Half a century back, even the smallest computer would have taken up an entire room. And today, thanks to the microchip, more powerful  computers than ever have been put in the pocket. But have you ever thought, if a house-sized computer is made with these small chips, how powerful it will be? Is it possible to make a super computer like this ? — that has a million times more supercomputing power than your or my laptop or desktop, and can solve any of the world’s most complex scientific problems? Today’s question is, what exactly makes an ordinary computer a super-computer? How is a supercomputer different from your current computer? Let’s know all the details about super computing.

What is a supercomputer?

So, what super features make a computer called a super-computer? Due to the computer flying? 😛 Because of the laser light from the CPU of the computer? 😛 No because of the buildin ammo and cannon with the computer? 😀 Hey one minute! It’s not super man, super-computer!

In simple words, computers that can solve complex scientific problems very fast, can predict the weather or can tell now what the climate will be like in 2050 and the computers that can do such amazing things are called supercomputers. Now you may ask, exactly how fast the computer works? If I add more RAM and more core processors, more GPUs to my computer, can I do super computing? In fact, in order to fully understand the definition I have given of a supercomputer in “simple language”, you need to be familiar with some terms.

Look, before entering today’s main question, we need to know what is a computer? —It is basically a machine, which completes any simple task. It first takes the input information (data) to do some work, then processes it in its own power, and finally presents you an output result. In fact, super-computer means that it has to be a huge giant computer or a million-billion times faster than your normal computer, but not at all; In fact it uses a completely different method to complete a task. Like your normal desktop or laptop, it multitasks rather than doing just one thing at a time. And this is basically the main feature that makes a computer a supercomputer.

Serial and Parallel Processing

What is the difference between serial and parallel processing? A normal computer can perform only one task at a time, i.e. to complete a task, it completes the processing one after the other and provides the output of the task. And this is called serial processing . Now you might say, “hey I play music, browse the internet and render video at the same time with my computer, and you say one thing at a time”? In fact, your computer can complete millions of commands per second, making any task seem real time to you, serially, one after the other. Let’s try to make the matter more clear with a nice example. Imagine an ordinary shopkeeper, suppose you go to his shop and ask him to give you a 1kg packet of flour. Now what he will do—he will first get up from his seat, then go to the flour packet rack, bring the flour packet, hand it to you, and count your cash and put it in the drawer. Now look, it doesn’t matter how fast he got up from his seat or how fast he got the dough from the rack and handed it to you. The point is that he is doing only one thing at a time. No matter how fast he is serving you, he is only doing one job at a time and serving you by doing one job at a time. This is how normal  computers complete their work.

But today’s modern supercomputers work using a completely different method. It divides a job into different parts and processes them simultaneously, and this system of doing work is  called parallel processing . As per the above example, now imagine that you go to a store to buy 1 kg of flour, 1 liter of oil, and 1 kg of sugar. Now suppose there are 3 people together in that shop to help the shopkeeper. Then everyone can choose a different job to deliver the things you want to you. Someone will bring sugar, someone will bring oil and someone will deliver flour and in this same time you can complete the money transaction with the shopkeeper. Notice again, each brings different things at the same time, but all come to you together. Thus, no matter how many things you want to buy in the shop, if there are many people working there, it is possible to complete all the work very quickly in the same time. And this is parallel processing, theoretically, our brain also works in parallel processes. At the same time you are reading this article, at the same time your brain is doing full activity inside your body.

Why is parallel processing necessary in supercomputing?

The computers we use every day for work do not need to be so efficient. Browsing the Internet, sending email, or typing requires very little of your  computer’s processing power. But if you want to do more complex tasks, such as changing the color of a very high resolution digital photo, your computer will need to do a lot of processing, sometimes taking several minutes to complete. Again, if you want to do gaming on your computer in addition to just browsing the Internet, sending emails, or typing, you will need a fast processor, dedicated GPU, and more memory (RAM), or your system will work slowly. Faster processors, and double the memory will make your computer run dramatically faster. But how fast? This speed also has a limit—because a processor can only do one task at a time.

Now imagine you are a scientist and you work in a meteorological office. Together you will work on weather forecasting, new cancer drug tests, and climate modeling up to 2050. Now you can upgrade your PC’s processor and more memory to complete these tasks, although it will work faster than before but still there is a limit to its working. But if your tasks are divided into one piece and each piece is handled by a separate processor, that means parallel processing, then your tasks will be done much faster and efficiently than before.

Parallel computing

Building supercomputers Supercomputing requires multiple processors—so that they can work together in parallel to complete multiple tasks at the same time. The name of the current fastest supercomputer is The Sunway TaihuLight  (The Sunway TaihuLight). It has 64-bit 40,960 multicore processors and each processor chip has 256 processing cores. That is, this entire computer has 10,649,600 processors!!!

