Topic: ### A ChainWorks Industries (CWI) Project - CWIgm | Simple Powerful Stable - page 36. (Read 67732 times)

[08:06:33]Mean netDiff: 18283.5
[08:06:53]Mean netDiff: 18204.3
[08:07:03]Mean netDiff: 18153.6
[08:07:43]Mean netDiff: 17965.8
[08:08:33]Mean netDiff: 17896
[08:08:59]Share: 23200d2c
[08:09:23]Mean netDiff: 17871.6
[08:09:33]Mean netDiff: 17884

looks like the worker diff is still being jacked up a bit too high.  this is from an x2 1070 rig.  pool currently has it at diff 409.6 (same exact diff as a x6 gpu rig), which is a bit too high (128 would be ideal, with 256 still ok but pushing it).

with this diff, the rig was only able to submit one share in the last 7 blocks.  it's actually more then 7.  the miner window only has so many lines, and there are at least a couple of blocks that already scrolled out of view, so it's closer to 8-9 blocks.

so this 2xgpu rig only contributed to finding one out of the last 8-9 blocks.

this is what i meant by gpu's being chocked by high worker diff.

it's rare for any pool to show your hash that precisely all the time.  

hashrate reported by the pool will always fluctuate.  when you mine coins with fast blocks (like sigt or dnr), it will often fluctuate wildly.

keep in mind, the pool doesn't know or care about your actual hashrate.  it can only count the hash by the shares you actually submit.  when blocks move fast, your gpu will sometimes submit a lot of shares, and sometime will not be able to submit any before a new block is already in play.

for example, my single 1070 that mines at 29'ish sometimes shows up at 50 and sometimes shows up at 10 etc.

all depends how many shares it managed to submit in whatever polling period the pool uses to measure hash.

in the beginning - yes ...

it does settle in to a more reasonably accurate figure the longer you mine and more shares are submitted ...

but the pool rates are never completely accurate as they estimate the miner hash by the submitted shares in a given amount of time ( with mpos that usually 5mins ) - whereas the miner is quite accurate because its the one producing the hashrate ...

the pool will always fluctuate somewhat - but should remain within a reasonable level of accuracy ... at time it will jump wildly - as many other factors affect the share rate submitted ... but over all - its ok ...

#crysx

[Rephrase]

How many of you have the pool showing same hash rate with the miner +/- 1-5MH/s?

we are working on it ...

#crysx

yes they do look reasonable ...

nice ...

#crysx

yeah, i figured you guys were playing with the stratum to update something or other, which is why i mentioned in in a ps.  not a big deal.

i really like this approach of the same people working on both the miner and the stratum to ensure that both of them can play together as nice as possible.  it does create a closed ecosystem, which typically is not a great thing, but in this case i don't really mind.  i'd rather have an efficient closed ecosystem that is run by people who know their shit, then a hosh posh of open solutions that don't play well together.

like i said, great job so far.

now let's see what you come up with next - no pressure

Hi all.
I see several messages like that in signatum miner

[12:09:24][GPU3]:Invalid nonce

Where can i read about mesages like that? I need to know what it mean

i run my 1070's at 60tdp.  

much like overclocking a cpu, there comes a point of diminishing returns, where you pump up extra power and heat for very little return.

in the case of my gpu's, i found that i can mine dnr at 58 at 100tdp, or 53-54 at 60tdp (110w).  the difference in hash is under 10 percent, the difference in power is over 50 watts, or just over a third.

but it's not just for the electricity savings.  with less power comes less heat - my cards pretty much never go above 60C.  i'm looking at my x6 rig now, and the hottest card is 58c.  this also means less worries about cooling the room the miners are in.  less heat produced - less ac needed to cool the room ec.

and less power also means i can run the entire x6 rig off just one 1000w power supply.  the entire rig only draws just over 700 w from the wall., according to a power meter.  this means that i can add more rigs to a circuit without overloading it, and can easily run 3x6 rigs on a single 20 amp line.  i would not be able to do that at full tdp.

point is - while not for everyone, there are many advantages (beyond just saving on cost of electric) to running at lower tdp.

Can someone please comment on my speed and settings

http://imgur.com/a/FpviD

Is it good or reasonable?

these are nice results ...

tanx for sharing ...

as for the disconnects - when i work on the pool or stratum systems - these interruptions happen ...

