What is a supercomputer? A super-quick overview in 500 words

What is a supercomputer?

First, it’s not AI. Artificial intelligence is software: a supercomputer is hardware.

Nor is a supercomputer something you’d necessarily recognise as a computer. (I.e., a portable machine sitting on a desk.)

Rather, a supercomputer is a distinct class of high-powered computer designed to crunch complex computations. It’s super-fast, super-sized, and works entirely differently to standard computers.

Supercomputers vs standard computers

At the simplest level, a computer takes input, stores and processes it, and generates an output.

A standard computer does so by performing one step at a time, following a distinct series of operations. We call this serial processing.

A supercomputer, meanwhile, can perform many actions at once using parallel processing. Rather than tackling every part of a task one step at a time, it splits the problem into pieces across multiple processors.

A processing analogy

We’ll use a supermarket self-service station to compare these two processing techniques.

  • Standard computer – serial processing

You take your item from the basket, you scan it, your add it to the weighing area, you repeat. Even if you work as fast as humanly possible, you can still only scan the bag a single item at a time.

  • Supercomputer – parallel processing

This time, you split your basket across a team of helpers. Each helper goes to their own self-service station with their own allocated item. You can all leave within moments, items scanned simultaneously.

Practically speaking

Back on a practical level, what this means is that supercomputers operate with vastly superior computational capacity than general-purpose computers.

They have multiple central processing units, massive storage capacity, and exceptional input/output capability.

They also tend to be super in terms of size. For example, a supercomputer may fill a server room the size of two tennis courts.

They also use a different performance metric than standard computers. (Using floating-point operations per second, or FLOPS, as opposed to million instructions per second, or MIPS – but we won’t dive into this at a layman’s level.)

What is a supercomputer used for?

So, what is all this good for? Namely, for powering complex applications that handle vast databases and perform a vast number of computations per second.

Sometimes, this might entail one specific purpose only. Think a supercomputer built only to play chess, for example. For each move, the supercomputer would need to evaluate millions – or perhaps billions – of potential options in a rapid space of time.

IBM Watson: a supercomputer system initially developed to answer questions on the quiz show Jeopardy!

Elsewhere, supercomputers are used for mathematically intensive scientific or engineering problems. This often includes things like weather forecasting, or climate simulations, or testing models for new chemical compounds.

In each scenario, the supercomputer must run an enormous amount of calculations at a time.

A high-level overview

What is a supercomputer? You could read a number of advanced explanations if you want more detail. But, to sum it up neatly:

A supercomputer is a powerful computation engine designed for advanced number crunching. It performs at or near the (currently) highest operational rate for computers, and it does so using parallel processing.

More quick tech overviews