RX580 Advice Needed For Lost Hashs

[kenbeannet]

Hello Gents, I am about 3 weeks into mining and trying to get everything running smooth.  I have 18 x RX580 XFX GTS cards.  Running latest Windows 10 Professional.

1 is Samsung Memory the other 17 are Hyniax.  
I have OC'ed each with UberMix Default 3.1 timings.  

https://i.imgur.com/rqedrSH.png.

I manually OC'ed each cards one at a time, starting a low values, and went up until they didnt crash for 2 hours straight.  
Each card is a bit different but mostly are around the 1160/835 and 2125/835 (Memory Clock).  
Target Temp around 60C using OverDriveNTool 0.2.3.
https://i.imgur.com/vE4e2jo.png

I am using the Crimson Beta Blockchain latest driver from Aug 23rd. This defaults everything to compute mode in AMD Settings.

I am using Claymore 10.2 Dual mode mining with ETH and DCR.  On Ethermine.org I send my reported hash rate to get an estimate of where I should be.

https://i.imgur.com/RokohsF.png

My issue is how crazy off my estimate to my reported rate.  I understand there is stale shares, and that I will never get pure 100% of the output.  But I am about 45Mh/s off my reported hash and very rarely do I ever go over.  I computed the difficultly with my outcome for the last 24 hours and the "Average Effective Hashrate" is near accurate of about 0.054ETH per day.

My claymore does have a high incorrect ETH share rate.  I took off all the OC and ran for 24 hours (much lower Mh/s) but found that I did not get any incorrect eth shares then.

https://i.imgur.com/a5V6f2n.png

I tried changing -dcri to increase DCR and lower ETH rate to help with Incorrect, but that didnt seem to help at all. (Went as high as -dcri 45).  

Is this normal?  I feel like 45Mh/s is way off.  Even if I do incur the 2% dev fee, that would be ~500Mh/s to 490Mh/s and there is still 35Mh/s missing.  If I missed anything, and more information is missing, I will add it, would love to get this solid.

P.S. I also tried ethminer just as a test, I get about 3Mh/s per card less.  

Thanks,

[MagicSmoker]

Hello Gents, I am about 3 weeks into mining and trying to get everything running smooth.
...
My issue is how crazy off my estimate to my reported rate.  I understand there is stale shares, and that I will never get pure 100% of the output.  But I am about 45Mh/s off my reported hash and very rarely do I ever go over.  I computed the difficultly with my outcome for the last 24 hours and the "Average Effective Hashrate" is near accurate of about 0.054ETH per day.
...
P.S. I also tried ethminer just as a test, I get about 3Mh/s per card less.  

I also just started mining, but am proceeding at a much more moderate pace than you (18 cards already? Ambitious!) but I already went through the same puzzlement with Claymore and finally decided that the program sucks. Ethminer might *appear* to give a lower hashrate, but what it reports and what I actually get mining UBQ or MUSIC (on Pool Sexy) are very close, and there is no devfee for Ethminer.

I don't mind paying a devfee for demonstrably better hashrate, but lying to me about the hashrate and/or submitting stale shares to artificially boost it isn't cool. I'm not saying Claymore is doing either of those things, just that I earn more coin per hour with the same rig and settings running Ethminer 0.12.0 than Claymore 10.2.

[kenbeannet]

Hello Gents, I am about 3 weeks into mining and trying to get everything running smooth.
...
My issue is how crazy off my estimate to my reported rate.  I understand there is stale shares, and that I will never get pure 100% of the output.  But I am about 45Mh/s off my reported hash and very rarely do I ever go over.  I computed the difficultly with my outcome for the last 24 hours and the "Average Effective Hashrate" is near accurate of about 0.054ETH per day.
...
P.S. I also tried ethminer just as a test, I get about 3Mh/s per card less.  

I also just started mining, but am proceeding at a much more moderate pace than you (18 cards already? Ambitious!) but I already went through the same puzzlement with Claymore and finally decided that the program sucks. Ethminer might *appear* to give a lower hashrate, but what it reports and what I actually get mining UBQ or MUSIC (on Pool Sexy) are very close, and there is no devfee for Ethminer.

