China's supercomputing power

1.

The E-Class prototype of tianhe-3, a new generation of 10 billion times supercomputer independently developed by China, has recently completed the development and deployment, and successfully passed the sub item acceptance. The prototype system adopts three kinds of domestic independent high-performance computing and communication chips, and the computing speed of its six cabinets is equivalent to the computing power of 120 cabinets of Tianhe-1

When it comes to domestic chips, Chinese people first think that it must be Loongson. Loongson is not the earliest domestic processor, nor the most successful domestic processor. However, when it comes to domestic chips with the highest exposure rate, Loongson undoubtedly has developed three generations of Loongson processor chips from Loongson 1 to Loongson 3, But most of them only live in the exhibition stands where CAS shows its achievements to the outside world. There are almost no commercial applications, and they can't keep up with the mainstream performance. Often, new models of chips fall behind as soon as they come out. Although Shenwei and Feiteng both belong to the national team, the latter two are still in the first echelon of the national chip team. Although Loongson is famous, it has no chance with the first echelon

2. Unknown_Error

3. On June 19, Beijing time, the latest global top 500 supercomputers list was officially released: China's "Shenwei · Taihu Lake light" and "Tianhe 2" jointly won the third place, and the US "Titan" fell out of the top three for the first time in more than 20 years
Fu haohuan, deputy director of China's National Supercomputing Center in Wuxi and associate professor of Tsinghua University, told the media that China's domestic system has won the first place in the world three times in a row, highlighting China's independent research and development capability in the field of supercomputing; Since last year, he has made a breakthrough in the Gordon Bell Award, which represents the highest level in the field of high-performance computing applications. It also shows that China's hardware and software research and development have improved simultaneously
in the future, with the support of domestic software, the rapidly developing domestic hardware will bring more powerful scientific research ability and engineering test ability to various fields, promote instrial upgrading, and form the sustainable development of domestic supercomputing
in addition, the floating-point computing speed of "Shenwei · Taihu Lake light" of China's National Supercomputing Center in Wuxi is 930 million times per second, which is described by the list editor as "undoubtedly the most powerful digital computing machine on the planet" &; More importantly, all of its core components are made in China; The floating-point operation speed of tianhe-2 developed by China University of Defense Science and technology is 339 million times per second, still ranking second this time

4. There is only one Supercomputing Center in China. It's a private enterprise, controlled by the party and authorized by the military. At present, supercomputing is only available for national key projects, and some enterprises related to military science and technology will also use it one after another. However, it will take several years for ordinary people to use it. China needs a large number of modern manufacturers to process it. At that time, China will have a series of domestic procts such as big brand computers, mobile phones, televisions and automobiles with independent intellectual property rights, We don't need to import.

5. Mainly e to the CPU design and CPU processing problems
CPU processing requires not only a very good process, but also the manufacturing of high-performance micro nano electronic devices and instruments. These instrial instruments and equipment cost hundreds of millions of dollars. There is also a huge existing interest chain behind it. For example, lithography machine is the core device of CPU. The best photoresist in the world is made by ASML in the Netherlands, which has been supplied to a few enterprises such as Intel and Qualcomm for a long time. The Dutch dare not offend these big money owners, so they are unlikely to sell the best equipment to China.

6. Of course, it's our own, and it's the first in the world, surpassing Intel
the United States doesn't care much about this. Only Japan and other neighboring countries with developed electronics instry are particularly nervous
we can only say that supercomputing is mainly used in the field of national defense and aerospace, with high investment and poor results. Supercomputers and processor chips are still different, which must be clarified
it's still too far away from public life. Intel has no rival at all in civil use, and AMD is still living on the European Union's fine against Intel.

