How to Build a Supercomputer with LEGO and Raspberry Pi

Raspberry PiBuilding a multi-node supercomputer is a dream shared by geeks around the world. Imagine having your own supercomputer with which you can solve the mysteries of the cosmos and put to work calculating Pi to the next decimal. This is possible with with the Raspberry Pi, a tiny computer you can purchase for $25-35 that takes very little power to operate.

At the top of my Reddit list yesterday was an interesting story that actually broke back in September. Since then, some additional information has come out, including a revision to the project’s instructions.

Back in September, the University of Southampton assembled a 64-node supercomputer using LEGO and 64 different Raspberry Pi devices. In November, a step-by-step guide was updated and published on the project’s website so you can build your own.

The Raspberry Pi Supercomputer, a project led by Professor Simon Cox, has been called “Iridis-Pi” after its larger supercomputer and runs off a single 13 amp power source and has 1 TB of memory, 16 processors, and takes up about as much space as a traditional server tower. As far as actual computing potential, the latest word from the project is that it quickly calculated Pi, a common test for a new supercomputer.

The linking of the nodes is made possible via ethernet and MPI (Message Passing Interface). Each individual Raspberry Pi has a 16 GB SD card installed. A base installation of Debian Wheezy got the ball rolling with additional configuration noted in the complete how-to released by the university.

For under $4,000, you could have a supercomputer in your very home.

Why Does This Matter?

Developing nations have a hard time obtaining and supporting advanced computing hardware. Learning institutions need tools like powerful computers in order for students to test theories and make calculations that a standard laptop or personal computer couldn’t.

While the Raspberry Pi isn’t exactly a powerful computer, a 64-node (or larger) supercomputer of any kind could provide a useful tool for learning institutions and companies operating on a very tight budget. Many of the cures and scientific breakthroughs made in the past century are attributed to computational processes made possible by supercomputers.

Now if only someone would write a Folding@Home program optimized for 64-node Raspberry Pi supercomputers.

Image: University of Southampton

Leave a reply:

Your email address will not be published.

Site Footer

Ryan Matthew Pierson