Black technology starts from the steel suit

Chapter 182 Chapter 182 The Bottleneck of Computing Power
Unit 1 is already in place

Unit 2 is already in place

Unit 3 is already in place

……

This time the experiment once again lasted for more than 28 hours, but unfortunately the results were not very satisfactory. Looking at the report submitted by the engineer, Mr. Fu had a headache. Looking at the letters on the A4 papers one after another, He didn't know what to say.

"Professor Wang, your model... is really complicated, can't you simplify it?" If it wasn't for the fact that the big man in front of him is too high-status and has a lot of money, they really don't want to serve this guy any more. A big shot.

You must know that supercomputing is a very scarce resource. If their supercomputing center wants to apply for use, it must apply at least one month in advance, and it may not be their turn. The preciousness of supercomputing resources can be seen It's clear.

However, it’s okay if you don’t want to serve them. After all, the supercomputing center is a national resource, and they are just managing it on behalf of them. If there are really important tasks, then the task must be the priority, and it is impossible to violate this original intention.

In addition, Professor Wang is really rich, so he paid all the liquidated damages for one month, so they can't say anything. After all, the boss is so willful, what else can they do?
"Let me try again!" Wang Feng looked at the experiment report, and his brows were also frowned. He had spent a lot of effort to optimize the model, but unfortunately the simulation was still not accurate enough, and the experiment report did not have a strong persuasive.

After so many simulations, Wang Feng also gradually discovered the defects of general-purpose computers in molecular dynamics simulations. Unless his model has achieved a major breakthrough, otherwise, the simulation efficiency of general-purpose supercomputers may not be enough. way to meet his requirements.

"From this point of view, we can only think of a way here!" Mr. Fu took the report in his hand and began to think about the possibility of success. In fact, what he wanted to say was whether to consider finding cooperation with foreigners. After all, general-purpose computers are not very good at this aspect of work.

But he couldn't say that. After all, the country spent so much money to build a supercomputing center, but in the end you said it couldn't be used, and you still need to find foreigners. Then...

Wang Feng looked at the way the other party hesitated to speak, and probably understood that this has basically reached the limit, and there should be no better way.

"It's difficult. Do I have to build a supercomputer specifically for molecular dynamics?" Wang Feng looked at the flashing lights of Shenwei. He had been at the Wuxi Supercomputer Center for half a month. , but no effective progress has been made so far.

At this time, he felt that he might have to think about David Shaw's suggestion with a high probability. After all, from the current point of view, this guy's Anton is indeed a scorpion shit—the only one.

To be honest, this is the first time that Wang Feng encountered a bottleneck in computing power.

Most of the difficulties he encountered before were at the logical level. Even if he occasionally encounters more complicated calculation problems, he can quickly find the results with the help of a computer.

But now, the situation he encountered was completely different from before.The mathematical model established for a single problem can easily complete the calculation work with a computer with a decent CPU and software such as Molpro, so his research and development work can be carried out very well with the help of this imperfect theory. smoothly.

However, when a "single problem" is extended to a "certain type of problem", the amount of calculation involved will increase geometrically along with the "unknowns" that appear in the theoretical model. A simple explanation Think about it: It can be considered as simply calculating the behavioral characteristics and patterns of a certain individual to calculating the behavioral characteristics and patterns of a certain type of people or people in a region.

The amount of calculation is so large that even the top five supercomputers in the world, such as Sunway TaihuLight, can hardly bear it, which is a bit too outrageous!

Of course, this does not mean that the supercomputing ability is not enough. The Sunway TaihuLight supercomputer is a supercomputer developed by the National Parallel Computer Engineering Technology Research Center and installed in the National Supercomputing Wuxi Center.

It installed 40960 "Shenwei 26010" many-core processors independently developed by China. The many-core processors use a 64-bit independent Shenwei instruction system, with a peak performance of 12.5 billion billion operations per second and a continuous performance of 9.3 million million operations. /Second.

