specialized computers for folding or other distributed computing projects

[FastFSB]

I remember that a few years back the eff created "Deep Crack" to participate in the distributed.net DES challenge. it was incredibly fast and was able to break 56 bit DES in under a day, and cost less than $200k to design and build. any thoughs as to why we dont see any specialized computers being built for folding or even distributed.net's rc5-72? is folding too complex for a cost effective specialized machine? if so, why not rc5-72?

[Christoph]

The whole idea of DC is that you let people do the work for you. There are supercomputers that are used to simulate protein folding too, but they're very expensive. With a solution like F@H, all Stanford has to do is split up the work into manageable chunks and put the finished WUs together. (gross oversimplification, of course)

[Carnil]

For RC5-72, the problem is getting a fast enough box. There's a 50% chance that the RC5-72 project will be wrapped up in the next 800 years, and no single box made with current technology will have a significant impact on that.

[FastFSB]

Christoph: Good point. I guess that would defeat the point of folding.

Carnil: Deep crack was able to reduce the time needed to break DES from months to hours, and DES is a much slower algorithm to work with. i dont think that rc5-72 could be solved any time soon but maybe a few specailized machines could speed it along greatly. DeepCrack tried 245billion keys/sec while the overall rc5-72 rate is just 64 billion. i understand that you cant compare differant algoriths that easily, but des is slower and computer technology has come a long way in the last few years since deepcrack was developed.