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

[marks1976]

I am running a 5830 and 6950 in windows machines and hits 700M/h steady on bitminter and a bit lower on other miners.  I want to add more hash and power is a concern for cost and supply to my house.

[ZodiacDragon84]

I'm considering a small solar array to power a remote fpga cluster in the future months. You have your host computer, connected to an inverter you can easily pull from a used gas generator. have a relay that kicks to grid power when inverter cant keep up. The inverter is hooked to a small bank of batteries (marine 12 v. deep cycle) The batteries are hooked to a cheap charge controller found on ebay, which is in turn hooked to solar cells. Host computer already performs well on the network wireless with a virgin mobile cell usb stick, so one would not even need to technically have a grid connect if you had enough sun and a big enough bank of batteries. (my host is my 4890, and my ghetto solar rig was cobbled together from ebay deals.)

[marks1976]

You are running FPGA's on here with no problem?  Are you using the Bitminter software?

[P4man]

Bitminter (app) doesnt support FPGAs. Yet, and I dont even know if its planned. Its still a relatively niche market, and bitminter (the app) aims to make mining easy for the less technical, so Im not sure it makes a lot of sense for DrHaribo to invest time in supporting them at this point. Wouldnt be easy either without access to the hardware.

Of course you can use the pool with whatever miner app you want, or whats supplied with the FPGA.

[Turbor]

on all 1.4 ghash?

No, only 800 MH/s.

[DrHaribo]

I'm considering a small solar array

Wow, this is interesting stuff. People are always saying solar is too expensive and can't be done profitably. How much for this equipment on ebay? And how many watts do you get out of it?

Bitminter (app) doesnt support FPGAs. Yet, and I dont even know if its planned.

Supporting FPGAs is on my list. But I have to prioritize first what matters for large groups of users. In the beginning I made the miner the fastest on 6970/6990, but it was slow on older cards. That wasn't a good idea. I think the big masses of miners were on 58xx and probably still are. So, in short: first things first. Hmm, and "BitMinter client" is a stupid and confusing name. Maybe I should find a new one to better tell the miner apart from the rest of BitMinter.

By the way, block 401 (namecoin at height 39183, built by Fefox) could have produced a bitcoin block as well. But the proof of work was stale, probably just seconds too late. I also noticed we were up to 0.5% rejected on bitcoin proofs of work.

I am working on some things to improve this. I have prioritized long polling running on my test server. Fast miners get long poll messages first. This may seem unfair to the slowest miners, but it's for the good of the pool and all of us. When fast miners produce stale work they produce more of it than slow miners, so it makes sense to give them fresh work first. When donations are implemented I'm thinking of making that part of the prioritization as well.

I'm also making some other changes to improve long polling and speed up the server in general. Hopefully there will be a new version of the pool back end sometime in the next few days.

[P4man]

I am working on some things to improve this. I have prioritized long polling running on my test server. Fast miners get long poll messages first.

Good idea.

When donations are implemented I'm thinking of making that part of the prioritization as well.

Bad idea. Think about it, having high priority LP is no benefit to any individual miner, whether he donates or not; particularly if he has a low hashrate. Its a benefit to everyone in the pool if he has a high hashrate.

Im sure you're thinking hard of ways to give people an incentive to donate, but this isnt the one. If you would implement a bounty system for finding long blocks, perhaps you could match that with donations. For instance only donators are eligible for the bonus. Other things to consider, would be extra services, if you can implement it, like email or sms notifications for solving blocks, hashrate alerts and the like.  Or at least a JSON interface that would allow others to implement it, currently you can only get general pool stats over JSON, not individual stats.

[jake262144]

Biggest-miners-first is a tried and true algorithm beneficial for the pool and its operator.
P4Man already made his point about donations and I believe he's right.

If you choose to proceed with incorporating donations into the LP system, Doc, make sure there's a hard limit as to where donations can move any miner on the priority list.
Don't do something bone-headed that might negatively impact the whole pool (*cough*Triplemining jackpot farming*cough*).

