[1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool

[kano]

Meanwhile regarding your rant full of inaccurate information ... if you spent the money of your electricity bill, that you using mining on the nVidia card, to buy BTC, you'd get more BTC and also help increase the price of BTC ... and reduce Carbon output at the same time

But if I didn't mess around with these miners, and learn all about this stuff, and just purchased the BTC,
 i would be like ever other noob who hasn't a clue of how and why bitcoin is a better option than our current Federal Reserve System.
(corrupt officials can't exactly counterfeit BTC)

Like I said I am not here to mine, just to learn about the whole process, and what exactly is being "mined"

I disagree that my COMMENT was complete inaccurate, but my time and electrons are worth more than to argue with you over something that I already know.
There are no good cuda miners for windows, and the rpcminer was last updated nov of 2011.

Well I guess CoinLabs must be hiding the CUDA miner that ckolivas wrote for them months ago ... ?
(I don't know I don't have anything to do with CoinLabs, I just know that ckolivas wrote a proprietary CUDA miner for them some months ago)

...

... and your previous post started with

I have the only cuda miner in existance. (rpcminer)

then

So this means that there are a ton of nvidia users out there with outdated miners (because they have been abandoned)

... anyway ... cgminer compiles on all sorts of platforms - even windows

[Clock Loop]

Ya know what.. dude...  since you seem to imply that there are tons of options for cuda miners....

Link me to some windows CUDA miners that don't require me to switch to linux or switch pools.

Otherwise, kindly f-off.

[Clock Loop]

Oh, and since were on the topic of cgminer.....   BEGIN RANT.

Someone f-ed up big time when programming/designing the cgminer to bitforce interface.


Its funny, because I happen to find a program that was transmitting on all FTDI serial ports every 10 seconds at 115200 baud.

It turns out that one of the "miners" called cgminer has code in it that wrecklessly opens ALL FTDI ports and transmits on them.
Its a program that is trying to find a device called "bitforce". (its a fpga hashing cpu) to mine digital hashes.
The problem is, that it opens ports to other devices that use the FTDI chip ALSO, and transmits ZXG every 10 seconds on ALL FTDI ports at 115200 baud.
From a bitforce perspective, they don't give a shit, they just want their product to work.
From a 3rd party hardware engineer, that is really really mean to do to other hardware developers that develop using the FTDI CHIP.
 (tons of microcontroller programmers and devices use it)

I know no one will change the code to be more "serial port" and other device friendly, and I also know you won't mention it to the developer.
But it really IS bad practice to open ALL ports every 10 seconds, not just open them, but actaully transmitt, and operate the DTR lines(many devices use the DTR to operate the reset line on the microcontroller/chip.

https://github.com/ckolivas/cgminer/...ver-bitforce.c
This code is included in almost all "miners" and so most computers that run a miner program most likely have this FTDI port access every 10 seconds
It will transmit "ZGX" at 115200 baud every 10 seconds on all FTDI ports.

Why a simple switch couldn't have been implemented to turn the bitforce port attacking ON, instead of having it run 24/7, till it finds a bitforce device, beats me.

Most of you have no idea that your cgminers are doing this to every serial port on your computer, every 10 seconds, because most of you don't do hardware development.
But if a hardware developer were to see this activity from a program, that hardware developer would have a few CHOICE words to use, and they wouldn't be pretty.

[kano]

Ya know what.. dude...  since you seem to imply that there are tons of options for cuda miners....

...

Quote me where I said that?

Then re-read my last post.

[Clock Loop]

Im good, lets just end this.

[kano]

Oh, and since were on the topic of cgminer.....   BEGIN RANT.

Someone f-ed up big time when programming/designing the cgminer to bitforce interface.


Its funny, because I happen to find a program that was transmitting on all FTDI serial ports every 10 seconds at 115200 baud.

It turns out that one of the "miners" called cgminer has code in it that wrecklessly opens ALL FTDI ports and transmits on them.
Its a program that is trying to find a device called "bitforce". (its a fpga hashing cpu) to mine digital hashes.
The problem is, that it opens ports to other devices that use the FTDI chip ALSO, and transmits ZXG every 10 seconds on ALL FTDI ports at 115200 baud.
From a bitforce perspective, they don't give a shit, they just want their product to work.
From a 3rd party hardware engineer, that is really really mean to do to other hardware developers that develop using the FTDI CHIP.
 (tons of microcontroller programmers and devices use it)

I know no one will change the code to be more "serial port" and other device friendly, and I also know you won't mention it to the developer.
But it really IS bad practice to open ALL ports every 10 seconds, not just open them, but actaully transmitt, and operate the DTR lines(many devices use the DTR to operate the reset line on the microcontroller/chip.

https://github.com/ckolivas/cgminer/...ver-bitforce.c
This code is included in almost all "miners"

WTF drugs have you been taking?

and so most computers that run a miner program most likely have this FTDI port access every 10 seconds
It will transmit "ZGX" at 115200 baud every 10 seconds on all FTDI ports.

