[NEW POOL & MINER] - BitcoinPool.com - Jump In! ~NO FEES~ :) - page 36.

geebus

sr. member

Activity: 258

Merit: 250

Quote from: jgarzik on March 08, 2011, 11:02:12 PM

This efficiency number is a bit misleading, IMO.

There is no guarantee that a solution exists in a 'getwork' data unit.

Better to rename "efficiency" as "luck." It is more clear.

It's "efficiency" in the sense that it's the percentage of how close you are to a 1:1 ratio of getworks asked to answers submitted. It's more "efficient" in regards to a pool because it causes less traffic on the server, and allows for more users to connect and participate, thus increasing the potential for blocks to be solved at a faster rate.

The miner itself yields results at the same speed as the core version of m0mchill's miner, but with a greatly reduced network load.

ef·fi·cien·cy [ih-fish-uhn-see]
–noun, plural -cies.

1) the ratio of the work done or energy developed by a machine, engine, etc., to the energy supplied to it, usually expressed as a percentage.

By dictionary definition, our miner fork is more efficient if you're looking at the getworks as "energy" and answers as "work done".

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: nelisky on March 08, 2011, 11:58:52 PM

Well, it is worth what it is worth. I'm now using your miner exclusively, so there you go

Don't let my word be taken the wrong way, you've done an amazing work. Not only did you get the miner to search the full key space, which is the way I like it, you also give out very helpful information while mining, something that was really lacking. Kudos to you, and btc's too once this miner renders me some

Thank you very much!

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: jgarzik on March 08, 2011, 11:02:12 PM

There is no guarantee that a solution exists in a 'getwork' data unit.

True. Then again, some getwork will have 2, 3, 4, 5, 6, or more answers. The most I've seen is 7 answers being found in 1 getwork request. The more time you give to a sample period, the closer you get to a 1:1 ratio of [shares submitted]:[getwork request].

Quote

Better to rename "efficiency" as "luck." It is more clear.

No, cause the whole process of mining is based around "luck".

When we talk about "efficiency", we're talking about how much bandwidth and resources the miner is causing the server to use. Our goal was to reduce the amount of used resources to a minimum while making sure the client isn't negatively impacted.

Sorry for the misunderstanding or if we were not clear enough.

nelisky

legendary

Activity: 1540

Merit: 1002

Quote from: FairUser on March 08, 2011, 11:51:07 PM

Quote

The 17% vs 90% description is really misleading!

I don't believe so. Perhaps our description of what we mean by "efficiency" was a bit misunderstood or lacked a sufficient explanation. Sorry for the misunderstanding.

Well, it is worth what it is worth. I'm now using your miner exclusively, so there you go

Don't let my word be taken the wrong way, you've done an amazing work. Not only did you get the miner to search the full key space, which is the way I like it, you also give out very helpful information while mining, something that was really lacking. Kudos to you, and btc's too once this miner renders me some

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: nelisky on March 08, 2011, 11:00:20 PM

Quote from: geebus on March 08, 2011, 10:31:29 PM

10 Getworks, 9 Answers = 90% Efficiency.

For a good basis of comparison, the original poclbm sits at around 17% efficiency on my cards, with a 10s askrate.

While I've always felt more comfortable with iterating the full getwork, to the point of having done my own opencl miner integrated on bitcoind just for that, the mathematics you are describing are really a flawed kind of statistical proof. Yes, you do get a much better coverage of the getwork (which is what you call efficiency) and the original miner just throws away the rest of the work once a solution is found, so it is never as "efficient". But that doesn't mean it is slower or worst, as we are basically trying to find a solution on a random pool of numbers.

You are correct, it is not slower or faster in finding shares. Our focus was simply to reduce the amount of resources (bandwidth and CPU time) on the server. That's what we're focused on; reducing the amount of bandwidth and resources used by and for the server.

Quote

So you count the number of solutions vs the number of value pools, whereas the original miner counts the number of value pools with at least one solution.
In the end I would argue the success rate will be larger for your miner because the network lag is smaller, as the miner doesn't request a new work set for each solution found, but that's about it.

I don't think that our modified miner finds more shares or blocks. All the tests I've done show about the same number of shares. We're hoping that by having more miner's that can get the same number of shares in a given time and reduces the load on the server, we'll be able to support more users. We're also hoping that more users means more blocks solved.

Quote

The 17% vs 90% description is really misleading!

I don't believe so. Perhaps our description of what we mean by "efficiency" was a bit misunderstood or lacked a sufficient explanation. Sorry for the misunderstanding.

jgarzik

legendary

Activity: 1596

Merit: 1100

This efficiency number is a bit misleading, IMO.

There is no guarantee that a solution exists in a 'getwork' data unit.

