It's a common misunderstanding that score based reward systems like Slush's underpay infrequent miners. While it's true that shares in a given round devalue if you mine towards the beginning of a round and then leave long before it finishes, you have to also take into account the fact that the length of a round is random. So you have equal chances of arriving late in a round and having your shares worth more than they would have been by straight proportion, or starting out a round and having it find a block right at the beginning (making each share worth a lot more because there were fewer of them). In the long run, these all average out perfectly.
The only circumstance in which you will be receiving less from a score based pool is if you consistently start out rounds but then switch to other pools when the rounds "take too long" for your liking. You can't do this by accident. It's a specific attack that we identified and analysed back in January, and unless you specifically program a miner that does this it's not going to happen consistently by random chance.
That said, there is a slight difference with the score based system for very low-capacity, infrequent miners: increased variance in your payouts. Sometimes they will be bigger than normal, and sometimes smaller. Mathematically, it all ends up coming down to the fee you pay, but psychologically this might be hard to take as sometimes you will feel ripped off, while you might not notice that you also get lucky just as much. So it might be easier on your brain to go with a pay-per-share system or similar, it just won't be easier on your pocketbook.
The biggest underlying problem with this is that you probably shouldn't be mining at all. Even if you only use it when your computer is on anyways, a 5670 almost certainly draws more power when mining than you earn back in bitcoins, unless you're paying very low electricity prices (5 cents or less per kilowatt-hour) or using someone else's electricity (in which case all you're doing is taking money from them, not earning it). Probably it's time to leave the mining to high-efficiency rigs located in very low-cost electricity zones.
For a complete analysis of different mining pool schemes, Meni Rosenfeld has written an excellent paper.
Having only an 5670, I received 0.0+0.0+0.0004 btc for mining 2.5, 3 and 4.5 hours. With increased difficulty this can be even worse. So much for the "edge case". – mbx – 2011-09-01T16:12:10.357
2True, I hadn't considered miners so slow (or rounds so fast) that they may not submit any shares in a given round. Mr. Schwartz may be correct in choosing PPS for such a scenario. I suppose as someone pushing 3 GH/s I need to revisit my definition of "casual miner" before making a judgement call. Still, for anyone with even a single 5830 or higher my advice still holds. – David Perry – 2011-09-01T16:14:28.543
See answer below--size of miner doesn't actually impact the fairness of score-based mining, only the variance. – eMansipater – 2011-09-02T15:45:43.657
It does if the size is extremely small. While CPU miners are almost nonexistent these days, mining at 1-10 MH/s only submits a share every several minutes. If a user has only a few shares per round their placement can make or break profits in any scoring method that involves time in its calculations. With a 5670 the OP should be pulling ~70 MH/s which is obviously above this threshold, but the lower your hashrate the more noticeable the effect can become. Better for low-hashrate miners to move to something like PPS and take luck out of the equation entirely. – David Perry – 2011-09-02T15:50:05.960
Wow, I never imagined the models to be this complicated. I thought you're just ... getting paid your fair share of hashing power per X blocks. – ripper234 – 2011-09-02T17:55:28.220
@ripper234 Unfortunately the simpler models were also vulnerable to various forms of abuse, so they've evolved into more complex models to prevent such abuse. – David Perry – 2011-09-02T18:06:46.407
@David - interesting. could you elaborate? Or should I open a new question? – ripper234 – 2011-09-02T18:35:09.573
@ripper234 It's more complicated than should be explained in a comment but if you'd like to open a new question I'll be glad to answer it. – David Perry – 2011-09-02T18:42:57.377
I got ~80Mhash/s with XP, now it's ~75Mhash/s. – mbx – 2011-09-02T20:09:06.020
@David - here you go: http://bitcoin.stackexchange.com/questions/461/what-are-the-problems-with-naive-pool-sharing-mechanisms
– ripper234 – 2011-09-02T20:23:28.943I mine very slowly, faster than a GPU, but only about 26Mhash/s so I very much prefer the PPS method, however even that is luck based as all mining is.. as some times i might go 20 minutes without finding a share, while other times i get one every 5 seconds. – Evil Spork – 2011-09-04T02:11:46.887