Why a simple switch couldn't have been implemented to turn the bitforce port attacking ON, instead of having it run 24/7, till it finds a bitforce device, beats me.

Most of you have no idea that your cgminers are doing this to every serial port on your computer, every 10 seconds, because most of you don't do hardware development.
But if a hardware developer were to see this activity from a program, that hardware developer would have a few CHOICE words to use, and they wouldn't be pretty.

LULZ

Go complain to the person who wrote the miner you are complaining about.

If it is cgminer you are talking about, then it's the windows driver that you are using at fault, not cgminer (... if anything above you have said it true)
Cgminer searches for BFL devices at startup ... but never again.

When a BFL device fails, it will attempt to restart it then, but otherwise, "ZGX" isn't sent for any other reason in cgminer.

What's the next fairytale you are going to make up?

--

Aside: if I look around the forum and IRC logs I'll find my comments about using serial, at all, being crap.

Lucky for you I've rewritten the MMQ code in the past week to use libusb, once I finish off the tidy up of the resource tracking, it will be available for use for the Icarus and Bitforce driver also.

Yeah I'm one of the cgminer developers
I wrote the API, the majority of the Icarus driver, and shortly most of the MMQ driver code also.
I'll be writing the ASIC drivers in cgminer too along with ckolivas.

If you have some actual useful knowledge or suggestions come visit IRC and help
Otherwise - fuck off.

[Clock Loop]

I wouldn't want to help someone with such a negative/attacking attitude.

I don't even care if my comments are false, as you can clearly see, I am new here, and being so harsh, and attacking with new people is a bit much dude.

Even if all I said was bad, then just try to correct me, instead of blatantly attacking me (in a general way too, saying its all completely false)
 Even still, you failed to mention any solution, you say switch to linux, or use cgminer(which uses 75% of cpu also)
So exactly .....WHAT solutions do you present, if rpcminer isn't the ONLY viable cuda miner.

You say that cgminer dosen't attack all FTDI ports constantly?
Did one version of the code DO this at one point in the past?  

After looking into it, looks like the ufasoft code is the culprit here..  sorry...


So its actually ufasoft(ufasoft_0.37) that is abusing the ports..  my bad..
I thought it was cgminer, my bad.

I thought I WAS running cgminer, but I must have switched to ufasoft at some point.

I also am not much help, I do hardware/firmware development, I try to stay away from the software side of it if I can.
But if someone wants to send me a fpga/asic dev board, id be glad to help.      
WHOLY CRAP ARE High speed FPGA/ASIC's expensive. (even if you just buy 1, bga ONLY, not even on a pcb)

Stratix IV fpga =  Cheapest single chip is $1,000.   FOR ONE CHIP.
http://www.digikey.com/scripts/dksearch/dksus.dll?FV=fff40027%2Cfff80166&vendor=0&mnonly=0&newproducts=0&ptm=0&fid=0&quantity=0&PV-5=22897&PV-5=22898&PV-5=21079&stock=1

(but I do dig into software code when I MUST, to help determine if bugs are software or hardware.)

[Clock Loop]

BACK to the topic of this thread.....  P2Pool...

I have more questions...

If i set my refresh rate to 100ms, at my low has rate (150mh/s) won't I never find a share because my GPU's don't have enough time on the current block to find any shares?
Doesn't the refresh rate REQUIRE a minimum hash rate to find any share or block?
What would increasing my refresh rate to say 1000ms do?
Give me potential dead/orphan shares? (local dead on arrival)

I have tried refreshrates of 100ms, 750ms, 4000ms, 8000ms...
If I go longer refresh rates (above 1000ms) then I will get lots of dead on arrival, but I find that my share # goes up.
 Some arrive as dead, orphans, but more show up as good shares.
If I set my refresh rate to say 100ms... I find that I produce NO shares, no dead, and no orphans, in a 24hour period.
If I set it to 750ms, I find I can produce 1 share every 4-8 hours or so.
If I increase it to around 4000ms I have lots of local dead on arrival, with even more shares than a lower refresh rate.  (but more dead and orphan shares show up.

