Author

Topic: [POOL][Scrypt][Scrypt-N][X11] Profit switching pool - wafflepool.com - page 155. (Read 465716 times)

full member
Activity: 168
Merit: 100
From <1% rejected consistently to >10% rejected so far.

I am testing different configs on 2 rigs now to see if I can get the rejects down - hard to sample over a short period of time.  I do know that I hated this block reject scheme on clevermining, however.  I liked the reject scheme on waffle before. But, I will admit I am very interested to see what kind of reject rate sfire would have under this new scheme - it might made a big different on profits depending on how badly people were abusing this.

Can we get some reject % data, and maybe ourselves vs pool (along with shift info maybe)?

So that would mean you were one of the miners who were taking up to 10% more of your fair share of profits from the pool, however unknowingly.  For the others who had their equipment tuned correctly for mining almost anywhere else than wafflepool, they were receiving up to 10% less of profits directly due to their efforts, depending upon the ratio of efficient to inefficient miners in the pool.  The actual percentages depend upon whether wafflepool was actually a pool of fools, and only poolwaffle knows that as he is the only one with access to the relevant statistics.

I too would be very curious about sfire's 20GH/s mining efficiency prior to the server changeover, which I consider to be a very welcome one.  Ironically, depending upon that statistic, it might not have been the largest miner that was keeping us from efficiently mining the lowest difficult coins, but the smaller miners instead!
hero member
Activity: 700
Merit: 500
Earnings rate for 2.6 MH/s on the charts for the past 2 hours is almost zero?Huh?
Lots of litecoin mining, but no litecoin blocks.  Variance sucks sometimes.
member
Activity: 70
Merit: 10
Earnings rate for 2.6 MH/s on the charts for the past 2 hours is almost zero?Huh?
newbie
Activity: 16
Merit: 0
Still loving the low reject rate on the server, Waffle.  Great job!

phzi-
what cards are you running and what is your config?

http://i.imgur.com/MMA2UjP.jpg
newbie
Activity: 56
Merit: 0
PW, i am sure that this is something pretty complex. but do you think it is possible to implement more intelligent version of "sell everything for btc" script to sell little coins proportionnaly to what miners have selected as their worker name - btc or ltc wallet.
is it possible at all without a huge amount of frontend and backend coding and making all miners to register?
Given the current activity around LTC right now, I wouldn't mind holding the LTC I am mining either.  But, it seems like it would be quite difficult to implement given wafflepool's no-signup system
you dont have to register to submit your L wallet as worker name instead of 1... i guess problems will be with coins wich are not directly tradable to ltc? lets wait for PW to do some more math for us )
i would love to add a picture of 3d cat with popcorn, but we ran out of popcorn.
newbie
Activity: 47
Merit: 0
If we could make LTC payout system that would be perfect!
Everything is that LTC will have bigger rise(in %) of profit in coming days/months comparing to btc.
hero member
Activity: 700
Merit: 500
PW, i am sure that this is something pretty complex. but do you think it is possible to implement more intelligent version of "sell everything for btc" script to sell little coins proportionnaly to what miners have selected as their worker name - btc or ltc wallet.
is it possible at all without a huge amount of frontend and backend coding and making all miners to register?
Given the current activity around LTC right now, I wouldn't mind holding the LTC I am mining either.  But, it seems like it would be quite difficult to implement given wafflepool's no-signup system
newbie
Activity: 56
Merit: 0
PW, i am sure that this is something pretty complex. but do you think it is possible to implement more intelligent version of "sell everything for btc" script to sell little coins proportionnaly to what miners have selected as their worker name - btc or ltc wallet.
is it possible at all without a huge amount of frontend and backend coding and making all miners to register?
if mined coin is not directly tradable from sh$tcoin to ltc -
1 you can trade it sh$itcoin-btc-ltc and i think that additional fees will be covered with no need to trade actually mined ltc.
2 you can trade less than 100% of mined ltc-btc to send what left to those with L wallet.
3 you probably now have enough btc/ltc for some kind of hot ltc wallet to use like a buffer to make payouts in ltc to those with L wallet. at least i hope you do. probably not, if our chinese cyborg will suddenly want to submit his L wallet))).
umm. it is complex indeed ).
hero member
Activity: 700
Merit: 500
What is "through the roof"?
From <1% rejected consistently to >10% rejected so far.