You wouldn't like to have 150 CPU miners appear out of the blue and ruin your LP system by setting their donation to 50%, right?
They aren't gonna get any work done anyhow and your LP system is hosed.

... I also noticed we were up to 0.5% rejected on bitcoin proofs of work. I am working on some things to improve this.
... I'm also making some other changes to improve long polling and speed up the server in general.

Thanks for taking a good care of us, Doc.

[DrHaribo]

Think about it, having high priority LP is no benefit to any individual miner, whether he donates or not; particularly if he has a low hashrate. Its a benefit to everyone in the pool if he has a high hashrate.

It is a benefit to the individual miner since you don't get paid for rejected work. 0.1% more rejects means 0.1% less income. Some miners might be happy to get 10% of their 1% donation back through 0.1% less rejected work.

But yes, fast miners having reduced rejected work is of benefit to all miners in the pool in the way that we make use of a higher percentage of their hash rate, which means lower variance for everyone. I am aware that variance is still the number one challenge for this pool.

Im sure you're thinking hard of ways to give people an incentive to donate

Yep, I have a list of ideas, this is just one of them. Nothing set in stone.

You wouldn't like to have 150 CPU miners appear out of the blue and ruin your LP system by setting their donation to 50%, right?
They aren't gonna get any work done anyhow and your LP system is hosed.

Yes, I have thought of this too. I have two options in mind.

First solution: Put miners in groups determined by their donations, arrange miners within each group by hashrate. Example: First long poll messages go to those who donate 1% or more, fastest miners first. Then long poll goes to those who donate below 1%, fastest miners first.

Second solution: Donation gives a boost to your hashrate as seen by the long poll system. It would have to be multiplied up a bit to be noticable. Example: If you don't donate, your hashrate is unmodified. If you donate 1% you get a 5% boost. If you donate 2% you get a 10% boost. So a 5 MH/s CPU miner donating 100% would still only count as 25 MH/s in the long poll priority queue.

Anyway, these are just some thoughts. The goal is for the pool to be better than other pools even if you don't donate, but for the donation perks to be good enough to make it worth donating. This is a difficult balance to make and you can see many internet-based companies struggle with this sort of thing. Well, who knows, some may even donate just because they like the pool.

[jake262144]

First solution: Put miners in groups determined by their donations, arrange miners within each group by hashrate. Example: First long poll messages go to those who donate 1% or more, fastest miners first. Then long poll goes to those who donate below 1%, fastest miners first.

Translates to:

Code:

First LP goes to the CPU drones with 99% donations.
Second LP - to the drones with 98% donations.
...
The 180-th and following LPs go to the legitimate miners with medium speeds and >1% donations.
...
The 1534-th LP goes - already late - to DeathAndTaxes who's got a crapload of horsepower at 1% donations.
...

Not what you want to see, I reckon.
That's the worst-case scenario, to be sure, but that's how algorithms should be always analyzed.
If you evaluate the best-case performance or even the expected performance, you might miss a situation where the algorithm falls flat on its face when slightly abused.


The second approach seems safer. Even though it doesn't give the big miners a huge incentive to set up donations, prioritizing them regardless is the way for the whole pool to be more efficient.

[DrHaribo]

Not what you want to see, I reckon.

No, and that's why there wouldn't be a whole bunch of groups, maybe just 2 like in the example. Just a slightly faster long poll if you donate above some small limit. The difference may not be very noticeable. But I think the trick is to make people feel that life is good even if they don't donate, but if they do donate a tiny bit then they get a little bit of VIP treatment. The disadvantage with the first approach is that a donating CPU miner gets ahead of a non-donating GPU miner.

The second approach seems safer. Even though it doesn't give the big miners a huge incentive to set up donations, prioritizing them regardless is the way for the whole pool to be more efficient.

The second approach I think is nice because by adjusting the multiplier (5 in the example) it's easy to adjust how much donations will let you move up the priority ladder. So it should be possible to find a place in between "no measurable effect" and "CPU miners slow down GPU miners".

Anyway, no worries, I have no intention of turning this into anything similar to internet games the makers call "free to play" and the players call "pay to win". I'm just trying to think of some ways to make a little money in the long run, or at least stop losing money. But first, of course, I need to make an actual donation system.