I guess I don't fully get it all, but my understanding is that if you set the refresh to a longer rate, you might produce old shares.
But having some dead/orphan shares is better than none at all.    So it seems that people with lower hash rates should use longer polling?
They may get more dead, but at least they will get something.
 I found that not all were dead, only say 10% were dead when I had a polling rate of 4000ms.
So what gives, does polling rate depend on your overall hash rate?

What I notice, is with very low polling rates(100ms or less), I produce no shares at all.   Even if I run it for two days.
But if I set it to 750ms or longer, I will produce about 1 share every 4-8 hours.
And if I set it to about 5000ms, I will produce about 1 share every 2-4 hours(with 25% of those shares ending up as dead/orphans)

So does polling rate depend on hash rate, if you want to produce any shares at all...(dead/orphan/good)

[aadje93]

BACK to the topic of this thread.....  P2Pool...

I have more questions...

If i set my refresh rate to 100ms, at my low has rate (150mh/s) won't I never find a share because my GPU's don't have enough time on the current block to find any shares?
Doesn't the refresh rate REQUIRE a minimum hash rate to find any share or block?
What would increasing my refresh rate to say 1000ms do?
Give me potential dead/orphan shares? (local dead on arrival)

I have tried refreshrates of 100ms, 750ms, 4000ms, 8000ms...
If I go longer refresh rates (above 1000ms) then I will get lots of dead on arrival, but I find that my share # goes up.
 Some arrive as dead, orphans, but more show up as good shares.
If I set my refresh rate to say 100ms... I find that I produce NO shares, no dead, and no orphans, in a 24hour period.
If I set it to 750ms, I find I can produce 1 share every 4-8 hours or so.
If I increase it to around 4000ms I have lots of local dead on arrival, with even more shares than a lower refresh rate.  (but more dead and orphan shares show up.

I guess I don't fully get it all, but my understanding is that if you set the refresh to a longer rate, you might produce old shares.
But having some dead/orphan shares is better than none at all.    So it seems that people with lower hash rates should use longer polling?
They may get more dead, but at least they will get something.
 I found that not all were dead, only say 10% were dead when I had a polling rate of 4000ms.
So what gives, does polling rate depend on your overall hash rate?

What I notice, is with very low polling rates(100ms or less), I produce no shares at all.   Even if I run it for two days.
But if I set it to 750ms or longer, I will produce about 1 share every 4-8 hours.
And if I set it to about 5000ms, I will produce about 1 share every 2-4 hours(with 25% of those shares ending up as dead/orphans)

So does polling rate depend on hash rate, if you want to produce any shares at all...(dead/orphan/good)

Hi because of your low hashrate the P2P share chain (similair to blockchain of bitcoin) has already found new shares. You can see share difficulty on the website of your miner node. With the current difficulty of writing it gives you around 4 shares each day But you could connect to a public node of P2Pool (check the p2pool servers list) and connect to there. This will "give" you 1 difficulty shares if i look correctly (i see alot of accepted on my gpu's but not so much on the miner node) and this will count your count of the shares from the miners node and give a payout to your adress (use adress as username, and no password) Then you still have some variance, but its less .

Its (if im right) some kind of proportional from the miner node wich mines at p2pool in PPLnS system 

[kano]

BACK to the topic of this thread.....  P2Pool...

I have more questions...

If i set my refresh rate to 100ms, at my low has rate (150mh/s) won't I never find a share because my GPU's don't have enough time on the current block to find any shares?
Doesn't the refresh rate REQUIRE a minimum hash rate to find any share or block?
What would increasing my refresh rate to say 1000ms do?
Give me potential dead/orphan shares? (local dead on arrival)
...

P2Pool has an average share LP of 10 seconds.

I've no idea what refresh rate is in your miner, however:

A ) if that affects how often the GPU returns success/fail results, then the lower it is the better, with an obvious exception:
as it goes lower, then the CPU will be doing more work for the same amount of GPU work (and the GPU will be more idle)

However, from a statistical view, it doesn't matter if you only hash 1,000th of a nonce range or all of it.
You will still find shares as often.

