You may have heard the term “supercomputer” before, but have you ever given any thought to what it takes for a computer to be called “super”? Supercomputers are more than just vague concepts found only in science fiction; there are actually hundreds of real supercomputers in the world constructed to assist in solving the most challenging scientific problems. This month, let’s explore the realm of supercomputing and see what these big ol’ machines can do.
What makes them so super?
A supercomputer is more than just a big computer; it’s actually lots of big computers in racks and cabinets all linked together with optic or copper cables. This “hivemind” design gives supercomputers unparalleled processing power to tackle the most complex mathematical calculations. So how do they compare to typical PCs?
We measure computer performance in floating-point operations per second (FLOPs) – that’s how many equations involving real numbers (particularly decimals) a processor can solve in one second. A typical PC has a single processor (CPU) and a single graphics card (GPU) which, combined, provide around 8 cores to process information. That allows for a performance of a few teraflops (one trillion FLOPs). And while that’s certainly a lot of calculations, the fastest supercomputer in the world has thousands of CPUs and GPUs with the equivalent of more than 8 million cores. That gives it a data processing speed of over 1 exaflop – that’s one quintillion calculations per second.
That’s a million trillions … every second.
There’s also a tremendous difference in the amount of RAM that a supercomputer requires. We’ve discussed before how having 16 to 32 MB of RAM in your PC is ideal to make it feel nice and fast, but that’s nothing compared to the Petabytes (PB) of RAM in a supercomputer (a Petabyte is a trillion Megabytes). It’s difficult to convey just how vast the gulf is between those values, but if you think of a Megabyte as a meter, then a Petabyte is the distance from the Earth to the Moon and back … 13 times.
How are they being used?
Different supercomputers are designed and run by different groups or companies, so their intended purposes vary.
The IBM supercomputer called Watson was initially designed as an AI “question answering” machine, and in 2011 it even made history on Jeopardy! when it defeated two human champion competitors live on national television.
The supercomputer called Sierra, located at the Lawrence Livermore National Laboratory in California, performs military research and nuclear weapons simulations.
Fugaku, located at the Riken Center for Computational Science in Japan, has helped develop Japanese language generative AI models and was also used for research on masks during the COVID-19 pandemic.
Frontier, run by the Oak Ridge National Laboratory in Tennessee and currently the world’s fastest supercomputer, was designed to assist with cancer research, drug discovery, and the future design of new medicine technologies.
Weather forecasting, data mining, complex decryption, climate modeling, pharmaceutical research, and big data analysis are just some of the challenges that supercomputers are used to tackle. In summary, supercomputers can provide predictions, perform analysis on large data sets, and chew through complex calculations that would otherwise be impossible, allowing us to further our knowledge in every field from biology to engineering to national security.