Topic: [OLD] Eligius: ASIC, no registration, no fee CPPSRB BTC + 105% PPS NMC, 877 # - page 50. (Read 458370 times)

Will pop on IRC can test on dev

Excellent!  Thanks for the rapid response, I love your pool!

The 128/256 second hashrates rely on the reward system, and as noted in my last post, it went into failsafe mode earlier.  They should be correct in about an hour.  The 22.5 minute and longer time period hashrates should always be correct.

So with the help of many volunteers I believe I have found the cause of the issues with KNC miners running cgminer.

cgminer seems to be "forgetting" about work if new work is not given often.  Eligius uses what I believe was an originally defined standard of 55 seconds between stratum work updates.  Work is valid for 120 seconds.  However, when cgminer doesnt get a work update every 30 seconds or so, it occasionally seems to forget the work it gave to some hashing cores and this manifests as false hardware errors.

So far three affected KNC users have confirmed that changing the work update interval to 30 seconds fixes the false HW error issue and gives a boost in hash rate.  As for tracking down the actual cgminer bug I will try to work with conman and kano on it, and hopefully keep it civil.

I will be updating the pool to do stratum work updates every 30 seconds instead of 55 seconds like it does now.  Anyone else having KNC HW error issues willing to test this before its on the live server (dev server, still get full credit for work) shoot me a PM on IRC.

I'm unsure if this issue affects non-KNC devices running cgminer as my efforts in tracking this down have been focused on KNC.

---

Side note, Bitminter beat us by about a second on a block earlier and it ended up throwing a failsafe.  Stats and payout queue are catching up. 

Basically, I'm going to *try* to dedicate most of this weekend to Eligius improvements.  So, more to come.

-wk

This issue may have been gone over already, but as this is currently at like a billion pages, I'm hoping someone might be able to help me out.

I just recently added 2 Antminer S1 units (~400 Gh/s), but my 128 and 256 second hash rates don't seem to be showing my new hash speed correctly.  As I've had my new machines up and running for nearly 3 hours, the 3 hour average actually appears to be correct.  Is this a problem other people have had?  Or should I be doing something to resolve this?

Thanks,
bmoconno

At this point I can't definitively say what I have noticed is true, but it really does look like that is the case. Also, this seems to be a more recent thing. This device has been mining for a few months now, and in the past I would typically only have a hard crash once every two weeks at most. In the past 4 days, it has happened 5 times.

I keep a pretty close eye on the ambient temps and they really haven't changed. I have a Arduino with a temp sensor next to the device and can log the temps. Is their a easy way to log BFGMiner output to a file?

From my understanding they are days away from getting the flag working.

EDIT:(20 minutes after last reboot)
Just had it crash again, I wonder if this thing is just thinking about giving up the ghost or what?  Here is a screenshot of what it does when it crashes.(from a month ago, but it's the same thing)

Changing the diff does not change the operation of your device at all.  It still generates shares exactly the same way it was already doing.  The difficulty simply determines which shares actually get submitted to the pool.  If the difficulty is 64, then only shares with a difficulty over 64 are submitted.  If the diff is 128, then only generated shares that just happen to have a difficulty of 128 or higher are submitted.

Changing difficulty does not make your hardware work harder, or less hard, or change anything at all about the hardware operation - it just defines which generated shares actually get submitted to the pool by the miner software.

PS - The last time I mined on Eligius, the '--request-diff' flag was not honored by the pool.  I don't know if that is still true or not.

So for example, when it's finished, if I set the '--request-diff' flag to '64' for example, it will still give a automatic value? What is this flag really accomplishing?

Just a FYI, '64' diff. seems to be the safe limit for my device, when it starts using '128' the temp runs higher and then it is more likely to completely crash, requiring a cold boot.

Ok, then I don't understand what you're asking for. 

The my payments are always 0.2xxxxxxx.  The only case where I would receive a dust payment is if I stop mining altogether at eligius right after I had received a payment, then eventually I would get my balance (no matter how small and dusty) as a payment.

Are you asking for payments to only ever be in clean multiples of some payment threshold?  So 0.20000000 but never 0.21738261?  I guess I can see the point if when you spend bitcoins, you only ever spend them in nice round numbers.  That hasn't been typical for my purchases (via BitPay), because of exchange rates, the merchants are often asking me to send them a payment of 0.162534 or whatever, and so even if my wallet was initially full of nice clean inputs, they wouldn't stay clean for long.

A simple minimum payout does not avoid the problem of receiving dust.