B ) What you will do however, if you are processing without break for 1000ms, is you will on average lose 500ms per share LP
i.e. when a share LP hits you will on average be half way through your 1000ms of work and thus have wasted 500ms of work (assuming of course the miner aborts current work on the share LP)
When an LP hits, any work in progress is throw away (unless it's a network block - and then only if the miner program is smart enough to send it - which hopefully it is for your miner)

Thus on average with that figure I just mentioned, 5% of your processing time is wasted on average and one in 20 1-difficulty shares will appear after they are no longer valid - thus 1 in 20 of the longer term share chain shares of much higher difficulty (that you will only find a very few each day at a low shash rate) will be discarded also.

If you are processing without break for only 100ms, then on average each LP will represent 50ms of lost work or 0.5% - so much better.

It's obviously a balance between the two A ) and B ) however.

[Clock Loop]

 This will "give" you 1 difficulty shares if i look correctly (i see alot of accepted on my gpu's but not so much on the miner node) and this will count your count of the shares from the miners node and give a payout to your adress (use adress as username, and no password) Then you still have some variance, but its less .

Can't I give p2pool an argument that forces specific difficulty shares?  
And if setting lower difficulty gets more shares, why doesn't everyone just set theirs like that?
(perhaps lower difficult shares are worth less?)  

Even if they are, at least i'd get some shares, instead of none.
In all the documentation I read, no miner documentation ever said that one should use servers with difficulty to a low number if you are hashing less than X....

I have read people suggest low hashers use server pools who serve lower difficulty.
Can't p2pool be configured to serve low difficulty?   (i understand that the miners cannot choose difficulty tho)

[Krak]

 This will "give" you 1 difficulty shares if i look correctly (i see alot of accepted on my gpu's but not so much on the miner node) and this will count your count of the shares from the miners node and give a payout to your adress (use adress as username, and no password) Then you still have some variance, but its less .

Can't I give p2pool an argument that forces specific difficulty shares?  
And if setting lower difficulty gets more shares, why doesn't everyone just set theirs like that?
(perhaps lower difficult shares are worth less?)  

Even if they are, at least i'd get some shares, instead of none.
In all the documentation I read, no miner documentation ever said that one should use servers with difficulty to a low number if you are hashing less than X....

I have read people suggest low hashers use server pools who serve lower difficulty.
Can't p2pool be configured to serve low difficulty?   (i understand that the miners cannot choose difficulty tho)

The only node that actually does this is P2PMining.com. They count the amount of difficulty-1 (or higher; they support variable difficulty) "pseudo" shares and pay out with their own PPLNS system. The reason p2pool can't do this for everybody is because there'd be thousands of shares getting transferred between all of the nodes on the network every second, causing a lot of orphans/DOAs and extremely high bandwidth usage. This is why you can't set your p2pool share difficulty lower than the current network difficulty, however some of the bigger miners volunteer to set their difficulty higher to lower variance for the rest of the network.

[sharky112065]

Ya know what.. dude...  since you seem to imply that there are tons of options for cuda miners....

Link me to some windows CUDA miners that don't require me to switch to linux or switch pools.

Otherwise, kindly f-off.

I'm not Kano, but...

Cgminer does mine on Nvidia hardware and has a Windows version supplied on each release. You can also compile your own binary by following the included "windows-build.txt" file as a guide. You can download it from one of the links in the OP at this forum thread https://bitcointalk.org/index.php?topic=28402.0 .

As far as Cgminer using  too much CPU, that seems odd to me as Cgminer is designed to use very little CPU.  Maybe you have a driver issue?

Kano is Australian, and in my opinion/experience they tend to come off a bit brash. Once you realize that and account for it, communicating with them will go much smoother. I used to play an online game with several Australians and that is how they were as well.  Remember, the Bitcoin community is made up of many nationalities/cultures.

[rav3n_pl]

Cgminer uses opencl drivers on any device that supports it, so hash rate on various devices depends mostly on opencl drivers.
Nvidia is more concentratd on cuda, so they opencl drivers are not so good as ATI ones.
Also 10s longpool is one of P2pool features needed to keep pool integrity/payouts and can not be changed w/o totally change pool rules.

[K1773R]

Ya know what.. dude...  since you seem to imply that there are tons of options for cuda miners....

Link me to some windows CUDA miners that don't require me to switch to linux or switch pools.

Otherwise, kindly f-off.

I'm not Kano, but...

Cgminer does mine on Nvidia hardware and has a Windows version supplied on each release. You can also compile your own binary by following the included "windows-build.txt" file as a guide. You can download it from one of the links in the OP at this forum thread https://bitcointalk.org/index.php?topic=28402.0 .

