Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
Compare cell phone plans using Wirefly's innovative plan comparison tool ×
Bitcoin Math The Internet Hardware News Technology

Best Way To Mine Bitcoins - Allow Errors! 167

An anonymous reader writes: A recent paper from the University of Illinois at Urbana-Champaign shows that bitcoin mining profits can be increased considerably if mining hardware is allowed to produce occasional errors. The research shows that mining hardware that allows occasional errors ("approximate mining") can run much faster and take up less area than a conventional miner. Furthermore, the errors that are produced by the miner do no corrupt the blockchain since such errors are easily detected and discarded by the bitcoin network. Mining profits can increase by over 30%.
This discussion has been archived. No new comments can be posted.

Best Way To Mine Bitcoins - Allow Errors!

Comments Filter:
  • BeauHD (Score:1, Offtopic)

    What are your working hours? You have to post an article once and hour while you work?
    • by BeauHD ( 4450103 )
      Nope but thereabouts is a good spacing wouldn't you say?
      • I like it! Keep up the good work, we appreciate it!
        • by BeauHD ( 4450103 )
          Thanks, obviously still getting the hang of things but I appreciate it!
      • This the first article of the day. The last one was 14 hours previous. Are some articles not showing up? Or did I misunderstand the question?

  • by Anonymous Coward on Sunday February 14, 2016 @12:12PM (#51506019)

    So, they basically shift part of the workload necessary for mining onto the nework, and thus pay less?

    • Re: (Score:1, Funny)

      Sounds like a Republican thought this scheme up.
      • by Anonymous Coward

        Why expect me to do the work? It's on your ass if you're not satisfied!

        See also Flint, Michigan. If you can't afford to test your own water, you shouldn't make us pay the price.

      • by JustNiz ( 692889 )

        Lol I wish I had mod points to give you.

      • You got it wrong. Democrats are the ones creating jobs to fix their mistakes.

      • by Solandri ( 704621 ) on Sunday February 14, 2016 @03:04PM (#51506675)
        Both parties are the same. Republicans shift workload onto society via corporate welfare. Democrats shift workload onto society via government welfare programs.

        There's nothing "wrong" per se with either type of program as long as you can keep inefficiencies down. In that respect it's good to have both R and D at each others' throats - helps keep their pet programs honest. Corporate welfare helps to guide the direction of economic development and technological progress (renewable energy being the latest example). Social welfare helps to act as a safety net allowing people who fall on hard times to pick themselves up and become productive members of society again more quickly.

        But when any of these programs are given a free pass (as the most staunch conservatives and staunch liberals do), inefficiency (corruption) begins to creep in and their cost to society eventually exceeds their benefit. Corporations take the subsidies without producing a meaningful product. People mooch off welfare with no intent to re-enter the workforce. If you can see the possibility of that happening with only the half of these programs your political party opposes, you're part of the reason our government wastes so much money.
        • by Dahamma ( 304068 ) on Sunday February 14, 2016 @04:16PM (#51506963)

          Both parties are the same. Republicans shift workload onto society via corporate welfare. Democrats shift workload onto society via government welfare programs. ...
          But when any of these programs are given a free pass (as the most staunch conservatives and staunch liberals do), inefficiency (corruption) begins to creep in and their cost to society eventually exceeds their benefit.

          True, and an insightful point. Though I would argue that the result of corporate welfare abuse is that the already rich 0.1% benefits the most. The result of social welfare abuse is that the poorest 25%+ benefits the most.

          Neither form of waste and corruption is ideal, of course, but if I had to choose the lesser of two evils I'd prefer to direct the inefficiencies to the 25% of the poorest people in society living day to day with underfed kids and horrible health care than I would to millionaires wanting to add an extra comma to their net worth.

          • by cadeon ( 977561 )

            I look at it this way: Any government spending hurts the poor and benefits the rich.

            When the government spends money, it borrows it. Borrowed money is created out of thin air by the banks. Reserve requirement.

            Money creation causes inflation.

            Inflation means it costs more for the poor to buy stuff, and it means the stuff the rich own becomes more valuable. (if you bought a $100k house 30 years ago, it's worth $300k now thanks to inflation... you made $200k simply by owning it).

            Unfortunately this means huge go

            • by Dahamma ( 304068 )

              Except many of your points are just plain nonsense, sorry. Borrowing does not by definition create money out of thin air and cause inflation. The US government borrowed tons of money in 2015, and inflation was way below average. Inflation can be due to a government printing money because it CAN'T borrow enough. If lots of people are willing to lend the government money at low rates (like in the last couple years), then inflation can be really low even if govt borrowing is high.

              Inflation means it costs more for the poor to buy stuff, and it means the stuff the rich own becomes more valuable. (if you bought a $100k house 30 years ago, it's worth $300k now thanks to inflation... you made $200k simply by owning it).

              Whoa, that's absurdly inco

              • by cadeon ( 977561 )

                I don't believe my premise is wrong, but I may have not made it all clear due to brevity. Of course reality is more complex than a few sentences in a slashdot comment; if it were this simple we'd all be economy experts. Let me try to clear things up.

                Inflation can be due to a government printing money because it CAN'T borrow enough.

                The government doesn't print money. Well, it does, but not nearly enough to matter, and in order to print it, the money has to be created first via a borrowing event. Look up how that all works.