Was wondering if there was any update on the Hashbuster Micros.  I placed an order and have not received confirmation or any other information.  Thanks

minimum payout can be configured here (I have mine set to 0.2):

http://eligius.st/~wizkid057/newstats/mystats.php?cmd=options

coinbase payouts do not matter to me...  I would rather have regular-sized payouts that reduce dust in the wallet (more friendly to the network).

i.e. payout at 0.1 BTC threshold.

Any chance of having that option?

Implementing that would have the side effect of not needing to put a set of addresses in the coinbase.

thats it

people only want kncminers works good here

Difficulty does not affect devices at all.
Even if you could set difficulty, it is unlikely to allow setting it lower than the automatic variable difficulty (which is lower than most pools use on Eligius).

I didn't read the KNC issues in detail. But one issue I seem to be having lately is if the 'pool difficulty' is high for my device, it seems more likely to crash(Temperature runs higher). Which requires a full power down of the device(BFL 60GH).  I'm looking forward to the set-able difficulty, so I can do some testing.

Greetings miners,

So, on to some minor fixes/changes.

Since I dropped the minimum payout amount down to 40 TBC (~0.04 BTC) the coinbase/generation transaction has been reaching the set maximum outputs almost every block.  This means that the balance ends up in the pool's cold wallet and needs to be paid manually later (which I have been doing).

In an effort to address this without increasing the size of the coinbase transaction (which can cause problems with miners that have low memory or slow hosts) I've modified the way the coinbase transaction is created slightly with regard to the payout queue.

Here is how it worked previously:

• Sort addresses in queue by balance age, oldest to newest
• Pay addresses until either the full coinbase/reward amount was consumed or 128 outputs were paid
• If less than the full reward were consumed, and less than 2000 TBC remain, send the balance to the offline "change" address.
• If more, pay 2000 TBC to the offline change address, balance to the main offline wallet

Now it works like this, changes highlighted in green:

• Sort addresses in queue by balance age, oldest to newest
• Pay addresses until either the full coinbase/reward amount was consumed or 100 outputs were paid
• If less than the full reward were consumed, re-sort the addresses in queue, highest balance to lowest balance
• Pay addresses until either the full coinbase/reward amount was consumed or 28 outputs were paid
• If less than the full reward were consumed, and less than 2000 TBC remain, send the balance to the offline "change" address.
• If more, pay 2000 TBC to the offline change address, balance to the main offline wallet

This has the effect of better filling the coinbase/generated payout transaction with more miners that are due payouts by effectively allowing miners with the largest balances due to jump ahead in the queue, only when needed, to better pay the full block reward directly to miners, instead of to a pool offline wallet for manual payment.

Summary: The payout queue should be paid out automatically the majority of the time and faster than it has been with less manual payouts needed.  More fresh coins for miners.

Happy mining!

-wk

Well, there are in fact people using this issue with KNC devices to pull people away from Eligius for whatever

That's the thing, the KNC devices do work well with it, as there are quite a few mining Eligius, including many November ship jupiters.  There are only a handful of people actually experiencing this issue.

Its definitely not related to the pool and I'm sure you're not actually getting better long-term hash rates elsewhere.  The issue is with the hardware/software.  Any illusion otherwise just means you either didn't test long enough, just got lucky/unlucky, or your pool is lying about or miscalculating your hash rate.  (Eligius bases this directly on shares accepted, as it should.)  If your Nov. jupiter gets high HW errors on Eligius, then based on my tests (of over a dozen jupiters now, thank you very much for the support folks) it does everywhere else also.  There's no difference in the data sent from Eligius than there is from any other pool aside from the obvious (who gets paid for found blocks).

It's been confirmed that the same issue that affected some Avalon users a while back with Eligius regarding a large coinbase transaction does not affect KNC units as they have tons of CPU headroom. (I tried with a coinbase transaction 50x larger than Eligius's allowed max and KNC devices didn't even flinch.)

Spoke with someone from KNC and they say the issue is likely related to the DC to DC converter temperature, which would explain why the issue is intermittent.

-wk

I agree. I don't think the problem is with the Eligius pool, but unfortunately the pool ends up taking a hit when the fastest mining rigs on the market don't work well with it using default, non hacked, firmware. The reason November Jupiters don't work well with the pool is kind of irrelevant.

I love the pool and will be back when Knc issues a firmware fix.

@wizkid

Thanks for your endeavor and your explanations. I think you misunderstood something, and I think, I speak for the other KnC Nov rig owners also, no one blamed or wanted to blame you or the pool. We just stated that there is a problem, but did not claim that it's a problem of the pool itself.

Eligius is a fine pool. Keep up your good work and, again, thanks for your work.