When does a really fast computer become super?
In: 5
What is super about a super computer is how quickly they can do calculations. FLOPS (floating point operations per second) is a commonly used measure of computing performance – it’s a measure of how many mathematical operations a computer can do every second. Current supercomputers can perform over 10^17 FLOPS, compared to desktop computers which are in the 10^13 range. So a supercomputer can do like 10,000 times as many calculations every second, as compared to a desktop computer.
It’s not about speed. It’s about amount of computation it can do in a period of time. Those two are different things.
Speed matters for the average consumer device because you typically only have 1 chip for a CPU with a certain number of cores. A modern supercomputer is more like a collection of a lot of computers. We’re talking about hundreds if not thousands of CPUs wired together to crunch numbers. At those scales, things like power efficiency matter more than speed. There are a lot of supercomputers built on essentially slightly customized x86 or ARM CPUs.
There’s no hard limit on what defines a supercomputer. Because it’s constantly increasing as technology advances. More than anything it comes down to how it’s used. They’re more often than not purpose built to do a particular set of tasks. Like doing trillions of floating point calculations (like multiplying decimals) for things like protein folding. Projects like Folding@Home were essentially a distributed supercomputer built on the internet using donated CPU time.
Kinda the short answer is that if you build a computer out of lots of computers, you’ve made a super computer. They call them a cluster, sure. Cluster is more a specific term.
Supercomputer is kinda an old term. Today you have cloud computing which you can rent like a hundred servers & make your own on-demand super computer. Though we’d call that a cluster.
Supercomputer sorta came about when the idea was ‘well, we cant make things efficient & small. Good thing is we dont need to today. Lets just put as much computing power together as we can just to solve this problem’.
Latest Answers