                When the news says the government is 'creating money' they really me

    • They say the network would detect the errors, but more likely, an adjacent, more reliable processor would—there is no reason not to double-check the relatively few attempts that would make it through the first stage (as happens already e.g. with pooled mining, where you get credit for things that are not as hard as the official difficulty).
    • I don't get it. I thought the whole idea was that a hash was quick to calculate, but it was hard to find a nonce that yielded a correct hash since there are so many nonces to try. So once your approximate hardware finds a nonce, why don't you just check it on a regular processor before sending it to the network? Sending off a possibly bad block to the network seems like a lazy (not to say sleazy) thing to do.

      Hey, I've got an even better idea to increase profits: just send loads of random blocks to the netwo

      • Because that would defeat the whole purpose of this method. This method only works faster by simply offloading the actual checking (recalculating the hash) onto the network (which has to do it anyway but now you've increased the workload on the network with a ton of false positives).

        • Re: (Score:3, Interesting)

          by slashping ( 2674483 )
          No, the false positives are easy to detect locally. A regular CPU in your PC could detect a false positive in much less than a microsecond, and the number of false positives would be small to start with.
      • by tricorn ( 199664 )

        From reading the linked page (which is woefully short on details), this isn't approximation, it's more akin to overclocking hardware so it goes faster even though it occassionally will make a mistake. If it can run twice as fast, but only create 10% false positives and negatives, it's worth it (as long as the overclocking doesn't use twice as much power).

        Sending it on to the network without double-checking that it's valid would be stupid, though, considering it takes less than a microsecond to check a resu

        • False negatives are a problem, but as the percentage of them is less than the gain in total amount of hashes, it's still a net win to do approximations.
        • by tricorn ( 199664 )

          Reading the paper further, while they do discuss overclocking, they are also allowing for occasional errors in an adder, specifically by assuming that long carry chains won't occur, thus avoiding the need for carry lookahead circuits to ripple the whole word length. That means occassionally an add will get the wrong result, but not very often, but the add circuit can be made faster. It's still deterministic, just that a few specific bit patterns aren't handled correctly.

        • by AmiMoJo ( 196126 )

          The real problem isn't the false positives but the false negatives, as those are missed wins.

          Could BTC be "lost" this way, I wonder. Do candidates get checked by more than one person, or is it organized like other distributing computing systems where blocks of work are allocated to avoid duplication of effort?

          Allowing errors is somewhat like how GPUs work. The reason the professional cards cost more despite having the same hardware is that they are guaranteed to produce correct results. Gaming grade cards might produce the occasional error, but it will likely be unnoticable to the player and the ca

    • No, that part of the paper is bullshit. Once you think you've found a hash, it's easy to verify with a regular CPU before you'd claim to have found a block.
    • by dbreeze ( 228599 )

      BRILLIANT!!!

    • Basically, it is the same sort of thing that has been done in medicine for ages -- use a cheaper test with a high rate of false positive, but low rate of false negative, and verify the (few) positives with a more expensive and reliable test.

    • No, verification is trivial. I have no idea why he wouldn't verify the result.

      Mining pools use a similar scheme to distribute work, sending each miner what amounts to a partial problem only, while checking answers to see if they're valid before submitting to the network and awarding any bounty to the miner. This was done so that miners with varying hardware could all effectively help the pool without their tasks expiring and so that miners couldn't submit work to the pool and then submit any blocks found

  • by Drethon ( 1445051 ) on Sunday February 14, 2016 @12:20PM (#51506061)
    Approximations can solve problems faster than exact algorithms but with occasional errors? Who would have thought? https://en.wikipedia.org/wiki/... [wikipedia.org]
    • Yeah, so great that you published all those scientific papers. Thanks for your wonderful contributions to science!
      • You are welcome, I just don't choose to post my published paper on slashdot since it shows the kind of crap that is accepted in publications.
    • You make it sound like it was obvious, but SHA-256 hashing requires 64 rounds of a hash function, and bitcoin mining requires a double SHA-256 hash. Even a small error in the hash function is likely to propagate quickly to the point where practically all of the results are bad. So, in this case the approximation must be nearly perfect to be worthwhile.
  • Just scam them from Martin Shkreli's dumb ass. [imgur.com]

  • by John Allsup ( 987 ) <<oc.puslla> <ta> <esuoh.anni.rotcod>> on Sunday February 14, 2016 @02:36PM (#51506553) Homepage Journal

    That is what is done with writing research papers these days: pay sufficient lip service to the notion of science to get published, and if errors turn up, just ignore them. Increases rate of publications significantly, and that's what people get paid for these days.

  • This is cheating (Score:4, Insightful)

    by mysidia ( 191772 ) on Sunday February 14, 2016 @02:46PM (#51506601)

    The research shows that mining hardware that allows occasional errors

    All you are doing is making other people do the checking work for you.

    This is wasting other people's CPU and bandwidth.

    Your mining pool should ban you if you're caught doing this.

    Completing 'fake' shares, which ultimately enrich yourself at the cost of the total profits of your pool and other miners.

    • by dbreeze ( 228599 )

      I think the exploiters have already grokked those issues...

  • If the hardware misidentifies a solution as valid, the network will quickly identify that the solution is invalid when the miner broadcasts it. “These are essentially immediately rejected.”

    Why broadcast the block straight away? Surely checking if a given solution is valid is quick enough that it could be checked by a different system before being broadcast?

  • We've seen the results of an occasional error in that work!!!
  • This is super obvious...how did this paper get published?

The possession of a book becomes a substitute for reading it. -- Anthony Burgess

Working...