CWIgm and the CWI-Pool system are two projects of the CWI-EcoSystem that are very much still in heavy development ... so disconnects and interruptions will happen from time to time ... but we are determined to maintain a stable continuous system for as long as possible ...

a new pool server has been commissioned - and will be finalized by the end of the month if all goes well ...

this has been planned for a while - and full details will be in our main CWI-Thread when w are closer to the launch / test period before putting it into full production ...

the CWI-Slack is also a place to get together and exchange info for ALL  our projects - not just CWIgm ... each has a subchannel and a small but nice group of people ...

we are testing a couple of things tonight - and if successful - may become part of the next revision of CWIgm ...

#crysx

this was explained in an earlier post ...

https://bitcointalksearch.org/topic/m.20741756 ...

if there is no need to work with this parameter - please remove it ...

#crysx

running for well over 24 hours now.

speeds are same as in my last report (29 and change per 1070 - 125/500/60tdp).  everything has been stable, with no crashes and no other wacky behavior.

last night worker diff was still set too high, leaving my single and x2 gpu choking.  a diff of 256 is just waaaaay too high for a single gpu

that said, i checked this morning, and it looks to have already been fixed.  at this moment my single gpu rigs run at diff 64, which through my personal experimenting on both yiimp and nova looks to be the sweetspot for a single 1070.  the x2 rig runs at 128 etc. excellent.

my x6 rig however is set at 258, which is just a tad too low.  ideally, i would love to see it at 384 (6x64).

overall, very impressed so far.  the high worker diff issue was brought up yesterday, and is already fixed.  miner runs great, pool keeps finding blocks, and everything is working as it should... so all, good.

great job, guys.

ps... looks like the pool was down for a bit this morning, since i see a bunch of "failed to connect, retrying in 30 seconds" lines in the miner.

nice ...

#crysx

oops ...

that was my mistake ... i didnt change that also on dnr ... doing that now ... will be a small stratum / connection interruption ...

tanx ...

#crysx

we are currently working on another idea we have ...

let see what we can do with it ...

#crysx

just a couple of things with your parameters ...

'My launch config is:
cwigm -c sigt -u shubaduba.100 -p d=0.1 --lodiff --max-temp=72 -d 1,4,5,6,7 --cpu-priority 3 --cpu-affinity 0xC -i 25 -s 3'

lets a few easy issues sorted ...

-p is your password ... this being an mpos pool and custom CWI-Stratum - the vardiff settings are always enforced and a static setting like what you ave there will not work ... the d=0.1 will work on yiimp and yiimp clones because thats the way tpruvot / epsylon3 has coded it ... our stratum 'could' support static difficulties - but a bit of a rewrite is required in order to support that ... for now just replace the d=0.1 with your actual worker password ...

--lodiff is fine ...

--max-temp is an experimental feature ... this parameter pauses the processing cycles of the miner to allow the gpu to cool enough then reconnects ... these ups and downs on the power graphs is actually the PID controller putting the card into sleep for some milliseconds in order to not exceed the 72c ... remove this parameter and test again ...

--cpu-priority and --cpu-affinity are fine ...

-i 25 is also fine ... though you can mess with this figure to get the best hashrate according to your cards ... but you already know that ...

-s 3 - this is the scantime parameter which does NOT need to be here ... the miner automatically adjusts scantimes and forcing a manual change would only be benficial IF there is an issue with it ... remove this one also ...

once you make these changes - please test again and let us know how that goes ...

#crysx

How many watts does the 1070 draw at that speed according to the miner?

i'm not the dev, so he'll give you a better answer.

But i did play with DNR miner, and i noticed that it has a condition in the bat file that pauses the miner when the netdiff gets high.  When the condition is met, the gpu idles for a few blocks, no shares are submitted etc.

I took --mean-net-diff=1.1 out of the bat, and that behavior stopped.  I'm guessing that most people leave that in, so all (or most) of the miners pause for a few blocks here and there, which adds to the time it officially takes the pool ti fond a block, which screws up with the already wacky luck calculation on DNR.

I have a feeling that if everyone changed their bat file, the luck would improve drastically.

Profitability today is 10s times more than electricity cost (at least in most of the areas).
Mine 1070 run at 56-57Mh/s without no tweaking more than juast move the slider in Afterburner and no overheating. The difference covers all additional costs. So what's the point of that?