Parallel computing is certainly an ideal system for supercomputing, but parallel computing also has some significant problems. Let’s go back to that store example, but this time let’s take the example of a supermarket. Suppose you go to the supermarket with some of your friends to buy some things. Now if you send each of your friends to a different counter to get some things, then in a very short time your things will be bought, and at the end of shopping all things will be together with you. In the same way, a supercomputer  divides a task into pieces and completes it with different processors and then puts the pieces together to give the output. But if a task is very large, then sending that task to the processors in pieces and putting all the pieces together at the end of the work becomes a very difficult task for the computer. To do this, another management system is needed so that each piece of work is delivered to the processor and the returned pieces are assembled for output. But processing this management system also requires a lot of processing power, which can overload the system. Such as Windows OS’s Task Manager—it basically displays the processing of all programs, but even displaying them requires some processing power.

By sending your friends to different counters, you may be able to complete the shopping in much less time, but the payment time will be delayed. Imagine you are shopping for a million items together and 50,000 friends are collecting items from different counters. If you shopped from one counter at a time then the cash pay problem would have been solved at one counter, but shopping at different counters has split one problem into 50,000 problems.

Similarly, imagine you command a computer to forecast the weather for the entire world for the next week. Now your computer will send all this work piece by piece to different processors for processing, if you were to do the work in one processor then all the processing would be done at once. But for processing in different processors, the weather forecast of different countries will be processed repeatedly in each processor separately. But the problem can be more serious, because the climate in different countries of the world can be completely different from another country. As a result, one processor will have to wait for another processor to finish its work—and one will need help from another processor to solve a problem. So some tasks in supercomputers need to be done in parallel and some tasks in serial processing.

Super computing

Types—

By installing thousands of processors, RAM and GPUs in a huge box, it is possible to make a supercomputer and do many complex tasks by doing parallel processing. Or if you want, you can build a supercomputer by interconnecting many PCs together in the same room with many fast LAN (local area network) cables. Such a supercomputer is called a cluster .  Google uses this type of computer to provide search results to its users.

Another type of supercomputer  is called a grid  —which is similar to a cluster, but different computers are located in different parts of the world and are connected to each other using the Internet or other computer networks. This is an example of distributed computing, where computers in different parts of the world work together in a network. It can also be called virtual super computing. Different universities or research centers around the world connect their computers to each other to create grid supercomputers. Not every computer on the grid works at the same time, nor do they have to work, but together they form a powerful computing system.

Administration—

Cluster supercomputers are placed together over a large area, sometimes the size of 3-4 football fields. And to run these giant sized  computers , a huge amount of electricity is required, enough electricity to power thousands of homes. Millions of dollars are needed to run supercomputers that consume so much electricity.

And supercomputers generate a lot of heat, because when electricity flows through a cable, it also generates heat energy. And for this your normal computer also needs a cooling system. But it is not possible to cool these huge computers with ordinary fans. Today’s modern supercomputers are cooled by liquid, similar to how a refrigerator works. It consumes a lot of electricity to operate this cooling system, secondly it is harmful to the environment.

What software does the super computer run on?

As mentioned earlier, a computer is a machine that can do all the normal tasks (virtually of course 😀). If you want, you can run the supercomputer using your computer’s operating system, such as Windows! But most supercomputers run on Linux-based operating systems. These operating systems are highly customized, meaning they are built with only specific working features. Since super-computers are mainly used to solve various scientific problems, why should there be other working features?

What is the use of super computing?

Since the computer is a machine for all general tasks, you can do Internet browsing, send mail, photo editing, video editing, gaming and even read articles on Techhubs with a supercomputer. Just like your normal computers, you will need to run various programs on your system to do these tasks. Just as you need different apps (applications) to do different things on your Android or iPhone, these are computer programs, but they have different names!

Supercomputers do not do these things for now. It is used for solving various complex mathematical problems, solving scientific problems, nuclear missile tests, weather forecasting, climate change, and testing the accuracy of encryption. But in theory, supercomputers can do all the things that your normal computer does.

How powerful are supercomputers?

Common computers express their performance in terms of MIPS or Million Instructions Per Second. This indicates how many commands (read, write, data store) the system can process to complete a task. MIPS makes it easy to compare two processors, the more MIPS the processor can handle, the more powerful it is. This is commonly known as processor speed, which is usually expressed in gigahertz.

But the performance of supercomputers is expressed differently. Since these computers are used for scientific purposes, its capacity is calculated as FLOPS ( Floating Point Operations Per Second). It is an efficient way to calculate the working capacity of a  computer . Let’s try to understand these two terms with a small example…

Imagine you’re watching a soccer match, you see player #1 currently in possession of the ball pass the ball to player #2 in the middle of the field, player #2 pass the ball to player #3 near the goal—and player #3 makes a smart hard hit. He scored by putting the ball into the goal. And in this way, 3 more goals were scored through the hits of different players. Now, MIPS here means how many hits the ball has to enter the goal to complete a goal. And FLOPS means the total number of goals scored by that team. That is, how many million commands the processor has to follow to perform a task per second is expressed by MIPS and how many tasks are completed per second is expressed by FLOPS.

5 fastest supercomputers in the world