After 9 years beyond the Sino-American Japan “FuYue” super-e-class crown who can take the lead?

富岳(ふがく),既是日本精神文化象征之一、日本第一高峰“富士山”的别称,是二战时期日本中岛飞机公司设计的战略轰炸机,也是日本超级计算机 Supercomputer Fugaku。而正是这款超级计算机,使得日本时隔 9 年后重登全球超算宝座。

On June 22, 2020, local time, the latest global list of top500 high-performance computers was released, with the Japanese supercomputer Fugaku topping the list. The 2nd to 5th places on the list are The Summit developed by IBM, the Sierra developed by IBM, the “Shenwei Taihu Light” developed by china’s National Parallel Computer Engineering Technology Research Center, and the Tianhe II development by China National Defense University of Science and Technology.

“Fu Yue” Opens “Post-Beijing Era”

A computer that can perform massive amounts of data and perform intensive computing operations that ordinary PCs can’t handle, we call them supercomputers.

Many supercomputers around the world are good and bad, and their performance can be based on a uniform standard , the TOP500 list. The TOP500 list began in 1993 and updates the rankings in June and November each year by benchmarking and refining the world’s 500 best-performing supercomputers.

Japan’s Fugaku is the world’s No.

When you talk about Fugaku, you have to start with another Japanese super-calculated “K Computer” – in June 2011, the top-ranked super-calculation was Japan’s “King”.

In fact, the name “King” makes sense. In Japanese, 1 Beijing represents 1 000 times 1 trillion.

The speed measurement unit of supercomputer is “Flops” per second floating-point operation, which is also a key indicator of supercomputing performance. At that time, “King” became the world’s fastest supercomputer with 8,162 trillion computing speeds per second.

In November of the same year, “Beijing” successfully accelerated, with a maximum computing performance of 10.51 PFlops, or 105.1 million floating-point calculations per second, the first time that humans have crossed the 100 million computing mark.

It is reported that the “King” processor is the number one IT manufacturer in Japan Fujitsu manufactured SPARC64 VIIIfx, eight core, main frequency 2.0GHz, thermal design power consumption 58W, peak floating point performance 128GFlops.

In 2012, Jing was officially operational at its institute of Science and Chemistry in Kobe, Japan. Since its official launch, more than 11,000 researchers from universities or businesses have used Beijing for space exploration, weather prediction, and new material development.

But at the same time, with the continuous refresh of the performance of countries, the wave continues to emerge, “Beijing” in the TOP500 list also gradually lost the glory of the past.

On August 16, 2019, researchers and businesses stopped using “Jing”. On August 30, “King” was removed after powering it down, and its historic mission was completed.

However, as early as 2014, Fujitsu and the Japan Institute of Science and Chemistry began the design and development of Fugaku, the successor to The Initaku, with the aim of ushering in the “post-Beijing era”.

Timeline published by the Japan Institute of Science and Chemistry

In fact, the goal of the Japanese study of Fugaku is clear – the computational performance is 100 times higher than that of Jing. To this end, the total budget is also known to reach 130 billion yen.

In terms of configuration, Fugaku consists of approximately 400 computers, each weighing about 2 tons, about 2 meters high, 1.4 meters deep and 80 centimeters wide. Fugaku’s 48-core Fujitsu chip A64FX is the first ARM-based supercomputing in history.

Fugaku is exceptional, reaching the TOP500 list with a peak speed of 4.155 billion floats per second (2.8 times that of the runner-up Summit supercomputer), the first time Japan has returned to the top of the global super-computing since “Jing” in 2011.

It is understood that Fugaku is suitable for deep learning and other AI areas of the wide range of applications, comprehensive performance, can be said to be epoch-making super-calculation.

Fugaku was originally scheduled to open in April 2021, but due to the outbreak of the new crown, Fugaku was activated early due to the outbreak of the new crown, which is mainly used for outbreak prevention and control at this stage.

Multi-country competition in the field of overcomputing

In fact, supercomputers in meteorology, life sciences, military, space and many other fields have made important contributions, can be said to be a country’s scientific and technological strength in one, so it is often referred to as the “national heavy instrument.”

All along, the field of supercomputing has maintained a pattern of fierce competition between China and the United States.

Prior to this ranking update, Top 5 was the Summit developed by IBM, the Sierra developed by IBM, the “Shenwei Taihu Light” in China, Tianhe II in China, and the Frontera developed by DELL EMC.

Among them, Summit has been at the top of the list since the first half of 2018.

Before The Summit became the real Summit, it was china’s ten-game winning streak – “Tianhe 2” won six in a row since June 2013, and then “Shenwei Taihu Light” became the world’s first to reach a billion-times super-calculation, surpassing “Tianhe 2” and four straight wins.

In the latest list, after China and the United States, Italy, Switzerland, Germany, France, South Korea and other countries among the top 20, it can be seen that the field of super-computing competition is very fierce.

Comparison of the world’s top number of super-calculated data

The latest rankings have just emerged, and Horst Simon, executive director of the Lawrence Berkeley National Laboratory in the United States, said:

Given the overcalculations of the U.S. Department of Energy’s Oak Ridge National Laboratory and Lawrence Livermore National Laboratory (respectively, Summit and Sierra), and China’s overcalculations, Fugaku’s position as the world’s fastest supercomputer won’t last long.

It’s worth noting that Horst Simon’s reference to China’s supercomputing refers to the E-class supercomputer, the 10 billion-time supercomputer, which is also seen by the industry as “the next crown of the supercomputer world”.

As early as 2016, China’s first 10 billion times over-computing prototype system development work has been officially started.

In the development of E-class supercomputers, China is likely to become the fastest developed country. There are several e-level supercomputer projects in China:

Tianhe 3: On May 19, 2018, the National Supercomputing Tianjin Center unveiled its prototype, with its fully configured host scheduled to be operational in 2020;

Shenwei E-Class: In July 2018, its prototype will be located in the National Supercomputing Jinan Center, which is scheduled to be put into operation in 2021;

Sugon E: On July 4, 2016, Sinoco Sugon announced the official launch of its Class E Super-Prototype System project, which will be finally deployed beyond 2020.

Recently, at the 4th World Intelligent Congress “Cloud Smart Technology Exhibition”, the “Tianhe 3” prototype was unveiled as a 3D model for the first time, causing widespread concern.

It is worth noting that the United States, Japan and other countries are also actively layout E-class supercomputers. It remains to be seen who will be the first country to make history by developing a successful E-class supercomputer, from the “Beijing” that crossed the 100 million-magnitude threshold for the first time to the first “Shenwei Taihu Light” that has reached 1 billion calculations.