In November 2020, in the 11th global supercomputing TOP56 list, Sunway TaihuLight ranked fourth. It can be said that there are very few computers in the world that can compete with him. All but a few computers.

A scientific research team led by Tsinghua University has for the first time achieved a global 10-kilometer high-resolution numerical simulation of the earth system at a million-core scale, which will comprehensively improve China's disaster reduction and prevention capabilities in dealing with extreme climate and natural disasters;
The numerical simulation of the return path of "Tiangong-[-]" by the National Computational Fluid Dynamics Laboratory will provide accurate predictions for the smooth return of "Tiangong-[-]";
The drug screening and disease mechanism research carried out by Shanghai Institute of Materia Medica has greatly accelerated the progress of drug design in the fields of leukemia, cancer, and bird flu.

It does look very powerful, and its ability to simulate in the virtual world is indeed unique, but it is useless. He has a specialization in surgery, so he is not used to do this job, so there is nowhere to use his power, and he can't show it.

But Anton is different, he is specially designed to do this.Judging from the floating-point computing power of the evaluation index of general-purpose computers, it will be beaten down; but once it reaches its field, it is a god-like existence and is known as the nuclear weapon of computational chemistry!

Thinking of this, Wang Feng took out the letter again:

Dear Professor Wang, hello!
My name is David Xiao, maybe you haven’t heard of me, but it doesn’t matter, as one of the authoritative scholars in the field of computational chemistry, I assure you that you must know Anton, which was created by our team.People in the same industry call it a nuclear weapon in the field of computational chemistry. I agree with this very much and feel very honored!

During this period of time, I have seen your latest achievements, and I am very interested in your research, so I sincerely hope to join your research, at least let our team join your team, and contribute to your research. Research and make a contribution to the progress of all mankind.

I don’t need your investment, although you are indeed not short of money; I don’t even need to sign the paper, the results are all yours; I am willing to provide you with technical support and help you build your own supercomputing center. Experts in the field of computational chemistry must need a supercomputer specifically designed for computational chemistry, and I can provide you with a full set of technical support for free in this regard, and even provide ASIC chips.

My only request is that you add the name of our DEShaw Institute to the latest results. This is my only request, and I hope you can accept it.

To be honest, any scholar would be ecstatic with excitement after hearing this simple request.

After all, supercars are very expensive, even if they are rented, they are not cheap. In this short period of half a month, Wang Feng did not count the liquidated damages paid to others. It cost more than 3000 million yuan!

This is after he comes down with the documents. Otherwise, even if he is willing to spend so much money on such a troublesome thing, people probably won’t be willing to do it for him: the money he makes is not his own, but these troublesome things are needed. Solved it by myself, and it is very likely that I will have to work overtime!

And David's promise is equivalent to millions or even tens of millions of funds, and it also runs the risk of equipment damage.In contrast, his request is only to add the name of the DEShaw Institute to the position of the thesis research institution.

What an internationalist request this is!

Of course, as a person in charge who is involved in many confidential projects and attaches great importance to confidentiality work, this is not good news. It means that his research project may be leaked!

This is a very serious matter, and it can even be called an accident, because since this kind of leak can happen in a non-secret project, it may appear in a confidential project, and this matter must be strictly investigated.

However, Wang Feng speculates that it is more likely that the other party came out of it. After all, he has already laid the groundwork many times when he published the paper. It is not surprising that those scholars who are still fighting on the front line can guess. It's weird not even guessing!

"Do we still need to continue the experiment?" Mr. Fu couldn't help but ask as he watched Wang Feng fall into deep thought.

"No more, everyone has been fighting for half a month. Even hard-working people should not be able to take it anymore. Let's take a rest first." Wang Feng shook his head and refused: after all, the research institute has been working hard for half a month. The engineers working in almost three shifts (two days), both physically and mentally have reached the limit.

