Vardiff and gridseeds

edited April 2014 in General

I have had some users mention that when a large ammount of gridseeds are on one mining instance that diff is being pushed up too high. Could i please get anyone with this issue to comment on this thread because i am about to do some tests and possibly setup a private stratum server for gridseeds as this could become a issue further down the track.

Thanks for the help :)

Comments

  • Not 100% sure it's a problem today (based on conversation in #give-me-coins) but yea, visually I see less accepted numbers with a higher vardiff.  As serraz and I discussed though, based on how vardiff works, I don't think it's a problem at the moment.  Would be willing to test a different stratum so we could track the results.

  • Same here, accepted shares seem to drop over diff 256.

  • Same here...have about 7MH of gridseed and when I run them on a server where I can set diff at 32 or 64.
    I see more consistent numbers, more accepted shares and less stales.

    Would be willing to try a separate server here....especially if my auto payout from yesterday is fixed and sent.  ;)

  • accepted share numbers dropping isnt a issue its normal of course a higher diff will mean less accepted shares but they are weighted at the server end. What im looking for is to see if payments are changing or being les sin some blocks due to less shares over time.

    Remember vardiff is there to restrict shares it raises diff so miners submit less shares which is less load on the server. However once a share hits the server is is weighted in its current difficulty.

  • edited April 2014

    So i'm only going off estimates, but according to vitbot, at approximately 9500 KH/s , I should have received about 1.6ltc yesterday, yet I only received 1.05ltc.  Based on the block count yesterday (relative, I know), I should have received about 1.1 to 1.3ltc.

    It could simply be a bad luck day - not sure.

    If you've got a stratum to test with, it's be interesting to see the difference.

    bp

  • actually the pool only found 28 blocks the day before was 36 at least in my timezone anyway.
    The way to tell is to check the rewards per block as in go into the Paymetns tab and see if the payments per block vary and if they do greatly was that block low shares.

    Basically you will never miss out per say i should not have mentioned the term miss out at all. you just may not submit shares in short and fast blocks due to too high diff beign given to the little grids.

  • ill remove the 512 failsafe and we will see what 128 or 256 roof will do

  • How hard would it be to allow miners to set their own diff?  I've seen pools with this option.  Just trying to provide a solution that would benefit all miners.

  • im not a fan of letting users choose their diff for the following reasons.

    Miners will set their diff to 1 thiniking they will get paid more (yes i have seen this on multiple pools) so a solution will need to be setup to push their diff up so they dont spam the server with redicliously low diff shares. It could also poentially be one step back in reducing server loads as chatter will greatly increase this is the same issue i face with bringing up a new startum with a smaller difficulty it will go against the whole point of vardiff and raise load on our servers.

    i can look into our options and see what would be best.

    Also i think i ned to mention that shares being submitted less oftern as vardiff give out higher diff shares is completly normal some people seem to be confused on that fact. when share diff goes up you submit less shares thats vardiffs purpose. i am just trying to gauge if it is actually an issue with small blocks (despite how uncommon they are) or just to put minds at ease.

  • After about a week of mining here with my 22 gridseeds, even with the lower accepted shares, I've seen my hashrate on the site catch up (i.e. submitted shares).  The average hash is almost exact with the hash I'm showing on cgminer.  Looks like the vardiff is working correctly with diff 512 being selected most of the time.  I guess it just took a little bit for it to show up everywhere.


    Just my input.
  • Noticed on my graph for the average peaks and dips throughout the day that when using a higher m/h setup (ie a gridseed blade) It is really noticeable and can change over 1 m/h within the time interval that the stats update. with the 5 chip gridseed miners it does not do it as drastically. Need to look into it more, but i can figure only a couple things things

    1. when it changes the difficulty (vardiff) the lower the diff the more shares i get (like it should be) the bump between the 2 worksizes might be to much for a good ammount of shares to be accepted.
    2 just a lot of variance in the blades with lots of spikes
    3 ??? something i dont understand.
    maybe set up a couple tests with a fixed diff . 1 for 256, 1 for 384 to test and see (these are the 2 diff the blades run at so I want to keep it high as possible) or if possible one in between and see where they run best
  • Thanks for the heads up. check teh black line if it matches yoru miner then all is fine. i have some blades ont eh way myself so i will be bale to do soem better testing there without intrupting you guys too much (like i have been doing on my 5 chip seeds).

  • ok my blades just arrived. ill put them on mega to see how they go as they seem to be spamming too many shares on the normal servers at present.

    This can also confirm that we are fully ASIC ready which i believe we are very close.

    If anyone would liek to test please let me know

Sign In or Register to comment.