Parallella low-cost personal supercomputer
The Parallella project is developing an open, affordable, low-power, high-performance parallel computing platform. It sounds too good to be true, but Adapteva, the team behind the project, has a history of achieving aggressive engineering goals in exactly this space. Moreover, the supercomputer is based on the existing Epiphany line of chips.
This EE Times article explains it better. They’re planning two versions of the computer; a 16-core (based on the Epiphany E16) version for $100, and a 64-core version (using the Epiphany E64) for $200. The latter would deliver over 90 GFLOPS (90 billion floating-point operations per second), on a board the size of a credit card, while consuming only 5W of power.
If you’re a robot hobbyist, a system like this should excite you. Many of the algorithms we need to make our robots truly intelligent are well known, but require lots of computing power. Vision is a good example; speech recognition is another. As our robots get physically more capable, it will soon be how well a robot can see, converse, and interact with its environment; a beefy board like this makes such problems a lot easier.
But the Parallella project is even more important from a broader perspective: parallel programming. Individual CPU cores aren’t likely to get much faster; increases in computing power these days come from parallelism. But parallel programming requires a quite different mindset and approach than traditional programming, and frankly, as a species, we suck at it. Adapteva is making this low-cost, high-performance, highly parallel system available with all sorts of development tools and support for primarily one reason: to help build a culture of developers skilled in parallel programming. I think this is an important and worthwhile goal.
However, time is running out. I just found out about their KickStarter project today, and its deadline is tomorrow. At the time of this writing, they have $612,075 pledged toward their goal of $750,000. Their project page is a little dry and technical, and I don’t think really conveys the importance (at least, not as well as the EE Times article does). If they don’t reach their goal by 6PM (Eastern) tomorrow, the project is cancelled, none of the backers are charged, and, quite likely, the world misses out on a great opportunity.
So please do the world a favor: head on over to the Parallella page at KickStarter, watch the videos, and consider making a pledge!
UPDATE (10/27/2012): the Parallella project was successfully funded. I can’t wait to see what people do with it!