[Kain]

Even without any incentive, this pool is profitable to the point where I'd willingly donate 20%, maybe more.

[Proofer]

I'm also making some other changes to improve long polling and speed up the server in general. Hopefully there will be a new version of the pool back end sometime in the next few days.

Possibly old news, but I'm seeing quite a few "not providing work fast enough" messages from cgminer.  In this instance "quite a few" means an average of one per eleven minutes over a 14-hour period, but they're bunched such that most of them were in the last couple of hours during that period.  I had a similar problem in a session early yesterday.

Code:

[2012-01-21 22:09:47] Pool 1 not providing work fast enough
[2012-01-21 22:13:07] Pool 1 not providing work fast enough
[2012-01-21 23:45:17] Pool 1 not providing work fast enough
[2012-01-22 00:48:35] Pool 1 not providing work fast enough
[2012-01-22 01:08:24] Pool 1 not providing work fast enough
[2012-01-22 01:20:07] Pool 1 not providing work fast enough
[2012-01-22 01:32:20] Pool 1 not providing work fast enough
[2012-01-22 02:14:14] Pool 1 not providing work fast enough
[2012-01-22 03:03:38] Pool 1 not providing work fast enough
[2012-01-22 03:21:46] Pool 1 not providing work fast enough
[2012-01-22 04:25:48] Pool 1 not providing work fast enough
[2012-01-22 05:30:33] Pool 1 not providing work fast enough
[2012-01-22 07:15:21] Pool 1 not providing work fast enough
[2012-01-22 07:37:20] Pool 1 not providing work fast enough
[2012-01-22 08:45:17] Pool 1 not providing work fast enough
[2012-01-22 09:09:15] Pool 1 not providing work fast enough
[2012-01-22 09:28:18] Pool 1 not providing work fast enough
[2012-01-22 09:28:46] Pool 1 not providing work fast enough
[2012-01-22 09:37:39] Pool 1 not providing work fast enough
[2012-01-22 09:39:33] Pool 1 not providing work fast enough
[2012-01-22 09:39:47] Pool 1 not providing work fast enough
[2012-01-22 09:40:02] Pool 1 not providing work fast enough
[2012-01-22 09:40:17] Pool 1 not providing work fast enough
[2012-01-22 09:40:25] Pool 1 not providing work fast enough
[2012-01-22 09:40:30] Pool 1 not providing work fast enough
[2012-01-22 09:53:28] Pool 1 not providing work fast enough
[2012-01-22 09:53:33] Pool 1 not providing work fast enough
[2012-01-22 09:54:24] Pool 1 not providing work fast enough
[2012-01-22 09:54:43] Pool 1 not providing work fast enough
[2012-01-22 09:55:21] Pool 1 not providing work fast enough
[2012-01-22 09:55:26] Pool 1 not providing work fast enough
[2012-01-22 09:56:36] Pool 1 not providing work fast enough
[2012-01-22 09:57:07] Pool 1 not providing work fast enough
[2012-01-22 09:57:11] Pool 1 not providing work fast enough
[2012-01-22 09:57:21] Pool 1 not providing work fast enough
[2012-01-22 09:57:43] Pool 1 not providing work fast enough
[2012-01-22 09:58:07] Pool 1 not providing work fast enough
[2012-01-22 09:58:11] Pool 1 not providing work fast enough
[2012-01-22 09:58:28] Pool 1 not providing work fast enough
[2012-01-22 09:58:50] Pool 1 not providing work fast enough
[2012-01-22 09:59:14] Pool 1 not providing work fast enough
[2012-01-22 09:59:21] Pool 1 not providing work fast enough
[2012-01-22 09:59:38] Pool 1 not providing work fast enough
[2012-01-22 10:00:00] Pool 1 not providing work fast enough
[2012-01-22 10:00:20] Pool 1 not providing work fast enough
[2012-01-22 10:00:28] Pool 1 not providing work fast enough
[2012-01-22 10:00:43] Pool 1 not providing work fast enough
[2012-01-22 10:01:19] Pool 1 not providing work fast enough
[2012-01-22 10:01:31] Pool 1 not providing work fast enough
[2012-01-22 10:01:53] Pool 1 not providing work fast enough
[2012-01-22 10:02:08] Pool 1 not providing work fast enough
[2012-01-22 10:02:13] Pool 1 not providing work fast enough
[2012-01-22 10:02:21] Pool 1 not providing work fast enough
[2012-01-22 10:02:36] Pool 1 not providing work fast enough
[2012-01-22 10:03:12] Pool 1 not providing work fast enough
[2012-01-22 10:03:23] Pool 1 not providing work fast enough
[2012-01-22 10:03:46] Pool 1 not providing work fast enough
[2012-01-22 10:04:02] Pool 1 not providing work fast enough
[2012-01-22 10:04:06] Pool 1 not providing work fast enough
[2012-01-22 10:04:14] Pool 1 not providing work fast enough
[2012-01-22 10:04:29] Pool 1 not providing work fast enough
[2012-01-22 10:05:05] Pool 1 not providing work fast enough
[2012-01-22 10:05:16] Pool 1 not providing work fast enough
[2012-01-22 10:05:39] Pool 1 not providing work fast enough
[2012-01-22 10:05:55] Pool 1 not providing work fast enough
[2012-01-22 10:05:59] Pool 1 not providing work fast enough
[2012-01-22 10:06:07] Pool 1 not providing work fast enough
[2012-01-22 10:06:22] Pool 1 not providing work fast enough
[2012-01-22 10:06:58] Pool 1 not providing work fast enough
[2012-01-22 10:07:09] Pool 1 not providing work fast enough
[2012-01-22 10:07:32] Pool 1 not providing work fast enough
[2012-01-22 10:07:52] Pool 1 not providing work fast enough
[2012-01-22 10:08:00] Pool 1 not providing work fast enough
[2012-01-22 10:08:15] Pool 1 not providing work fast enough
[2012-01-22 10:08:51] Pool 1 not providing work fast enough
[2012-01-22 10:09:02] Pool 1 not providing work fast enough
[2012-01-22 10:09:53] Pool 1 not providing work fast enough

[jake262144]

Update your cgminer Proofer.
There was some minor glitch in the code resulting in the "Pool not providing work fast enough" condition being falsely triggered right after having received an LP message.
I believe Conman mentioned this specific glitch lately in cgminer thread.

[ZodiacDragon84]

no fpgas on my end yet, was making a theoretical post there. solar is pulling about 1 kW Dr. Haribo, for the inverter from an old generator, enough batteries (refurbished forklift batteries) to run a day and a half no grid, and a home built 1kW solar grid made with cells sold on ebay and bussed together myself, cost me about 2k. This setup runs my 4890 for a day and a half when sun not shining. I plan on using the 4890 rig as my fpga host, and I think the company has  their own software to connect to pools of choice.

[rjk]

no fpgas on my end yet, was making a theoretical post there. solar is pulling about 1 kW Dr. Haribo, for the inverter from an old generator, enough batteries (refurbished forklift batteries) to run a day and a half no grid, and a home built 1kW solar grid made with cells sold on ebay and bussed together myself, cost me about 2k. This setup runs my 4890 for a day and a half when sun not shining. I plan on using the 4890 rig as my fpga host, and I think the company has  their own software to connect to pools of choice.

I might note that the 4890 is horribly inefficent, so you might be better off just removing it to save a few watts.

[ZodiacDragon84]

true, but off of solar power, It makes no real difference, and every little hash helps! (I occasionally have the card do other parallel compute tasks when needed)

[ZodiacDragon84]

Another loooong block. Sad face

[DrHaribo]

Argh! Did someone hook up 1000 computers this morning? The server can't keep up with all the getwork requests and we are getting very high reject ratio now.

In a meeting at work, difficult to do much right now.

[DrHaribo]

And after I wrote that the load kept increasing. At this point the server is almost completely unresponsive. Not sure if it is a DoS-attack or if a botnet is trying to mine here. But the current server can't handle either.

Sorry guys, not much I can do at this moment.