Liquidware announces modular, scalable Arm-based motherboards
If your idea of fun is developing software for obscure computing platforms, then you're probably already down with Liquidware. It seems that Arduino-based projects like the GamePack and OpenBerry have become a little too easy for the gang, so they've decided to embark on a little project to revolutionize computers by "turning them inside out," so to speak. The Illuminato X Machina is a 4-inch square upon which sits a 72MHz ARM-based microprocessor, a dedicated EEPROM chip for data storage, RAM, LEDs for output, and 14 I/O pins that line each of its four sides -- in short, a fully functional computer. The true magic of the operation, however, lies in the fact that these things are designed to cluster together, with each cell operating in parallel. "Like living organisms," it says at the group's website, the cells are "social." They autonomously communicate, distribute processing power and programming, and even isolate faulty neighbors -- when one of the cells crashes, it won't bring the whole cluster down with it. Believe us, we've only scratched the surface here -- hit that read link order up some of these bad boys for yourself, but not before you check out the videos after the break.
[Via Wired, Thanks Justin]
[Via Wired, Thanks Justin]
See more video at our hub!


















it's like a Mesh network topology!
That guy looks like Moses...
or Santa Claus considering that "oh-oh" thing he says
Sorta, there's already a lot of research in this direction - trying to make the hardware of a computer itself highly parallel. The difficulty of this is not linking the processors together, but implementing the machine language (and compilers that create this) that allows for highly parallel operation.
The theory behind this:
http://en.wikipedia.org/wiki/Explicit_Data_Graph_Execution
A research processor that implements EDGE:
http://wavescalar.cs.washington.edu/wavecache.shtml
this is cool but i wish the arm processors were more powerful. wouldn't writing software for this be hard since theres so many processors involved?
It's not hard that is a myth. All developers have a PhD in mathematics so splitting a complicated logic process and synchronizing across 8-256 cores it is a piece of cake. German chocolate cake.
Embarassingly parallel problem then yea it's easy. Inefficient solution to a complex problem then it's fairly easy. A good, efficient solution to a complex problem then it's not easy.
I get the impression from "They autonomously communicate, distribute processing power and programming, and even isolate faulty neighbors -- when one of the cells crashes, it won't bring the whole cluster down with it." that you load the code into the 'cloud' of cores, and let them do the sorting out - no need to worry if you've got 2, 10 or 100.
However, whilst the redundant nature of the beast is handy, I don't see why people are going to need the parallel processing aspect - You're probably going to use another solution if you need more computing power than say 2-4 of these clustered together provides.
One to watch though - I'm sure some bright spark will identify a need....
So basically this is the same idea as Creative's Zii "stem-cell computing". Yawn.
Yes, except with slower older processors.
It seems if you want software that works with this, than you must rethink software on a fundamental level. How it works and how WE see the data. If this is based of off cells, than the software should be self evolving. Rather than software that is designed on legacy interfaces and such. We need to come up with a new solution that can make this hardware work better for us. Basically we should not try and make this hardware just recreate the current OS environment we are already used to. It should also possibly use something else as a visual interface device rather than a moniter. Perhaps we could use 3d goggles that are floating around?
Wow amazing!
I just can't wait to get one of these and watch each board tell the other to flash different colours by checking the other one!
In all serious-ness though, I do understand whats going on here, and why its significant. I just don't see any reason that this couldn't be achieved with a couple of desktop computers anyway. This sort of thing is nothing special running on made-for-purpose hardware. The real ingenious is making this stuff run on existing hardware.
Looks neat, but this sounds a lot cooler than it likely is. They hooked up some low power processors and wrote a firmware which supports ICSP.
I'm sure it was difficult. But, it's been done.
Make them smaller, faster, and able to connect in 3 dimensions. Then I think we're onto something.
From my understanding of ARM-based boards, these types of slower-speed processors are intended for embedded applications where an entire PC would be overkill. I could see how developing an auto-detecting and apparently easy to use method of setting up parallel processing networks for slower processors could be very useful. Imagine, your microwave could communicate with your freezer, letting you know when your freezer is low on its stock of hot pockets.
Sure, simply looking in your freezer could immediately provide the answer, but 50 years down the line, we are all going to be so fat from eating hot pockets all day that we won't bother to get up unless we know there's a hot pocket waiting for us.
Haaaaaaawt Pockets! http://www.youtube.com/watch?v=J9c9lAfXQHs
Connection Machine (http://en.wikipedia.org/wiki/Connection_machine) even got the groovey LEDs :-) The CM itself was very heavily influenced by the ICL DAP in the 70s. I also remember one of the home computing mags in the 80's giving details of how to build a farm of Zilog Z80s to compute the Mandelbrot set in parallel. More power to their elbows. Eventually we will hit a point where bespoke hardware is required as we finally hit the wall on Moores Law. Even language specific processors are making a comeback (OK not much but someone is touting a "Java Machine") dreamy memories of the Japanese 5th Gen projects :-) It's just more fun having a bit of variety and a bit of asynchrony will keep a programmer on their toes.