Better to rename "efficiency" as "luck." It is more clear.

nelisky

legendary

Activity: 1540

Merit: 1002

Quote from: geebus on March 08, 2011, 10:31:29 PM

10 Getworks, 9 Answers = 90% Efficiency.

For a good basis of comparison, the original poclbm sits at around 17% efficiency on my cards, with a 10s askrate.

While I've always felt more comfortable with iterating the full getwork, to the point of having done my own opencl miner integrated on bitcoind just for that, the mathematics you are describing are really a flawed kind of statistical proof. Yes, you do get a much better coverage of the getwork (which is what you call efficiency) and the original miner just throws away the rest of the work once a solution is found, so it is never as "efficient". But that doesn't mean it is slower or worst, as we are basically trying to find a solution on a random pool of numbers.

So you count the number of solutions vs the number of value pools, whereas the original miner counts the number of value pools with at least one solution. In the end I would argue the success rate will be larger for your miner because the network lag is smaller, as the miner doesn't request a new work set for each solution found, but that's about it. The 17% vs 90% description is really misleading!

geebus

sr. member

Activity: 258

Merit: 250

Quote from: qed on March 08, 2011, 10:22:26 PM

I'm using your miner and seems like ti's working well. I have 3 cards and i get around 90% effic3ency. What are the reason for Submitted Shares / Total Getworks to be 90%?

Thanks.

There can be more than one answer inside of a single getwork. A 90% efficiency means you've submitted 9 answers for every 10 getworks requested from the server.

It doesn't exactly mean that 9 getworks had answers, just that 9 answers have been submitted, it could have been any number of combinations.
For instance:

Getwork 1 - 5 resulted in no answers
Getwork 6 - 8 had 3 answers each
Getwork 9 & 10 resulted in no answers.

10 Getworks, 9 Answers = 90% Efficiency.

For a good basis of comparison, the original poclbm sits at around 17% efficiency on my cards, with a 10s askrate.

qed

full member

Activity: 196

Merit: 100

I'm using your miner and seems like ti's working well. I have 3 cards and i get around 90% effic3ency. What are the reason for Submitted Shares / Total Getworks to be 90%?

Thanks.

h00ters

newbie

Activity: 22

Merit: 0

sweeeeeeeet!!! new pool that does not force donations.

+10BTC gents for the new pool!

new client is good. getting about 90% efficiently!

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: nster on March 08, 2011, 09:47:12 PM

oh I use -v -w 128 -f 0 or -f 10 for my GPU...

And I do not think ufasoft's miner works, at least not for me... what SSE2 miners work?

Honestly I'm not sure. After spending $45 on 64 CPU's from Amazon's EC2 cloud for only 24 hours, and getting almost nothing in return, I abandoned all the CPU's miner....personally speaking. I spent the money, got a couple GPU's, and the entire time I was testing the pool I used GPU's. Geebus may have tested some CPUs.

So any and all feedback from the CPU's miners out there is a BIG help for us.
Thank you in advance.

geebus

sr. member

Activity: 258

Merit: 250

We tested the following CPU miners:

jgarzik's CPU miner
puddinpop's CPU miner
poclbm-mod running on CPU

Aside from that, we don't know.

nster

full member

Activity: 126

Merit: 100

Quote from: FairUser on March 08, 2011, 08:57:45 PM

Quote from: nster on March 08, 2011, 08:20:54 PM

Quote from: geebus on March 08, 2011, 08:15:53 PM

Current Round Stats are back up on the index page. We had to resolve an issue with slow response and timeouts on the database queries. All is well now.

For now I have the highest efficiency Cheesy

and I'm 3rd in shares if we don't count the geebuses and the goldys else I'm 6th

all that with a non-constant 270Mh/s

Also, can this be run with ufasoft's miner? I didn't get it to work. and the poclbm gets me only 3Mh/s (vs 20) AND lowers my GPU mining by 25Mh/s

I honestly haven't tried ufasoft's miner, so I'm not qualified to answer that question.

Also, you might want to have the "-v" flag added to the poclbm-mod.exe so it will attempt to use vectors. I noticed a speed improvement on all three of my boxes when I used vectors.

If you figure out what's going on, please post back here. I'm @ work right now and can't really test anything at the moment.

oh I use -v -w 128 -f 0 or -f 10 for my GPU...

And I do not think ufasoft's miner works, at least not for me... what SSE2 miners work?

geebus

sr. member

Activity: 258

Merit: 250

We're manually locking some user accounts that are excessively flooding our servers. See the list below:

burtyb requested 74653 (More than 2 per second) getworks and submitted 125 answers.

yomi requested 45050 (more than 2 per second) getworks and submitted 169 answers.

thistle requested 30075 (more than 1 per second) getworks and submitted 37 answers.

