I Found A Planet

May 1st.

A piece of good news came from the Supercomputing Research Center established in Tianjin City.

After nearly a year, Xinghai Technology's first self-developed supercomputer, Supernova 1, was successfully developed.

Since then, Xinghai Technology has a self-developed supercomputing system, which can greatly save the purchase cost of supercomputers.

Regarding the performance of "Supernova 1", Zhao Zhanglai, the person in charge of the project, explained it to Chen Jin in a remote video conference.

Zhao Zhanglai is in his forties, wearing glasses and wearing a white coat. He was originally an associate professor expert working at the National Defense University and participated in the research and development of the "Tianhe-2" supercomputer. In this project, Zhao Zhanglai served as the "Second Architect ", the role is very important.

In order to lure him, Chen Jin not only offered him an annual salary of 10 million, Xinghai Technology also signed a "Strict Confidentiality Agreement" with the University of National Defense Science and Technology, and paid 100 million in technology licensing fees, and the other party agreed to release him. .

But in fact, the mechanism used by Supernova 1 was brought over by Chen Jin from the Dakeng base, which is obviously much more advanced than the structure of "Tianhe 2". Zhao Zhanglai was dug here only because he is more familiar with this area... …No matter how advanced the technology is, it must be realized by experienced people.

As for the University of National Defense Science and Technology, they have developed the "Tianhe-3E" with a computing speed of exascale. Special care.

On the contrary, the tens of millions of annual salary offered by Xinghai Science and Technology caused a sensation within the University of National Defense Science and Technology, and many leaders were annoyed: "It's really disturbing the morale of the army!"

...

Zhao Zhanglai introduced:

"Mr. Chen, the supernova architecture you provided us is at least two generations more advanced than 'Tianhe-2' and comparable to 'Tianhe-3E'. In theory, based on this architecture, we can develop an exascale supercomputer of exascale. "

"However, you said that the company was in short supply of supercomputers, and we were in a hurry to use them, so we made a small fuss and first made a small 'Supernova 1'."

"The peak computing speed of Supernova 1 is about 1.25 billion times per second, and the continuous performance is 1 billion times per second."

"The overall level is comparable to the Sunway TaihuLight before the upgrade, and its performance ranks third in the country."

"Because of the advanced architecture, the power consumption ratio of Supernova 1 is very outstanding. We achieved 9G/W, which surpasses the 6.5G/W of Sunway TaihuLight A, and is comparable to the power consumption ratio of Tianhe 3E... the world The performance is top-notch, almost reaching the limits of traditional computers."

"With such a power consumption ratio, hundreds of millions of electricity bills are saved every year."

Chen Jin nodded. The most expensive thing about a supercomputer is not the hardware construction cost, but the high electricity bill!

For example, "Tianhe-2" is a big power consumer, with a total power consumption of 17,808 kilowatts. After being equipped with a water-cooling system, the power consumption will reach 24 megawatts...equivalent to the power provided by a small nuclear reactor.

The electricity bill is as high as 100 million RMB per year.

The continuous computing performance of Supernova 1 is three times that of Tianhe-2, but Supernova 1 also saves energy by 30%, and the annual electricity bill is about 66 million.

It is more economical to use.

"In addition, the supercomputing architecture provided to me by Mr. Chen has unimaginable stability. The general supercomputing failure rate is quite high. Every time a computing core is broken, it takes half an hour to repair it. There are tens of thousands to hundreds of thousands of cores, and one or two will often fail... Such a super can not meet long-term use, and can only run some short-term computing projects."

"And many of our company's program applications require long-term operation and calculation...it is not allowed for supercomputers to fail frequently."

"Our supernova architecture adopts fault self-diagnosis and intelligent parallel technology. The failure of one or two computing cores will not affect normal use. It can be replaced directly when it is turned on...even if half of the cores fail, it can still be used normally."

"In terms of stability, the supernova structure is not weaker than the structure used by Tianhe-3E."

"And the supernova architecture is extremely scalable, and it is entirely possible to develop an E-class supercomputer based on this architecture."

"..."

Zhao Zhanglai said that he couldn't finish talking about the various advantages of the supernova architecture. In short, he made the right decision to switch to Xinghai Technology. Researching supercomputing here not only has sufficient budget, but also has a bright future.

He is very satisfied.

Chen Jin nodded repeatedly.

Regarding supercomputing, in many people's minds, it is as simple as stacking a large number of CPUs together, referred to as "core".

In reality, it is far from easy. To make tens of thousands or hundreds of thousands of computing cores work together to process one thing at the same time, the massive data generated can easily exceed the transmission bandwidth of communication lines, and exceed the read and write capacity of memory and hard disks. Capabilities... These bottlenecks can only be broken through with advanced architecture.

In the field of supercomputing, the supercomputers of country Z have repeatedly dominated the rankings, and there is a reason why they often rank first in the world. It is because our chips are not good. Since the 1980s or even earlier, scientific researchers in country Z can only develop horizontally CPU parallel computing, supplementing quality with quantity... This is the only gift that the older generation of scientific researchers left to the younger generations in the computer field in that era of backwardness everywhere.

This is why after foreign countries banned the sale of advanced CPUs, researchers in country Z were still able to create the first supercomputer by using domestic CPUs that were several generations behind.

As for the "Supernova 1" self-developed by Xinghai Technology, the CPU, memory, hard disk and other core components used are all domestically produced. In order to reduce the purchase price, Xinghai Technology ordered enough to produce 15 "Supernova 1" at one time "The huge order stunned those manufacturers.

As a result, the cost of "Supernova 1" has been successfully reduced. Excluding R\u0026D expenditures, and based on hardware costs alone, the cost of Supernova 1 is about 500 million.

"It really is better to buy than to make."

Chen Jin sighed, and said to Zhao Zhang, "Assemble five units of Supernova 1 first, and the company will use them! After completing this task, you will immediately invest in the research and development of exascale supercomputers, and develop a computer with a computing capacity of tens of billions of billions." The next 'Supernova 2'."

"Understood, Mr. Chen."

Zhao Zhanglai nodded.

...

A few more days passed.

A not so optimistic news came.

Ma Xiaobing, the person in charge of the "Nebula System" project, said: "Mr. Chen, the number of programmers using 'Nebula Studio' to try to develop Nebula APP has dropped from a peak of 1 million to less than 100,000. fall."

"Because Gu Ge forced an 'anti-flash patch', any mobile phone with this patch, as long as it is flashed into the Nebula system, the underlying hardware of the mobile phone will be damaged and cannot be used... At present, 90% of the mobile phones in China Android phones have this patch... This is simply a rogue push! Virus push! The purpose is to block our Nebula system!"

This directly confirms the rumor - the rumor that there are backdoors in foreign software or operating systems is indeed true.

There are even backdoors not only in the software, but also in the underlying hardware... Otherwise, how can changes at the software level damage the hardware?

Therefore, after the "backdoor incident" was hammered, on the one hand, the information security department of country Z is investigating, and will soon give Google a ticket.

On the other hand, the majority of programmers who originally wanted to participate in the development of Nebula APP were greatly shaken in their confidence and chose to give up one after another.

Because things have developed to this point, it is no longer just a software-level problem. No matter how good the Nebula system is, it is useless. As long as the back door of the underlying hardware is still there, the Nebula system may still crash.

Unless Xinghai Technology Company has a way to make all hardware and software independent, it is impossible to build the Xingyun system.

A senior programmer lamented: "The death of the Nebula system will sound a wake-up call for us. Unless the software and hardware are fully autonomous, the back door will last forever."

But is the nebula system really going to die?

In this severe challenge, Xinghai Technology is helpless and helpless?