I don't mind paying a devfee for demonstrably better hashrate, but lying to me about the hashrate and/or submitting stale shares to artificially boost it isn't cool. I'm not saying Claymore is doing either of those things, just that I earn more coin per hour with the same rig and settings running Ethminer 0.12.0 than Claymore 10.2.

Do remember the ability of Claymore dual mining capabilities.  If the answer to my post is simply Claymore lies, than I can understand that and move on.  But I dont think I can run Ethminer as the way I look at it is Eth is 100% profit and DCR is the cover of the electric bill (Assuming I have 100% ROI on hardware).  From what I hear almost everyone uses Claymore though, and I havent heard many people complain about it lying about the hash.  But yet I do see people claim to get about 30Mh/s with a RX580 (I get 29.9~ when I dont have DCRI on).  But I don't see myself getting 30Mh/s with ethminer (26-27Mh/s).  I should look around and see if anyone was able to get that rate with ethminer with an RX580.  Then again, with Claymore I am really only getting 27Mh/s output anyway.

[MagicSmoker]

Do remember the ability of Claymore dual mining capabilities.

Meh, dual mining increases the power drain of each card considerably, and all to mine a shitcoin like Sia or Pascal or Decred. So your electricity bill is higher and your cards run hotter; a tradeoff I don't find compelling, but I admit the proverbial devil is in the details here.

But yet I do see people claim to get about 30Mh/s with a RX580 (I get 29.9~ when I dont have DCRI on).  But I don't see myself getting 30Mh/s with ethminer (26-27Mh/s).  I should look around and see if anyone was able to get that rate with ethminer with an RX580.  Then again, with Claymore I am really only getting 27Mh/s output anyway.

I tested Claymore again and it is, indeed, faster than Ethminer, even factoring in the devfee. However, it's a rather modest difference, with Claymore delivering 27.0 (not correcting for devfee) vs. 25.4 for Ethminer (on an RX 570). Unfortunately, to generate the devfee Claymore has to disconnect from the pool I am using, connect to his devfee pool, then generate the DAG for Ethereum, because I was mining Ubiq or Music, then mine Ethereum for 36 seconds, then switch back to my pool, generate the DAG for my coin, and finally resume mining for me.

One other thing - pushing memclock too high first results in more bad shares (from memory errors - use HWInfo to check) then leads to outright crashes; you may have just gotten an especially unlucky array of cards in the silicon lottery if they all have a lot of errors (and, really, more than 1 error per polling period is too much, I feel; dial back the OC if that is happening).

[kenbeannet]

When I checked with HWInfo, I have about 10 errors on 24 hours of running, which I believe is OK?

I am going to lower the MClock down then and give it a try over night.  I dont see much of an ETH performance difference going from 2100 to 2025 for example.

[MagicSmoker]

Well, before I got started mining I would have said that *any* memory errors would be too much, but it does seem that many like to play things a little closer to the bleeding edge. Still, you should see about a 1 MH/s difference for every 100 MHz change in memclock (from what I've noticed, not something I know for sure), so if dialing back 75 MHz doesn't reduce hashrate then that suggests the extra clock rate was only wasting power, at best, and maybe causing memory errors leading to stale/rejected shares and too many of those will get you locked out of some pools.

[kenbeannet]

So after just around 12 hours, lowering the MClock didn't seem to help.  I went from 2100 to 2000 and I am still producing errors.

https://i.imgur.com/ICn2o6c.png

[CryptoWatcher420]

a whole 14 rejected shares on dcr is quite good after 12 hrs and none on eth

[MagicSmoker]

Hmmm, I thought the OP was talking about memory errors - something reinforced by reporting what HWInfo said - and not merely rejected shares reported by the miner. Memory errors do result in rejected shares, but rejected shares are not just caused by memory errors!