"Except for the staff on duty, everyone should take a good rest first. By the way, there should be a lot of my advance payment left. Let's go to the account to withdraw a subsidy first. According to 3000 for each ordinary researcher or engineer, advanced A bonus of 5000 will be issued to researchers and engineers, and everyone should take a good rest for two days."

"This..." Mr. Fu was a little embarrassed: "This is not good!"

staff member:? ? ?
I don't think I have a problem, but you have a problem!
"It doesn't matter, just go to Huaneng Battery's account, and everyone will directly bring the invoice back for reimbursement, and the reimbursement amount will be based on this ratio!" Wang Feng probably figured out what the other party was worried about, so he changed the term.

"Okay, then I will let everyone go to rest first, and we will meet again next Monday!" Mr. Fu asked his assistant to announce: Today is Friday and this is the morning. If you go to work next Monday, you can move out A 3-day short vacation is enough for everyone to relax.

"Also, Professor Wang, what do you plan to do?" Mr. Fu asked curiously. His eyesight was naturally extremely precise. Wang Feng not only gave everyone a holiday from frowning, but also gave them a subsidy. , which shows that the other party has found a way to solve this problem with a high probability.

"You can take a look at this first!" Wang Feng handed David Xiao's letter to the other party: "Shenwei's computing power has been squeezed to the limit, even if we continue to optimize the model and algorithm, I am afraid there is no way to get this thing A fundamental solution."

"In comparison, maybe Anton can help me verify this model!" Wang Feng didn't make any euphemistic remarks. After all, everyone is engaged in academics, and there is no leader present, so everyone can speak freely.

"Sorry, I couldn't help you, Professor Wang!"

"It doesn't matter, you have tried your best, but we are indeed missing some classes, and now we have to pay for the previous bills!" Wang Feng comforted.

"Everything will get better slowly, don't be discouraged, people run ahead of us, it's normal to be ahead!"

"I failed everyone's expectations!" Mr. Fu couldn't let go, and was even a little ashamed. Although Shenwei once won No.1, and now it has not been squeezed out of the top five, but that doesn't mean that Shenwei is so capable. , but other countries are too incapable of fighting.

Needless to say, other countries either have no money or no demand in this regard, because they have completely given up, and even their own mushroom eggs in Europe are handed over to the United States for safekeeping. Normal.

However, whether it is the United States or Japan and other countries that are the most capable, there are quite large-scale supercomputers that continue to be used, and their performance is not outdated. It would be a waste to eliminate them now, so they just put the top few It was snatched back, and there was not much action after that.

Shenwei's floating-point computing ability is indeed outstanding, but it also has many problems, especially the chip problem. This is a flaw. No matter how good the algorithm and optimization are, the gap in this part cannot be smoothed out. .

Sunway TaihuLight is the first top-level supercomputer built under embargo conditions in China. The word Sunway comes from the Sunway SW26010 processor it uses. It was developed by the Sixth Research Institute of the Ministry of Network Systems (also known as Jiangnan Computing Technology Research Institute). ) developed.

SW26010 is the fourth generation CPU of Sunway series, which should be made using 28 nm process (there is not much information online, so leave a question mark), which is indeed one or two generations behind Intel's 14 nm.Therefore, the main frequency of SW26010 is indeed not high, only 1.45 GHz, while the main frequency of Intel Xeon series chips during the same period is usually 22~2.6 GHz.

Whether it is playing games, doing simulation calculations or doing video editing, everyone should know how much the main frequency of the CPU affects performance, and this part of the impact cannot be eliminated through software optimization.

Therefore, choosing Anton can be said to be an inevitable choice, unless Wang Feng wants to wait for them to develop this kind of specialized supercomputer, or wait for quantum computers to enter practical applications before publishing a paper.

"Perhaps, this is an opportunity!" Wang Feng pondered.

"Oh, you mean?"

"I do need a supercomputer, a supercomputer specifically designed for computational chemistry, and you probably also need some experience in this area, what do you think?"

"Will they agree?"

"That depends on the content of our transaction!"

(End of this chapter)