Slashdot Mirror
RC5-72 Clients Available on distributed.net
Yoda2 writes "From the distributed.net site... 'The RC5-72 project is now officially up and running, as of 03-Dec-2002! You will need to download a new client in order to participate. Our FAQ-O-matic has been updated with the beginnings of a new RC5-72 section.' Also, there is a $10,000 prize for the winner, but as with the other RC5 projects, the owner of the computer that finds the key does not get all of the money."
I'm torn as to wether or not I want to participate in this, or Folding@Home.
I ran the RC5-64 Project for a long time. I like it, In my year(s?) or participating I developed a habit of definding it, explaining it, and had grown to care for it.
But when the end of RC5-64 came along I was left idle. I believe that some good can come of these distributed projects, but I've never made the effort to install F@H on my assorted boxen, my own little garden.
I'm well versed in the cow though, and could be back on RC5 quickly...
argh, choices, choices.
Computational Madness in a round package.
Anyway, I think all of these efforts would benefit from some real competition. You can't believe how rewarding it was to race with distributed.net and the other efforts and to see who can develop best optimized code - for example.
But to build that spirit of competition (without doing duplicate work) between the efforts, we would need some fresh and new (reasonable, interesting) idea for: what to crunch? Any ideas there? I am sure the guys at distributed.net and the multiple other efforts would love to see the same "fighting spirit" again as well :) And as result, I believe everyones code will be optimized much faster and new ideas will be created faster, more people will be interested to join...but: what to crunch, what would be really really interesting? :)
I'd like to see the cost of electricity to power all those cycles, and compare it to the $10,000 prize.
Cancer Research - if you install the google bar you will be helping the google team with the folding@home project for simulating the folding of proteins.
- some how this is supposed to help with cancer but I am no doctor and have no idea what folding protein means - i just draw pretty pictures all day, but this makes me feel warm and fuzzy inside after passing by the bums on the street.
Ave Molech Setting
OK, granted that this project may be a waste of computing power (assuming that they're not going to be just sitting there wasting cycles anyway), but I saw a lot of people suggesting that users instead participate in the folding@home project. That got me to thinking...
I'm not against folding@home, but I don't think that the number crunching approach to solving protein folding is ever really going to give us the breakthroughs we want. We need to theoretically address the issue of folding and find more simple behavioral theories with which to approach the problem. I know a lot of work is currently being done from the physics front with spin glasses and other complex systems models.
The difference between these two approaches is the difference between the current encryption cracking projects, and a Sneakers-like approach to actually find a mathematical solution to the large number factoring problem.
(email addr is at acm, not mca)
We are Number One. All others are Number Two, or lower.
--The Sphinx
The principle of Optimum Slackitude points out that because of Moore's Law, the overall cost in time or money can be decreased by waiting to being. If current numbers predict 12 years to exhaust the keyspace, and we wait 18 months to start, then that first 18 months worth of effort will have to be made up at the end, but 12 years later computers will be 2^8 or 256 times faster. That first 18 months worth of effort will only take 2-3 days to make up at the end of the project.
I think that's probably what people object to about starting this project now instead of in a couple years.
Or you could just get 20 times as many people to run the client. There are LOTS of unused CPU cycles in the world. Probably 99.999% of all CPU cycles are doing nothing but spinning in main{} right now. Let's put 'em to work!
Or let's put 'em to sleep! I used to leave all my computers running all night just to crack RC5. I noticed a significant drop in my power bill when I started turning machines off. (Also, my laptop battery started lasting 2.5 hours instead of 40 minutes.)
Maybe if they started paying for my cycles, I'd reconsider, but I'd still have to look at peak power prices first.
Tim
(t/1.5) = log_2 (256) = 8
t = 8 * 1.5 = 12.
Uhm, not quite. That's how long it will take before our machines are 256 times faster, which is a very different question. (It would be tempting to just multiply this number by 4, the number of years it took to solve RC-64, but that would merely tell us how long it would take the computers of 2014 to solve RC-72 [answer: 48 years].)
You need a more nuanced answer that takes into account your exponential progress as you're ramping up to full speed.
Let C be the Moore doubling time. Let P be the number of computations required to solve RC-64. Let X be the instantaneous speed at which you can solve problems, in units of P/year. So for t = 4
1/2 x/C t^2 = 1
so x = (C/8 years) P/year
Given that, we can calculate t in this equation:
1/16years^2 t^2 = 256
t^2 = 4096 years^2
t = 64 years
When I moderate, I only use "-1, Overrated". That way, I never get meta-moderated!