Topic: [CLOSING DOWN] ScryptGuild Auto-Switching Pool - page 32. (Read 102241 times)

1% is good.

If you use whatever hashrate you get with minimal rejects as your baseline, you can determine where your "sweet" spot is.

So let's say you gain 5% hash rate by increasing your intensity, but you also get 5% more rejects, you aren't getting anywhere.
But if you get 5% more hash rate with 2% more rejects, then you're better.

Personally I don't have the time to play with it.  I go for minimal rejects and hardware errors.  I believe op said it's impossible to have less than 2% rejects here.  When I told him I was get 0.7% (0.007), he thought I had room to get more hashrate.  That may be true, but I think it would be minimal.  Because the hashrates I get match what I see on the litecoin hardware comparison page, I tend to think all is well.

ymmv.

M

aaaah ok thought there might have been more too it. 1.069% I suppose is good? Way better than what I was getting...

Sooo another noob question ...how would you work out how much the rejects cost you to know if you should push the intensity higher or lower? 0% rejects might not necessarily be as profitable as having a higher reject rate but get more accepted shares through. Is that what eleuthria was saying?

Thanks for the infos

uhm, i have the same question...

I would think that solo mining or being on a pool that doesn't coin switch would fare better with high intensities. Depends how often the block changes.

Reject % = rejects / (accepted + rejects) * 100.

M

Ok brilliant i'll sift back down the pages and have a read ...but yep makes sense! Thanks for sharing

Presumably on these types of pools then, a less powerful rig will actually outperform one that has a far higher hash rate in terms of BTC payouts. Conversely if you were solo mining or on a pool like middlecoin/wafflePool that switches less often, a higher hash rig will be better. Is that a fair assumption?

Out of interest ...how do you calculate the reject %? Are there a bunch of things you need to factor?

Cheers

There are good number of posts earlier here explaining this.  I'll try to summarize:

the pool is designed for rapid coin switching based on profitability
the higher the intensity, the more your GPU will be unavailable to process coin switching "block has changed" messages
therefore your hashrate may appear higher with higher intensities, but you'll get more rejects

You're job is the find the sweet spot.  I had mine at I=13 and had 0.7% rejects.

Op has some good posts in here explaining it.

M

why are u mining doge as we speak, when LTC mining is more profitable than DOGE ?

erm so setting my gpu's to 13 means I hash really low so what is the point of having semi powerful cards like the 7950s? Maybe im missing something blindingly obvious?

As a medium test I dropped them all down to 16 just to see and my rejects were really high which based on what your saying is expected!

After 15.5hrs mining...
3,136.41 KH/s    A: 2,278,976    R: 22,720 / 0 / 0 / 0    I: 16

On other pools I hash at 19 and I'm pretty sure my rejects are super low so I guess this leads to my follow up question of ...is the "intensity" value also reflective of what pool or coin your a mining compared to just the cards themselves?

Wondering why the LTC comparison chart has i=20 for the 7950s ...I assumed it was specifically to do only with the hardware. So guessing their is a pool setup factor also involved?

Cheers

Here are more accurate BTC/MH/day stats on my rigs for the last days (most recent last):

0.00728 0.00446 0.00571 0.00522 0.00448 0.00482

We're pretty much on par and above average compared to other pools on poolpicker.eu

Thank you for making an awesome pool.

Your reject rate should not be higher if you have it set higher.  Your rejects will hurt more, but you will have proportionally less of them in the long run.  The pool sends you work once.  That work is the same whether you're diff1 or diff65536.  All difficulty means is you won't bother submitting results that aren't high enough to match d ifficulty.

You don't have to set it.  In most cases all works well letting vardiff kick in.

You're getting higher rejects at 256 because of your intensity.  Lower it to 13 and you should be fine.

The way it should work, assuming everything is optimal, no bandwidth limitations on either side, and no processing limitations on either side...

the lower the share diff, the more bandwidth that will be consumed because they are sent more often, and the more "regular" your valid share submissions will be
the higher the share diff, the less bandwidth that will be consumed, and the less "regular" your valid share submissions will be, leading to higher variance

At your current hash power, you should be able to process difficulty 2048 shares just fine.  You just won't get them very often, leading to high variance.  But since your intensity it too high, your GPU is busy longer trying to process shares.  When it's busy, it's busy, it can not be interrupted.  Because it's so busy, it ends up missing the "block changed" message that comes very often from this pool, leading to stale shares being submitted.  (I think I stated that right.)

In rough mathematical terms:

                                     variance
share difficulty ==  -----------------------
                             bandwidth required

so variance is directly proportional to share difficulty, and inversely proportional to bandwidth used.  (there's probably a better visual formula for this)

Another thing to keep in mind the server side.  It can't handle feeding everyone share difficulty 1 shares.  That's why vardiff is enforced.

M

Maybe im missing something ...if the difficulty is variable and the pool monitors the worker and sets accordingly, why do you set something? I have a 3.5MH rig (if i=17 is correct - seems stable on this) ...if I set to 256 it generates a tonne of rejects! 128 seems ok so far.

So just to clarify... the difficulty in this sense is really just a guage on how often the pool needs to send work. Too low a kh then more works arrives than can be processed resulting in rejects, too high a kh and not enough work arrives so a lower than normal Accepted shares? ...and all this is governed by the WorkUnit value which is the "actual" hash rate? Am I in the ball park for how this works?

Cheers

Those shouldn't impact mining at all.  Ignore them.

In regards to your point above, My Rig is 2.5  Mh/s and i selected starting difficulty to 128. Vardiff has not changed it what so ever. Generally on other pools, Vardiff sets it to around 256 or 512

Depends on your hash rate.  My suggestion is to leave it alone and check your miners to see what the vardiff switched it to.  Then set your miners to that value.

M

Yep that did it ...I forgot I copied from the LTC wiki i=20 ...my old settings i=17 gives me 0% rejects (so far)   ...Thanks!

Any chance you know what the Difficulty should be set to on the Dashboard interface for the miners?

Cheers

Whats the expiry & Scantime settings are you guys using for this pool?

Most rejects are caused by your intensity being set too high.  Try lowering it.  

I was averaging 0.7% across 4 rigs when I was mining here.

M

my average rejects on my 3 rigs with 6x GTX750TI-OC is 1-2%, so i think it's all about settings.

I'm running at C 1250, mV 1178, MC 5800 (small overlocking) per card