I am testing different configs on 2 rigs now to see if I can get the rejects down - hard to sample over a short period of time.  I do know that I hated this block reject scheme on clevermining, however.  I liked the reject scheme on waffle before. But, I will admit I am very interested to see what kind of reject rate sfire would have under this new scheme - it might made a big different on profits depending on how badly people were abusing this.

Can we get some reject % data, and maybe ourselves vs pool (along with shift info maybe)?
member
Activity: 84
Merit: 10
If you are getting high reject rate AND you set intensity high, try reducing I by 1 (i.e. I=19, change to I=18), then observe the rejects for at least 30 min.

sr. member
Activity: 322
Merit: 254
Alrighty guys.
...new code...rejections...optimisation
OK, forgive me if these are daft questions but as I'm new I'll ask... Smiley
1) I know I can see the rejections in cgminer (A: xxxx R: yyyy for each card) but is there somewhere I can easily see the totals / % from the pool's point of view or do I need to add them all up across all my cards?  In other pools they have a stat which shows your total rejects but I don't know if I we have anything like this on WP
2) At what level should I start to take action?  What sort of % would you suggest should cause me to reduce intensity?
3) What action to take? Smiley  Now I know that seems like a silly question in that if the % is too high I need to reduce intensity but I assume I need to knock it back 1 or 2 at a time and then wait to see the impact on rejections but how long to wait?  Is it a pretty instance cause and effect or is it sensible to leave it for 1 hour or 1 day etc?  Or is there somewhere to go to see what intensity works for my cards?

I just want to make this as profitable for me and everyone else as possible and don't want my lack of knowledge to get in the way!  Especially as you're doing such hard work for us all Smiley

1) I'll be adding to the miners page shortly (it was never high enough to really put on there)
2) Its not % based unfortunately, its % compared to increase in hash power.  For example, early on bumps in intensity (say 8 to 9) will yield a TON more hashpower, for a very small increase in rejects (say +10% shares, +0.1% rejects, very obviously worth it).  Later intensity increases will yield a small bump in hash power, for a larger increase in rejects (say +1% shares, +2% rejects, probably not worth it).  It has to be done on a per miner basis unfortunately.  Typical ranges for rejects are around 2-3%, and aren't terribly worrysome.  10%+ is probably worrysome.  It also depends a lot on your latency to the stratum server...
3) Back off intensity, or try to use a "closer" (ping-wise) stratum endpoint.  If you're in asia, and you're using the USEast endpoint, you're probably doing it wrong Smiley

I'm in the UK so using the EU point as my primary.
I'm using the settings in Cryptobadger's blog :
 ./cgminer --scrypt -I 20 -g 1 -w 256 --thread-concurrency 24000 --gpu-engine 1050 --gpu-memclock 1250 --gpu-vddc 1.087 --temp-target 80 --auto-fan
With this I am getting fairly good hash levels but my rejects are between 2 and 9% (taking the % to be r/a for each of the cards and then averaging them) which seems high.
The cards are MSI 9750.  I guess I need to reduce my "I" setting?  Anyone got any suggestions as to what level to set it at to retain the performance?
2% seems fine, 9% seems high.  Unfortunately, you just need to tweak them and see what changes, it really depends on each card individually.  Turn it back a bit and see if you get overall more shares accepted than before.
newbie
Activity: 55
Merit: 0
Alrighty guys.
...new code...rejections...optimisation
OK, forgive me if these are daft questions but as I'm new I'll ask... Smiley
1) I know I can see the rejections in cgminer (A: xxxx R: yyyy for each card) but is there somewhere I can easily see the totals / % from the pool's point of view or do I need to add them all up across all my cards?  In other pools they have a stat which shows your total rejects but I don't know if I we have anything like this on WP
2) At what level should I start to take action?  What sort of % would you suggest should cause me to reduce intensity?
3) What action to take? Smiley  Now I know that seems like a silly question in that if the % is too high I need to reduce intensity but I assume I need to knock it back 1 or 2 at a time and then wait to see the impact on rejections but how long to wait?  Is it a pretty instance cause and effect or is it sensible to leave it for 1 hour or 1 day etc?  Or is there somewhere to go to see what intensity works for my cards?