[kenbeannet]

More to the point I guess, the Incorrect Shares seems a bit alarming for 12 hours.  It says in Claymore don't overclock, but at this point I am barely OC'ed as it is.  Default for the card 1405/2000 and I am at 1160/2000 yet I am still getting incorrect shares which is probably why I am lower than the suggested.

[MagicSmoker]

Your always going to get some incorrect/invalid/rejected shares if for no other reason than latency (aka "ping time") between you and the pool. In fact, this is the main reason why you should try to use pools with low ping times, which are generally, but not always, the ones closest to you.

[kenbeannet]

Your always going to get some incorrect/invalid/rejected shares if for no other reason than latency (aka "ping time") between you and the pool. In fact, this is the main reason why you should try to use pools with low ping times, which are generally, but not always, the ones closest to you.

In my opening post, I mentioned this.  But I think were being blinded by the truth of Claymore or maybe not?  As of recently, I just started using Gateless Gate Sharp (You can find his post on the this forum).  I am parsing at 29.23Mh/s on average per card, and on 4 hours of work, I am yielding 171~ Accepts per hour.  Compared to my average of about 139~ per hour on Claymore.  This is obviously a small sample size (only 4 hours).  But the proof may be starting to show itself with examples of other applications claiming a lower Mh/s but yet produce higher accepted results than Claymore.

Again, I guess only time will tell.  But I do think I am on to something.

[MagicSmoker]

I have Gateless Gate (the previous iteration of GGS) by only used it for Neoscrypt coins (TZC and FTC), where it performed very well (and devfee is optional). If GGS improves on GG enough to offset the devfee and has generates less invalid shares then I'm all for it.

The only halfway plausible way of evaluating miner speed - that I know of, anyway - is to see how many coins you earned over a 24 hour period and compare it to what you should have gotten based on the average difficulty for the last 24 hours on the minethecoin.com calculators. Otherwise you are trying to calculate a continuously moving target.

[kenbeannet]

I have Gateless Gate (the previous iteration of GGS) by only used it for Neoscrypt coins (TZC and FTC), where it performed very well (and devfee is optional). If GGS improves on GG enough to offset the devfee and has generates less invalid shares then I'm all for it.

The only halfway plausible way of evaluating miner speed - that I know of, anyway - is to see how many coins you earned over a 24 hour period and compare it to what you should have gotten based on the average difficulty for the last 24 hours on the minethecoin.com calculators. Otherwise you are trying to calculate a continuously moving target.

Agreed.  But the average hash rate changes for Etherum about 1-3%.  Which means if I allow for a 10% (way over-estimation) then I should still be within the variation of the moving target.

https://i.imgur.com/cJEQXHk.png

Proof of the last 4 hours.  Lucky the last hour I submitted a 191 while the other 2 running Claymore are still way under performing.

[MagicSmoker]

...
Proof of the last 4 hours.  Lucky the last hour I submitted a 191 while the other 2 running Claymore are still way under performing.

Interesting... was/is Claymore reporting any rejected shares to the console? Ie - "Total Shares: 287, Rejected: 3" or the like?

[kenbeannet]

...
Proof of the last 4 hours.  Lucky the last hour I submitted a 191 while the other 2 running Claymore are still way under performing.

Interesting... was/is Claymore reporting any rejected shares to the console? Ie - "Total Shares: 287, Rejected: 3" or the like?

I rarely get a rejected ETH.  Stale, yes, about 3-4%.  Rejected Decred is common.

[kenbeannet]

Wanted to do an update.  Since moving off Claymore, the only change is that I am not dual mining but solo mining.  The reported hash rate of both bots is identical around 175Mh/s per system.

https://i.imgur.com/84CJI2F.png

After only about 12 hours, my average (24 hour) hash rate has gone from 450 to 475Mhs and doesn't look like its even'ed out yet.  I would imaging my reported hash of 525 will be around 490~ when done and with the 1% dev fee its nearly accurate. 

I also just read a post about someone else claiming that Claymore takes between 5-6% of fees.  Starting to believe.