On 9/12/2012 3:44 PM, EBo wrote: > With all the discussion of the Raspberry Pi lately, I thought I would > forward this. So, when we are playing with message passing and latency > testing, there might be some other resources available. > > EBo -- > > Southampton Engineers a Raspberry Pi Supercomputer > University of Southampton (United Kingdom) (09/12/12) > > University of Southampton researchers have developed Iridis-Pi, a > supercomputer made from 64 Raspberry Pi computers and Lego. "We > installed and built all of the necessary software on the Pi, starting > from a standard Debian Wheezy system image, and we have published a > guide so you can build your own supercomputer," says Southampton > professor Simon Cox. Iridis-Pi runs off of one 13 Amp mains socket and > uses Message Passing Interface to communicate between nodes using > Ethernet. The researchers note the entire system cost less than 2,500 > pounds Sterling and includes 64 processors and one terabyte of memory. > The researchers used the Python Tools for Visual Studio plug-in to > develop software for the system. "The team wants to see this low-cost > system as a starting point to inspire and enable students to apply > high-performance computing and data handling to tackle complex > engineering and scientific challenges as part of our ongoing outreach > activities," Cox says. > http://www.southampton.ac.uk/mediacentre/features/raspberry_pi_supercomputer.shtml > > Jeez, why stop at 64? I calculate one could fit 128 RPis easily in the volume occupied by the PDP-11/20 and its "high speed" papertape reader/punch I brought into my lab ca 1971-1972.
The 128 RPis would cost less than we paid for one 8Kb (e.g., 4 Kw; the pdp-11 was by its nature a 16-bit word machine that could access 8-bit bytes directly) memory module. The power consumption of the RPis would be comparable or less. More to the point, according to some Internet notes of one John McCallum, the PDP-11/20 was good for about 0.057 MIPS. I've seen the RPi is claimed to be 965 MIPS, so this "supercomputer" would be roughly 5x10^6 faster in the MIPS race. Not that MIPS ratings equate directly to computing power, especially with such divergent architectures, but 6 orders of magnitude ought to be good for something:-) Of course, there's a big step between paralleling a bunch of uC boards and getting useful computations done. There was a bit of intellectual sleight of hand in the press release when Professor Cox went from "apply high-performance computing and data handling" to "tackle complex engineering and scientific challenges". I won't be the one to rain on their parade. The software challenges of massively parallel computation and its supervision will do that for me. Thanks, Ebo, for the note. I hadn't seen the announcement. For some more fun in the same vein, check http://www.designspark.com/content/my-raspberry-pi-thinks-its-mainframe Regards, Kent ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