I just want to make this as profitable for me and everyone else as possible and don't want my lack of knowledge to get in the way!  Especially as you're doing such hard work for us all Smiley

1) I'll be adding to the miners page shortly (it was never high enough to really put on there)
2) Its not % based unfortunately, its % compared to increase in hash power.  For example, early on bumps in intensity (say 8 to 9) will yield a TON more hashpower, for a very small increase in rejects (say +10% shares, +0.1% rejects, very obviously worth it).  Later intensity increases will yield a small bump in hash power, for a larger increase in rejects (say +1% shares, +2% rejects, probably not worth it).  It has to be done on a per miner basis unfortunately.  Typical ranges for rejects are around 2-3%, and aren't terribly worrysome.  10%+ is probably worrysome.  It also depends a lot on your latency to the stratum server...
3) Back off intensity, or try to use a "closer" (ping-wise) stratum endpoint.  If you're in asia, and you're using the USEast endpoint, you're probably doing it wrong Smiley

I'm in the UK so using the EU point as my primary.
I'm using the settings in Cryptobadger's blog :
 ./cgminer --scrypt -I 20 -g 1 -w 256 --thread-concurrency 24000 --gpu-engine 1050 --gpu-memclock 1250 --gpu-vddc 1.087 --temp-target 80 --auto-fan
With this I am getting fairly good hash levels but my rejects are between 2 and 9% (taking the % to be r/a for each of the cards and then averaging them) which seems high.
The cards are MSI 9750.  I guess I need to reduce my "I" setting?  Anyone got any suggestions as to what level to set it at to retain the performance?

Thanks

Miles
sr. member
Activity: 322
Merit: 254
My rejects are through the roof... I have discovered 1000s of blocks for WafflePool, of which only a few were orphans.

What made WafflePool seem so fair previously, IMO, was that share generation was statistically consistent between mining fast coins and slow coins.  Now, a share is "worth more" then a slow coin share, because the slow coin will have less shares lost to 'rejects'.

What is "through the roof"?  Share value shouldn't change, as slow/fast coins are spread amongst servers evenly, so you'll have some sections of time where you mine fast coins (and get a slightly higher reject rate - I'm seeing 5% avg rejects on fast coins), and times where you'll mine slower coins (seeing 1-2% reject rate on lite/doge).  But those will even out to be the same for all miners.
hero member
Activity: 700
Merit: 500
My rejects are through the roof... I have discovered 1000s of blocks for WafflePool, of which only a few were orphans.

What made WafflePool seem so fair previously, IMO, was that share generation was statistically consistent between mining fast coins and slow coins.  Now, a share is "worth more" then a slow coin share, because the slow coin will have less shares lost to 'rejects'.
newbie
Activity: 57
Merit: 0
But you're also mining litecoin which was increasing in value and therefore the exchange ltc/btc was alot more beneficial to us, thus the increase in profitability on the graph?

(not jinxing it, just asking Smiley)
sr. member
Activity: 322
Merit: 254
While its only a small amount of data, and it could _very_ easily be luck skewing the graphs, here's what my data is showing:



First yellow line is when we switched about 15% of the pool.  Somewhere after that (don't have an exact time), we switched to 25%.  Second yellow line is when we switched the rest of the pool (roughly).  Left side axis is BTC/MHs, right side is hashrate (you can see sfire pretty blatently on there) Smiley
newbie
Activity: 26
Merit: 0
I hope the lower difficulty coins work out I could use a few more fractions of a Bitcoin.

[16:13:41] Network diff set to 338M
[16:13:41] Stratum from Pool 0 detected new block
[16:13:41] Pool 0 stale share detected, submitting as user requested
[16:13:41] Rejected 1ac0d8b0 Diff 2.45K/512 GPU 0 Pool 0 (job not found bb)
...
[16:17:05] Pool 0 stale share detected, submitting as user requested
[16:17:05] Accepted 72412b6b Diff 574/512 GPU 0 at Pool 0

I might need to adjust the intensity.  Reject rate is currently 2.8% but it has only been about 25 min since I zeroed the statistics.

sr. member
Activity: 322
Merit: 254
You have to start selling servers dude, now that you have the stuff optimized they are underutilized Cheesy

Rediculously underutilized.  On our bigger boxes (EU/USE), we went from a load of 2-3 consistently, to a load of 0.3-0.35...  Can easily scale back our servers Smiley  Just want to wait until people start coming back from Ghash to make sure it stays that way Smiley
sr. member
Activity: 322
Merit: 254
i know there have been several posts but any ideas where all of our hash went?

Mostly GHash for the promotion, the vast majority will be back in 1 day 16hrs Smiley
newbie
Activity: 49
Merit: 0
You have to start selling servers dude, now that you have the stuff optimized they are underutilized Cheesy
Jump to: