[4+ EH] Slush Pool (slushpool.com); Overt AsicBoost; World First Mining Pool

[marcus_of_augustus]

Hi slush,

I am wondering if you had considered increasing the difficulty for your pool shares?, i.e. making it >1

It would reduce getwork calls across the internet, for the higher hash rate cards it seems to make more sense to be working on a harder problem, maybe a separate pool for bigger cards? They spend quite a bit of time waiting for the new work, taken in clock cycles versus wall-clock ...

(also I'm a bit biased because I seem to have a speedy core that likes to run hot, fast and not stop ... it hangs when it ramps up after a getwork 1 in 1000 times??)

[1714178145]

1714178145

[1714178145]

1714178145

[1714178145]

1714178145

[slush]

I am wondering if you had considered increasing the difficulty for your pool shares?, i.e. making it >1

It would reduce getwork calls across the internet, for the higher hash rate cards it seems to make more sense to be working on a harder problem, maybe a separate pool for bigger cards? They spend quite a bit of time waiting for the new work, taken in clock cycles versus wall-clock ...

Well, with well written miner, there is no reason why cores should wait on submitting shares; it can be done fully in asynchronous way (just send, don't wait - at least in blocking state - to response). Maybe the feature request to the miner developers? Blocking getworks are another issue, but I'm working on it - its long polling and next stuff comming soon.

The higher difficulty for fast cards are good idea, it is already on my list for some time. It will definitely come, I have some algorithms to find 'idea' difficulty for given worker in my head; will see how it will work in real life . Keep in mind that higher difficulty does not mean only lower network overhead, but also higher variance in round rewards and maybe also some new kind of pool attacks, so it need some care to don't break anything.

[Dude65535]

(also I'm a bit biased because I seem to have a speedy core that likes to run hot, fast and not stop ... it hangs when it ramps up after a getwork 1 in 1000 times??)

I have a fairly crappy net connection, so to deal with it I run one copy of poclbm with a -f of 30-60 connected to the pool and a second copy with a -f of 200+ solo mining. The second copy gets about 5% of the gpu power when both are running but it picks up the slack immediately if there is a delay in the get work.

Edit: I have a satellite internet connection plus I use a VPN to avoid the 250MB a day download cap, but that puts me last in line for getting any bandwidth. About the only way my connection could get worse is if i used Tor as well.

[marcus_of_augustus]

I am wondering if you had considered increasing the difficulty for your pool shares?, i.e. making it >1

It would reduce getwork calls across the internet, for the higher hash rate cards it seems to make more sense to be working on a harder problem, maybe a separate pool for bigger cards? They spend quite a bit of time waiting for the new work, taken in clock cycles versus wall-clock ...

Well, with well written miner, there is no reason why cores should wait on submitting shares; it can be done fully in asynchronous way (just send, don't wait - at least in blocking state - to response). Maybe the feature request to the miner developers? Blocking getworks are another issue, but I'm working on it - its long polling and next stuff comming soon.

The higher difficulty for fast cards are good idea, it is already on my list for some time. It will definitely come, I have some algorithms to find 'idea' difficulty for given worker in my head; will see how it will work in real life . Keep in mind that higher difficulty does not mean only lower network overhead, but also higher variance in round rewards and maybe also some new kind of pool attacks, so it need some care to don't break anything.

Anybody know if the poclbm.py miner is asynchronous? (I'll ask m0mchill also I guess) It is apparent on screen that hash rate drops after a block is accepted and new one is begun but is that what is happening on the core?

I wasn't going to bother with overhead of solo-mining set-up just yet but maybe it is way to work around buggy fast hot core ... or maybe put two processes on same core so it is always doing something regardless of getwork calls ....hmmm, thnx for the ideas.

[slush]

Anybody know if the poclbm.py miner is asynchronous? (I'll ask m0mchill also I guess) It is apparent on screen that hash rate drops after a block is accepted and new one is begun but is that what is happening on the core?

Currently the poclbm does not drop job when share is found. Afaik the whole network stuff is in separate thread, so it should work 'asynchronous'. But I didn't do any debugging, maybe it is really handing somewhere. m0mchil performs new getwork before whole nonce space is crunched, so unless you have really crappy line, it should work fine.

Btw ongoing improvements in the pool will make network load waay lower than it is now. And I'm not talking only about long polling .

[marcus_of_augustus]

(also I'm a bit biased because I seem to have a speedy core that likes to run hot, fast and not stop ... it hangs when it ramps up after a getwork 1 in 1000 times??)

I have a fairly crappy net connection, so to deal with it I run one copy of poclbm with a -f of 30-60 connected to the pool and a second copy with a -f of 200+ solo mining. The second copy gets about 5% of the gpu power when both are running but it picks up the slack immediately if there is a delay in the get work.

Edit: I have a satellite internet connection plus I use a VPN to avoid the 250MB a day download cap, but that puts me last in line for getting any bandwidth. About the only way my connection could get worse is if i used Tor as well.

Hey Dude, thnx for your idea. I launched two processes on the problem GPU and it is now stable for over 6 hours (at best it was 1 hour before). If anyone is interested here is my convoluted launch process to get this thing crunching ...
in one terminal
$export DISPLAY=:0.(speedy GPU adapter number)
$fgl_glxgears
wait for some output to indicate fps are being crunched in gears then in another terminal (should have these commands ready to go with terminals and shells prepped)
$nice -20 ./poclbm.py -u miner0_username --pass=miner0_passwd -o mining.bitcoin.cz -p 8332 -v -w128 -f10 -d(speedy GPU)
and in yet another terminal
$nice -20 ./poclbm.py -u miner1_username --pass=miner1_passwd -o mining.bitcoin.cz -p 8332 -v -w128 -f10 -d(speedy GPU)
and as soon as processes report back hashing numbers kill (CTRL+C) fgl_glxgears.

Dirty hack I know but it works and is now stable, I can do a little script to launch this with a timer included. It was worth it because this core actually hashes 1-2MH/s quicker than other cores and wasn't worth sending back 5970 card for protracted warranty claim as it passed all the graphics tests anyway.

I think it is a problem with Xserver, (Xorg) fglrx kernel module and poclbm and a hardware weirdness whereby GPU doesn't like stopping and ramping up after a getwork pause but is fine if it is kept crunching maxed out .... maybe multiple processes can optimise GPUs in other ways ... -w 32 x 4?

[cosix]

im getting some stales (1 per 20 valid blocks). should i set my askrate lower? right now, its 5 (poclbm default)

[slush]

im getting some stales (1 per 20 valid blocks). should i set my askrate lower? right now, its 5 (poclbm default)

No, leave it as is. I had to restart bitcoin daemons on the server, so the stale shares were probably caused by short outage.

[Groc]

Join ur on http://mining.bitcoin.cz!
EDIT 27.12.2010: wiki page about pooled mining

If you have a slower computer,... then pooled mining may be the only way that you will ever mint any bitcoins at all.

How do I get started?
You need less than 10 minutes to start mining in pool. Visit http://mining.bitcoin.cz and follow instructions.

I fit in to this boat but for the life of me can't figure out how to use these "miner programs" to contribute to mining.bitcoin.cz
 

p.s. I'm currently not generating coins. I use windows xp. I have an IBP p4 think center. 

[snedie]

Join ur on http://mining.bitcoin.cz!
EDIT 27.12.2010: wiki page about pooled mining

If you have a slower computer,... then pooled mining may be the only way that you will ever mint any bitcoins at all.

How do I get started?
You need less than 10 minutes to start mining in pool. Visit http://mining.bitcoin.cz and follow instructions.

I fit in to this boat but for the life of me can't figure out how to use these "miner programs" to contribute to mining.bitcoin.cz
 

p.s. I'm currently not generating coins. I use windows xp. I have an IBP p4 think center. 

Pretty much wasting your time to put in simply.

But if you want to do it anyway, download the poclbm GUI miner which will come in a .7z archive. Extract this to a folder of your choice, then run the guiminer.exe file. Go to File>New Miner and give it a name. Then enter mining.bitcoin.cz for the server address.

Go to mining.bitcoin.cz and register an account, and then make a miner (It's under the "My Account" page, just click on "Add new miner".

Now back in poclbm under the new miner you just made, enter the username and password for your miner. This can be found under the my account page on mining.bitcoin.cz.

The go to File>Save Settings, and then start the miner. It will say in the bottom right how many Hash/s you are doing.

[snedie]

Slush,

Do I actually need a seperate user entry for each miner? I just noticed that I accidentaly gave two of my miners the same username to use and it simply added both their shares together...

[slush]

I fit in to this boat but for the life of me can't figure out how to use these "miner programs" to contribute to mining.bitcoin.cz
 

Please select one from supported miners (in your case, it will be probably jgarzik's CPU miner) and follow instructions in his forum thread. The link is on the pool homepage. If you will have any troubles with that, don't hesitate to ask more specific directly in that thread.

[slush]

Do I actually need a seperate user entry for each miner? I just noticed that I accidentaly gave two of my miners the same username to use and it simply added both their shares together...

There are some technical reasons why it isn't good idea. In some cases your miners might duplicate the effort (and from two same submitted shares, only one will be accounted). Simply - don't do that, really .

[Lardie]

Correct me if I'm wrong, but if someone uses my worker's username they just generate bitcoins for me. So what is the purpose of the password?

[slush]

Correct me if I'm wrong, but if someone uses my worker's username they just generate bitcoins for me. So what is the purpose of the password?

Well, it is little complicated under the hood. Server must remember jobs which served you, to be able to perform some checks on share submitting. As the server has limited memory for every worker jobs, with knowledge of user/password, the attacker can make your mining impossible by erasing your worker queue. Well, it is nothing to too much worry about, but posting your full worker credentials on the web isn't generally good idea.

[demonofelru]

Hey Slush have you switched payout to only daily or something?  Usually it goes out hourly but have been over the payout minimum for a few hours now.

[datathe1st]

Hey Slush.  Would you consider charging only a 1% commission out of the goodness of your heart?

[snedie]

Do I actually need a seperate user entry for each miner? I just noticed that I accidentaly gave two of my miners the same username to use and it simply added both their shares together...

There are some technical reasons why it isn't good idea. In some cases your miners might duplicate the effort (and from two same submitted shares, only one will be accounted). Simply - don't do that, really .

Thanks bud, I knew there would be a good reason not to do it. 

[slush]

Hey Slush have you switched payout to only daily or something?  Usually it goes out hourly but have been over the payout minimum for a few hours now.

Thanks for reporting; I found that one wallet had insufficient funds for paying tx fees. It is fixed now, payment should be processed withing one hour.

[dishwara]

Active workers (at least one share in last hour): 0

For past one hour or more..