If any of the listed users want to email me, we can discuss adjustment of your askrate to more efficiently work with the pool and re-activate your accounts.

Shoot me a PM with the following info:

Your Askrate
# of CPUs/GPUs on the account
What miner you're using

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: nster on March 08, 2011, 08:20:54 PM

Quote from: geebus on March 08, 2011, 08:15:53 PM

Current Round Stats are back up on the index page. We had to resolve an issue with slow response and timeouts on the database queries. All is well now.

For now I have the highest efficiency Cheesy

and I'm 3rd in shares if we don't count the geebuses and the goldys else I'm 6th

all that with a non-constant 270Mh/s

Also, can this be run with ufasoft's miner? I didn't get it to work. and the poclbm gets me only 3Mh/s (vs 20) AND lowers my GPU mining by 25Mh/s

I honestly haven't tried ufasoft's miner, so I'm not qualified to answer that question.

Also, you might want to have the "-v" flag added to the poclbm-mod.exe so it will attempt to use vectors. I noticed a speed improvement on all three of my boxes when I used vectors.

If you figure out what's going on, please post back here. I'm @ work right now and can't really test anything at the moment.

nster

full member

Activity: 126

Merit: 100

Quote from: geebus on March 08, 2011, 08:15:53 PM

Current Round Stats are back up on the index page. We had to resolve an issue with slow response and timeouts on the database queries. All is well now.

For now I have the highest efficiency Cheesy

and I'm 3rd in shares if we don't count the geebuses and the goldys else I'm 6th

all that with a non-constant 270Mh/s

Also, can this be run with ufasoft's miner? I didn't get it to work. and the poclbm gets me only 3Mh/s (vs 20) AND lowers my GPU mining by 25Mh/s

grue

legendary

Activity: 2058

Merit: 1462

Quote from: nster on March 08, 2011, 07:10:14 PM

Have you tried ufasoft? It could be that it doesn't work, but if a CPU miner works with this pool, most likely ufasoft's will work

that is the one im currently using. i'm not sure what version it is, but the properties window is showing a version number of 5.0.1822.0, and the CRC32 of the .exe is 83CFD1E0.

also, im using this:
bitcoin-miner.exe -t 4 -u grue -p █████████ -o http://bitcoinpool.com:8334
as the command line parameter

geebus

sr. member

Activity: 258

Merit: 250

Current Round Stats are back up on the index page. We had to resolve an issue with slow response and timeouts on the database queries. All is well now.

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: kseistrup on March 08, 2011, 05:14:20 PM

If people with CPU miners are unsure if they should join this pool — i.e., if they can live up to 1% efficiency — here are some numbers from the fastest of my CPU miners (and you will see it's pretty slow):

I'm running jgarzik's cpuminer single-threaded with the sse2_64 algorithm on a Core2Duo E6550 @ 2.33 GHz and a scantime of 7 seconds, which gives it a speed of 2,452 ± 272 khash/s (p=.95, based on 56,533 getworks).

Now, out of those 56,533 getworks the miner has delivered 224 POWs. Ladies and gentlemen, that's a whopping efficiency of 0.396% (and so this miner doesn't qualify)!

That said, I do like the share-based payout system…

Cheers,

OK, we made a change. Auto-banning is turned off for now. CPU miners are safe.

We will no longer look at the 1% Efficiency as a means of banning people.
Instead, if you miner make more than 30 getwork request AND has not submitted a share in 60 seconds, you'll be temporarily banned for 5 minutes. If your miner is getting a new getwork every 5 to 10 seconds AND has not submitted a share in 60 seconds, you'll be OK.

We are trying to prevent getwork flooding or as some have called it the "fire hose" method for trying to find shares, because it's horribly inefficient and ultimately cuts the CPU miners short. Working on your getwork for more than 5 or 10 seconds is preferred, but you probably don't want to work on the same getwork for more than the current average time it take to find a block (based on a average of the last 100 or 1000 blocks).

Ahh, gotta love the CPU's too...

FairUser

sr. member

Activity: 1344

Merit: 264

bit.ly/3QXp3oh | Ultimate Launchpad on TON

Quote from: nster on March 08, 2011, 07:19:37 PM

I think I got banned from bitcoinpool.com (and the miner does work either) because I refreshed the page too many times

Could you unban me please Embarrassed

PM me your account name on the pool.
Wait, I see a "nster" account and it does not have the banned flag set. You should be able to mine with no problems.

Please keep in mind we are working on the site during this time too. So if a page says it has an error or something, that'll be temporary and you should try to refresh in a couple of minutes.

Topic: [NEW POOL & MINER] - BitcoinPool.com - Jump In! ~NO FEES~ :) - page 36. (Read 101960 times)