Topic: [OLD] Eligius: ASIC, no registration, no fee CPPSRB BTC + 105% PPS NMC, 877 # - page 51. (Read 458370 times)

Following up since I responded before checking all messages on IRC.

Seems a few people have done some tests on their own using bfgminer, with a mod to work around the bug with the latest jupiter's i2c issue.

Results were pretty solid that cgminer has to be the cause of the misreported HW errors.  My 'out in left field without looking at the code closely' guess is that somewhere along the line is it losing work it generated and when it goes to do the recheck it is checking using the wrong work, thus misinterpreting it as a HW error and consequently not submitting the should-be-valid share to the pool, resulting in a lower hashrate.  This is not a problem with Eligius or any pool and is a problem with the stock software.

Also, contrary to the crazed reports all over, this issue is not consistent, not with Eligius or any other pool.  So, enough with trying to blame Eligius for this.

I'll try to see if Luke-Jr can push a fix for the latest bfgminer so that people can switch to bfgminer to fix the issue on the latest jupiters.

-wk

KNC has even said that pool choice effecting HW error rate "does not make sense".

So, it definitely has to be a software (cgminer) problem somewhere, likely with something very subtle.

I would suggest anyone having this issue upgrade to the latest bfgminer and use that instead.  The KNC driver for bfgminer was written independently of cgminer and reportedly does not have any issues with any pools. (Latest version)

-wk

I tried a November Jupiter again this morning at Eligius, and it still has many hardware errors and pool rejects.

I also have an October unit mining at Eligius that is working perfectly.

Slush has the same effect. Way less HW errors and higher hashrate.

Hi

Just tried switching from Elgius to BTCGuild and HW errors went down from 12% to 2% and has been stable at 2% over the past 15 hours.

This only affects the faster running Nov Jupiter @ 650GH/s+

Gonna try slush's next as I'm not comfortable with the size of either BTCG or GHash.io

You can get bfgminer if you install Bertmod 0.4

I wish the Knc firmware came with cgminer and bfgminer.

I suspect bfgminer will not have the same issue with November Jupiters.

sorry eligius i like this pool
but 10% hw is a lot
some people like me will change pool

Eligius has a large buffer of namecoins built up from earlier merged mining, likely from people who haven't set a NMC address. It can afford to pay you extra NMC at the expense of the NMC buffer getting smaller.

Also, how is it 105% PPS NMC? Are you like....giving us NMC for free?

Hello! What is the namecoin payout threshold? I dont get any namecoins for a few days now...

Following up my earlier post about KNC...

Still not able to duplicate any issues on any devices I have ssh'd into.  Several folks have run their own tests with varying results which I'm looking into.

Thanks very much, I can see it has gone into the queue now.
I did adjust my min.payment downwards but that was back when the difficulty change happened and I am sure I have had a payment go through since then. Strange. Perhaps my memory is playing tricks with me. TVM anyway.

Looks like you had a minimum payout set, and then adjusted it downward.  Its not instantly effective.  I went ahead and pushed it through manually.  Check again in about 10 minutes (caching).

-wk

Is the payout queue working OK at present?
I have been above my minimum payout for two blocks now (more than 4 hours), but the stats is still saying :-
"Approximately 0.00000000 BTC remaining to enter payout queue. Maintaining your 3 hour hashrate average, this will take at least another a few seconds at current network difficulty of 908350862.44".

1DNqNN7TaauTSnaxigXNx11nntDBxLwS7n

Since my last post I've ssh'd into three different recently shipped KNC jupiters (thank you everyone) and ran many tests.

I tested against Eligius.  I also tested against btcguild and bitminter.  No noticeable deviation in HW error counts.

For more in depth Eligius testing:

I tried Eligius's getwork (disabling stratum redirect), and no increase in HW error counts.

I tried increasing the size of the coinbase transaction the KNC device used for stratum mining (this made old avalon firmwares have problems which is why Eligius limits to 128 outputs now) to 500 outputs.  No change.

I tried locking work difficulty at difficulty 1 to increase load.  No change.

I tried locking work difficulty at 1024, 2048.  No change.


Highest I saw cgminer CPU usage at was about 25%.

So far 3 out of 3 devices I've tested have functioned flawlessly, so, I don't know what is going on here (if anything).

-wk

I have to correct myself.  I did a test where I pointed cgminer-tune at only eligius (not load balanced with btcguild), and in that configuration, I also see the 10% lower WU (and the 10% HW which I agree and understand are not real hardware errors, but are reported as hardware errors for some reason in cgminer).

It seems that by sending a portion of my hashing to btcguild, the problem was being masked.

i think all people use eligius and me ofcourse are happy

but november batch knc not work well here

Hi Miners,

OK, so I need to definitely address recent posts, but I'm not going to go through and individually quote them.

First, as mentioned before, pool choice can no effect hardware errors.  Its not possible.  Hardware errors are when the hardware returns an incorrect result.  I will try and get some more experts in the field to chime in on this because obviously no one believes me on this and the FUD continues.  I don't care about screenshots or other such nonsense, as there are a hundred other things that do effect hardware error rates that also have nothing whatsoever to do with the pool.

As for difficulty adjusting constantly/periodically/etc... this is normal.  It does not effect your hash rate, reject %, etc in any way.  I could go in to detail about this again, but, I will not because I've done so many times in the past. 

In any case, *if* KNC devices are for some reason actually having some issue with Eligius, it is *not* because it is making it get more hardware errors because as explained, that is impossible.  It is most likely some issue with cgminer.  If anyone knowledgeable would care to chime in on how exactly a pool can cause hardware errors, I'm all ears.  (Not to sound like a broken record, but, I won't actually hear anything because its not actually possible.)

As for Eligius "encountering lots of problems recently", that is also nonsense.  Take a look at the pool-wide hash rate graph.  We've been pretty much rock solid for weeks.  Before that there were a few DoS attacks that did not last very long anyway.

I've added more hardware to the back end and have been making improvements constantly to be able to handle the influx of new users smoothly.  Eligius handles well over 100,000 connections continuously submitting without issues and processes over 200,000 diff 1 equivalent shares per second, all while retaining the ability to pay via the coinbase transaction with near-realtime balance/amount updates.  No other pool does this, probably because it is quite a bit of work to perfect.

Overall, most everyone who mines here is happy with the pool.  If there were really some widespread issue with the pool causing problems with certain miners, I'm sure I would know about it and would have reported it here.

I am, however, working on investigating the reports from a small number of KNC users.  I'd be willing to bet that this can be chalked up to some random issue with cgminer, but in the interest of thoughroghness I am in fact investigating in detail.  So far the people I have spoken with on IRC willing to run several tests for me show no issues or discrepancies between Eligius and other pools.

Thanks, and happy mining!

-wk

is time eligius put server on europe