- Supercomputers leverage much of the same technology as used in traditional computers
- Parallel computing allows supercomputers to process tasks faster than your PC
- Researchers share time on world’s biggest computers to solve science’s biggest problems
We’ve come a long way since MITS developed the first personal computer in 1974, which was sold as a kit that required the customer to assemble the machine themselves. Jump ahead to 2018, and around 77% of Americans currently own a smartphone, and nearly half of the global population uses the internet.
The devices we keep at home and in our pockets are pretty advanced compared to the technology of the past, but they can’t hold a candle to the raw power of a supercomputer.
The capabilities of the HPC machines we talk about so often here at Science Node can be hard to conceptualize. That’s why we’re going to lay it all out for you and explain how supercomputers differ from the laptop on your desk, and just what it is these machines need all that extra performance for.
The need for speed
Computer performance is measured in FLOPS, which stands for floating-point operations per second. The more FLOPS a computer can process, the more powerful it is.
For example, look to the Intel Core i9 Extreme Edition processor designed for desktop computers. It has 18 cores, or processing units that take in tasks and complete them based on received instructions.
This single chip is capable of one trillion floating point operations per second (i.e., 1 teraFLOP)—as fast as a supercomputer from 1998. You don’t need that kind of performance to check email and surf the web, but it’s great for hardcore gamers, livestreaming, and virtual reality.
Modern supercomputers use similar chips, memory, and storage as personal computers, but instead of a few processors they have tens of thousands. What distinguishes supercomputers is scale.
China’s Sunway TaihuLight, which is currently the fastest supercomputer in the world, boasts 10,648,600 cores with a maximum performance of more than 93,014.6 teraFLOPS.
Theoretically, the Sunway TaihuLight is capable of reaching 125,436 teraFLOPS of performance—more than 125 thousand times faster than the Intel Core i9 Extreme Edition processor. And it 'only' cost around ¥1.8 billion ($270 million), compared to the Intel chip’s price tag of $1,999.
Don’t forget memory
A computer’s memory holds information while the processor is working on it. When you’re playing Fortnite, your computer’s random-access memory (RAM) stores and updates the speed and direction in which you’re running.
Most people will get by fine with 8 to 16 GB of RAM. Hardcore gamers generally find that 32GB of RAM is enough, but computer aficionados that run virtual machines and perform other high-end computing tasks at home or at work will sometimes build machines with 64GB or more of RAM.
The Sunway TaihuLight once again squashes the competition with around 1,310,600 GB of memory to work with. This means the machine can hold and process an enormous amount of data at the same time, which allows for large-scale simulations of complex events, such as the devastating 1976 earthquake in Tangshan.
Another advantage of supercomputers is their ability to excel at parallel computing, which is when two or more processors run simultaneously and divide the workload of a task, reducing the time it takes to complete.
Personal computers have limited parallel ability. But since the 1990s, most supercomputers have used massively parallel processing, in which thousands of processors attack a problem simultaneously. In theory this is great, but there can be problems.
Someone (or something) has to decide how the task will be broken up and shared among the processors. But some complex problems don’t divide easily. One task may be processed quickly, but then must wait on a task that’s processed more slowly. The practical, rather than theoretical, speeds of supercomputers depend on this kind of task management.
Super powers for super projects
The truth is, many supercomputers are shared resources, processing data and solving equations for multiple teams of researchers at the same time. It's rare for a scientist to use a supercomputer's entire capacity just for one project.
So while a top-of-the-line machine like the Sunway TaihuLight leaves your laptop in the dirt, take heart that personal computers are getting faster all the time. But then, so are supercomputers. With each step forward in speed and performance, HPC technology helps us unlock the mysteries of the universe around us.