This does seem very interesting! Because of your comment I started reading the architecture document for the GreenArrays[0] chip, and it's totally different to anything I've seen: a network of processing units, and the whole thing without a clock anywhere!
I am interested what makes you say it "provides computational grunt comparable to an i7 for a few milliwatts" though - could you elucidate? Do you mean in terms of performance-per-watt?
> I am interested what makes you say it "provides computational grunt comparable to an i7 for a few milliwatts"
I've been quite interested in Green Arrays, but I'm quite sure that's basically nonsense. The Ga144 can do lots of interesting stuff, but comparing it with an i7 is ill adviced. If for no other reason than that you have to provide your own dram interface, while the i7 typically comes with that in the form of chipset support.
In short, I think such a claim is misbranding the Green Array chip, setting potential users up for disappointment and at the same time selling the chip short.
Interesting, it reminds me a bit of the game TIS-100[0], similar idea of an array of very simple units interacting by I/O ports, blocking until a read/write to another unit can complete. I didn't know there was a real architecture like that!
edit: It seems the transputer[1] was similar, hadn't heard of that either.
The transputer actually has a successor called the XCore[0], which is produced by XMOS. They're founded by the same fellow that came up with the transputer (David May).
It's a similar concept again - it's a multi-core chip for embedded applications, and the cores communicate using an on-chip network. Each core is like a regular CPU with a clock though - the GreenArrays approach seems to be completely asynchronous.
I am interested what makes you say it "provides computational grunt comparable to an i7 for a few milliwatts" though - could you elucidate? Do you mean in terms of performance-per-watt?
[0]: http://www.greenarraychips.com/home/documents/greg/PB002-100...