Interesting use of Beagleboards.

So, this guy a while back made a kickass graphing calculator out of a Beagleboard and a nice sized monochrome screen. I guess it wasn't quite powerful enough for his liking, so he made this Beagleboard cluster-in-a-briefcase running (iirc) the same software, and it apparently outperforms his laptop.

I'm not so sure this is a useful cost-benefit balance given that the whole setup cost ~$2,000USD, but I thought it was cool. :P I do wish he had shown the power usage while it was actually doing something though...and actually done a proper job making a case for it..

http://hackaday.com/2010/09/07/beagleboard-cluster/


...I have to point this out because I'm picky like that, but in the article he mentions that this thing outperforms his "$4,500 laptop"...that is not a $4,500 laptop... The laptop in question is a Lenovo Thinkpad X201, which even if you max it out with an i7-860, 4GB of RAM, and a 128GB SSD, 'only' comes to a tad over $2,500. ...anyways, enough with me picking apart his claims, just go read the article. :P

what's a beagleboard :?

It's a OMAP3530 (ARM Cortex 8 ) embedded system.
http://beagleboard.org/

...I just had a ridiculous idea...
What about a bifferboard cluster? it would be completely useless though, since it's just a normal 150MHz x86 chip, but it would be fun! :P It would be simpler than the beagleboard one too, since the bifferboards have NICs integrated.

I'm confused. It's bypassing the CPU and running through the cluster? What can you actually do with it?

Each beagleboard is a small ARM-based computer. They're hooked up in a Beowulf cluster, and the software is smart enough to distribute the processes across the different nodes. The ARM architecture excels at doing certain types of functions, in this case certain mathematical processes, and does so at incredibly low power consumption levels. I think each beagleboard tops out at 5W peak power consumption.

Neat concept but if I can't game on it, screw it. lol

You could game with a cluster of ARMs but I doubt it would perform as well as a decent PC. You could build things like RAM controllers, drive controllers, and GPUs into a beagle cluster ... and you'd get something like a PlayStation.

The cluster of 8 beagles does better (than the dual-core laptop) on his "semantic analysis algorithm" ... basically a parallel-processing number-crunching job. It's hooked up to the laptop so it can share storage and log hardware monitoring.

The laptop would likely score a lot higher (do the job a lot faster) if this parallel-processing benchmark was running in the GPU. Note that it's already running Windows and a number of open processes while the beagles aren't.