[1050 TH] BitMinter.com [1% PPLNS,Pays TxFees +MergedMining,Stratum,GBT,vardiff]

[DrHaribo]

Some server changes are in. CPU usage reduced by 20-25%. The server is now also using Luke-jr's trick of sending pre-computed "empty" blocks (only transaction is the generation transaction) through long poll. So now the load spikes from abusive miners are greatly reduced and long poll is faster. This results in much fewer rejects than we've seen in the last few days and the delay you would sometimes see with getting work after long poll seems essentially gone.

At the moment for the current round I have 0.04% rejects. Pool wide rejects are 0.43% though. I would have hoped to get this lower. Let me know if you still have a high reject ratio and I'll see what I can do.

Namecoin rejects are sometimes very high. I suspect some miners are configured to ignore namecoin long polls. Not a big deal I guess, with namecoin value being what it is.

Note: every round starts with rejected work. If you see 10% rejected in the first second of a round, that's normal. Round data is more useful to look at after it has gone over, say, 30 minutes.

[1714314187]

1714314187

[1714314187]

1714314187

[1714314187]

1714314187

[QuantumFoam]

kano, yea that thread is huge and I haven't been keeping up with it for several months. My bad. Thanks for the link! It'll be easy to keep cgm updated now.

[Shermo]

My efficiency seems to be significantly increased since your changes DrHaribo On my 6870 machine its gone from about 280% to 450%, rejects seem to be better too!

[narousberg]

My efficiency seems to be significantly increased since your changes DrHaribo On my 6870 machine its gone from about 280% to 450%, rejects seem to be better too!

yes, less rejects too (ztex quads)

[Shermo]

Seems the site and pool server went down briefly and now seems really sluggish...

[DrHaribo]

A bit of russian DDoS attack on the web server currently taking place. Sorry for the instability. Mining looks fine so far, though.

[Shermo]

Looking at the stats my home machines using cgminer seem to be fine. I only noticed because my work PC went quiet all of a sudden, it is using the BitMinter Java client with the port 80 switch... so I guess that doesn't use long polling as it keeps running out of work?

[EDIT]
Work PC seems to be OK now actually, it was working and then dropping but it seems to be going steady again now

[DrHaribo]

Looking at the stats my home machines using cgminer seem to be fine. I only noticed because my work PC went quiet all of a sudden, it is using the BitMinter Java client with the port 80 switch... so I guess that doesn't use long polling as it keeps running out of work?

Sure it uses long polling, or it would cause a lot of rejects. But it goes through the web server. Not recommended unless you really need to use port 80 because that's all your firewall allows.

[DrHaribo]

Small server update today:

• Increase shift size (PPLNS with N=4x difficulty). This should reduce variance for part-time mining and if you have some downtime on your rigs.
• Implemented getwork "midstate" extension. Most miners support this now. It reduces bandwidth a little and also CPU usage on the server.
• Small speedup for long poll
• Small speedup in checking proofs of work

Only shaved a few percent off the server CPU usage this time, but every bit helps.

EDIT: most browsers  of course I meant most miners support this now. Also, if your livestats don't show shift progression correctly (100% when only half done), try clearing your browser cache and reloading the page.

[Digigami]

sweet

[QuantumFoam]

Noticed my hash rate has increased by about 5Mh/s per gpu ever since the recent server changes.

[Shermo]

The pool seems to be running really sweetly with all the recent changes, my efficiency is over 500% now on my single GPU system, and my rejects are very very low compared to before Great work DrHaribo!

[loshia]

The pool seems to be running really sweetly with all the recent changes, my efficiency is over 500% now on my single GPU system, and my rejects are very very low compared to before Great work DrHaribo!

+1
You are the man DOC:)

[Mobius]

When will you implement var diff for miners?

[ralree]

• 170,100 shares accepted since last cgminer restart
• 139 rejects

That's 0.08%!  Cheers!

[QuantumFoam]

(5s):2480.2 (avg):2486.0 Mh/s | Q:7925  A:88271  R:112  HW:0  E:1114%  U:34.7/m

Pretty sweet

[DrHaribo]

Great to hear the new server changes are working out

Yes, variable difficulty will be coming soon. I'm currently working on some changes to support multiple mining servers. I thought that might be a good idea to do first as it will make it easier to test var diff.

It could also be useful to have a couple more servers when ASICs turn up.

[jamesg]

Great to hear the new server changes are working out

Yes, variable difficulty will be coming soon. I'm currently working on some changes to support multiple mining servers. I thought that might be a good idea to do first as it will make it easier to test var diff.

It could also be useful to have a couple more servers when ASICs turn up.

Are you sure you would even need extra servers with variable diff? You already have rollNTime support so the most requests coming to the pool server are share submissions. With a variable diff solution that uses X-Mining-Hashrate you could greatly reduce the pools current load. You could also fall back to the server calculating the var diff to keep share submission below a maximum threshold of say 24 shares per minute.

I guess my main point is that IMHO it is a higher priority to have variable difficulty working with either GBT or stratum than to have multiple pool servers. You might never need an extra pool server with these other pieces in place.

[Krak]

Are you sure you would even need extra servers with variable diff? You already have rollNTime support so the most requests coming to the pool server are share submissions. With a variable diff solution that uses X-Mining-Hashrate you could greatly reduce the pools current load. You could also fall back to the server calculating the var diff to keep share submission below a maximum threshold of say 24 shares per minute.

I guess my main point is that IMHO it is a higher priority to have variable difficulty working with either GBT or stratum than to have multiple pool servers. You might never need an extra pool server with these other pieces in place.

Yeah, but a US server would be nice. I can easily hit < 0.01% rejects on MaxBTC just because my ping to their server is ~30ms.

[-ck]

Are you sure you would even need extra servers with variable diff? You already have rollNTime support so the most requests coming to the pool server are share submissions. With a variable diff solution that uses X-Mining-Hashrate you could greatly reduce the pools current load. You could also fall back to the server calculating the var diff to keep share submission below a maximum threshold of say 24 shares per minute.

I guess my main point is that IMHO it is a higher priority to have variable difficulty working with either GBT or stratum than to have multiple pool servers. You might never need an extra pool server with these other pieces in place.

Yeah, but a US server would be nice. I can easily hit < 0.01% rejects on MaxBTC just because my ping to their server is ~30ms.

The ping time factor is overrated. I've tried different servers from the same pool located at 250ms and 50ms ping times and the reject rate was identical. On the other hand, changes to the pool setup itself cause larger changes to the reject rate.