7. In April 2015, the US Department of Commerce decided to ban the sale of Zhiqiang Phi computing cards to four national supercomputer centers in China. At that time, all kinds of voices were heard all over the world. Even though Tianhe 2 won the fifth consecutive title in the 2015 international supercomputer conference, it has been criticized by some people with ulterior motives for using the CPU of Intel company in the United States. Many people claim that without the CPU of the United States, Tianhe 2 will be doomed
the reality is quite the opposite:
at the isc2015 press conference, the University of Defense Science and technology announced the upgrade plan of Tianhe 2a, which will use the matrix 2000 (gpdsp) independently developed by the University of Defense Science and technology to replace the Zhiqiang Phi computing card. Matrix 2000 adopts 40nm process, has 16 cores, 1g main frequency, 2.4tflops double precision floating point, and 200W power consumption
at the beginning of the new year, good news was revealed from articles published by the Ministry of science and technology& mdash; The domestic multi-core chip designed by Shanghai high performance integrated circuit design center has reached the advanced level in the world. The multi-core chip adopts 28nm process, and the peak double precision floating-point operation speed exceeds 3 trillion times per second (3tflops), which is completely equal to Intel's second-generation Xeon phi (also Intel's best multi-core chip) & mdash& mdash; The second generation of Zhiqiang Phi computing card, code named "Knights landing", adopts 14nm technology, with double precision floating-point performance of more than 3tflops and power consumption of 250-300W
on January 22, Xinhua News Agency reported a more exciting news & mdash& mdash; China plans to launch a new generation of 10 billion times supercomputer in Tianjin this year (10 billion times is 1000p, tianhe-2 is 55P, if it's not for the reporter's mistake, it's really black Technology)... At the same time, the University of Defense Science and technology is designing a new generation of Feiteng chip< It can be seen that the US Department of Commerce's ban on the sale of computing cards has not restricted the development of supercomputing in China. In fact, in the past half century, history has proved Comrade Mao Zedong's words& mdash;“ Let's blockade it for ten or eight years, and China's problems will be solved. ". Everything that can be bought is difficult to make.
although most computers in China have no Chinese chip since the 1980s, even the supercomputers Tianhe-1 and tianhe-2, which once made Chinese people proud, basically use foreign chips, while domestic chips are only used for high-speed Internet
however, in the era of Mao Zedong, China had its own semiconctor instry and mastered the whole process technology of single crystal preparation, equipment manufacturing and integrated circuit manufacturing. Computers made in China also had Chinese chips
however, in the 1980s, China's IC instry suffered a devastating blow after the implementation of "building is better than buying, buying is better than renting" and "market for technology"& mdash; The family wealth accumulated in the Mao Zedong era was completely destroyed. Technical talents either went abroad, or went to universities to teach, and some were even transferred to guard the computer room. The technical talents cultivated in the Mao Zedong era were completely lost
ring this period of time, independent technology has been infinitely devalued, while foreign technology has been over glorified, and "foreign experts" have been regarded as God, and policy makers lack the most basic vigilance against them& mdash; Many promising scientific research projects were rejected after leaders listened to the opinions of "foreign experts"; Many scientific research projects are led astray under the "careful guidance" of "foreign experts"; A large number of valuable independent technologies have been stolen by "foreign experts" and turned into "foreign technology", resulting in the dilemma that Chinese enterprises still have to pay patent fees to foreign investors when using their own technologies< For example, in 1983, Inspur began to use imported intel8088 chips instead of domestic chips to assemble 0520 microcomputers. The Great Wall 286, Great Wall 386, Great Wall 486, 2780, Taiji 2220, Yinhe super minicomputer, hn2730 super minicomputer and other computers developed from the 1980s to the early 1990s all use foreign chips. The representative model of supercomputer in the fourth generation computer in China & mdash& mdash;“ Yinhe-1 supercomputer costs 100 million yuan, but it has not played much positive role in the progress of China's semiconctor instry e to the purchase of a large number of foreign hardware.

9. That of course is the unremitting efforts of the R & D department
in addition to efforts, we also need to have a forward-looking vision
in order to become a cutting-edge team in the science and technology instry