputer center in Tianjin where the Tianhe-1A CPU-GPU hybrid was deployed in 2010 and gave China its first top spot on the Top 500 rankings of supercomputers. There is no mention of using the Matrix2000 DSP accelerator with this system, but unless NUDT plans to create its own ARM chip with a homegrown floating point accelerator and embed it on the die, it stands to reason that this first pre-exascale machine will be an ARM-DSP hybrid.
The second pre-exascale machine is being developed by the same people who put together the Sunway TaihuLight system, and it will be deployed in the national supercomputing center in Jinan, where its predecessor, the Sunway Bluelight system, currently runs.
The third pre-exascale machine, and perhaps equally interesting, will be built by Chinese system maker Sugon and will employ an X86 processor licensed from AMD. We presume this is a licensed variant of the future “Zen” Opteron chip, due in 2017 for servers. It is not clear who is doing the licensing of the X86 technology from AMD, but back in April, AMD announced that it had inked a deal worth $293 million to license X86 chip technology to Tianjin Haiguang Advanced Technology Investment Co, which is itself an investment consortium that is guided by the Chinese Academy of Sciences.
In May, China committed to delivering an exascale-class machine by 2020 with 10 PB of memory, exabytes of storage, and 30 gigaflops per watt efficiency (about five times better than the new Sunway TaihuLight system), and greater than 60 percent efficiency on the Linpack Fortran benchmark test.
3 prototype systems for exascale will be ready by end of 2017 in China. Each has ~2.5PF in Peak and ~500-600 nodes.
The winner will be chosen to build the “exascale system” in peak performance by 2020.
One is by NUDT with ARM approach to be deployed in Tianjing national center, where hosts Tianhe-1A.
Another is by Taihulight team with the next generation SW CPU to be deployed on Jinan national center, where hosts Sunway bluelight.
The third is by Sugon with AMD licenced x86 CPU to be deployed in both Shanghai supercomputer center and Shenzhen national center.
“A complete [exaFLOP] computing system of the exascale supercomputer and its applications can only be expected in 2020, and will be 200 times more powerful than the country’s first petaflop computer Tianhe-1, recognized as the world’s fastest in 2010,” said Zhang Ting , Application engineer with the Tianjin-based National Supercomputer Center, when attending the sixth session of the 16th Tianjin Municipal People’s Congress Tuesday.
Exascale computers are capable of at least 1 quintillion (a billion billion) calculations per second.
Zhang said that using the exascale computer for cloud computing and big data applications, China could spur ahead with many key innovation and high-tech programs.
In June 2016, China revealed its fastest new supercomputer – the Sunway TaihuLight – with a peak performance of 124.5 petaflops, the world’s first system to exceed 100 petaflops.
China has been steadily building its supercomputing capacity, and independently developed all key technology including microprocessors.
Zhang said the next-generation exascale computer will not only lead in calculation speed, but also in data transmission efficiency.
Previously China had targeted 2018 for the exascale prototype. So developments are tracking a few months ahead of schedule.
The US the Department of Energy is running the Exascale Computing Project, which is looking to increase on the petascale level computers that exist now. The project awarded $39.8 million of grants to partners developing the technology necessary to build an exascale computer by 2023.
China’s Sunway TaihuLight is the current world’s fastest supercomputer which has a processing speed of 93 petaflops. At its peak, the computer can perform 93,000 trillion calculations per second. In total, 167 of the most powerful 500 computers in the world reside in China.
The US is developing a number of supercomputers that would be capable of beating the Sunway TaihuLight – a 200 petaflop machine called Summit is being developed at the Oak Ridge National Lab and is due to arrive in 2018. Japan is also heavily investing in supercomputing technology and has said it will spend 19.5 billion yen (£139 million) on a 130 petaflop computer.
SOURCES – Xinhau, Wired UK