As far as Cgminer using  too much CPU, that seems odd to me as Cgminer is designed to use very little CPU.  Maybe you have a driver issue?

Kano is Australian, and in my opinion/experience they tend to come off a bit brash. Once you realize that and account for it, communicating with them will go much smoother. I used to play an online game with several Australians and that is how they were as well.  Remember, the Bitcoin community is made up of many nationalities/cultures.

I dont think its about Australians, i do understand him. Im from switzerland and would react the same way. Its like "dont tell me how it works, i know how it works and your talking just bullshit. dont try to teach a geek if ur a newbie".

[organofcorti]

Kano is Australian, and in my opinion/experience they tend to come off a bit brash. Once you realize that and account for it, communicating with them will go much smoother. I used to play an online game with several Australians and that is how they were as well.  Remember, the Bitcoin community is made up of many nationalities/cultures.

Brash? I'll give you brash, arsehole!

I dont think its about Australians, i do understand him. Im from switzerland and would react the same way. Its like "dont tell me how it works, i know how it works and your talking just bullshit. dont try to teach a geek if ur a newbie".

I'm Australian, and I wouldn't react the same way. I mean I'd say much the same thing but I'd make the piss-take a bit more subtle. kano's lack of subtlety does make me cringe at times. I prefer the insults I use to be slow burners.

[cabin]

 This will "give" you 1 difficulty shares if i look correctly (i see alot of accepted on my gpu's but not so much on the miner node) and this will count your count of the shares from the miners node and give a payout to your adress (use adress as username, and no password) Then you still have some variance, but its less .

Can't I give p2pool an argument that forces specific difficulty shares?  
And if setting lower difficulty gets more shares, why doesn't everyone just set theirs like that?
(perhaps lower difficult shares are worth less?)  

Even if they are, at least i'd get some shares, instead of none.
In all the documentation I read, no miner documentation ever said that one should use servers with difficulty to a low number if you are hashing less than X....

I have read people suggest low hashers use server pools who serve lower difficulty.
Can't p2pool be configured to serve low difficulty?   (i understand that the miners cannot choose difficulty tho)

You should probably concentrate on the Local Rate Graph on this page of your p2p node:
http://localhost:9332/static/graphs.html?Day

It uses the Difficulty 1 shares to judge how fast your miner is. You should use the setting that gives you the highest number of non-dead shares. As long as you are submitting lots of non-dead diff1 shares, the real shares will also come in time. Trying to experiment with the real diff700 shares is not very helpful because the variability is too great.

As a ballpark though you could expect 5% DOA with 1000ms, 10% with 2000ms etc. 5-10% orphans would also be normal on top of that depending on how well your p2pool node is connected and how smoothly your bitcoind is serving work.

[kjj]

P2Pool release 9.0 tag: 9.0 hash: 2cb4d8381e179f71ea2075cdce948ea83cf0dc55

HARDFORK: Upgrade is required! Hardfork will happen after 95% of the pool's hash rate has upgraded. Everybody having not upgraded will be split off into a tiny P2Pool.

Windows binary: http://u.forre.st/u/urxoztyg/p2pool_win32_9.0.zip
Source zipball: https://github.com/forrestv/p2pool/zipball/9.0
Source tarball: https://github.com/forrestv/p2pool/tarball/9.0

Changes:
* Transaction preforwarding Transactions that you're mining are sent to peers before you get a share, so any block solution you find can be broadcast virtually instantaneously. This could theoretically get our invalid block rate below any other pool's thanks to our large network of nodes. This was implemented in v8, but was prevented from taking effect due to a bug being discovered.
* Stratum support in the P2Pool protocol. Stratum mining will be enabled in a future patch to v9.

Any chance that you could put a list of supported block versions on the main status page?

[TravisE]

As far as Cgminer using  too much CPU, that seems odd to me as Cgminer is designed to use very little CPU.  Maybe you have a driver issue?

cgminer on my system (64-bit Linux, official Nvidia drivers) also uses close to one full core all the time it's operating (even with GPU mining only). Never knew why, unless it does have something to do with the drivers.

[kano]

Kano is Australian, and in my opinion/experience they tend to come off a bit brash. Once you realize that and account for it, communicating with them will go much smoother. I used to play an online game with several Australians and that is how they were as well.  Remember, the Bitcoin community is made up of many nationalities/cultures.

Brash? I'll give you brash, arsehole!

I dont think its about Australians, i do understand him. Im from switzerland and would react the same way. Its like "dont tell me how it works, i know how it works and your talking just bullshit. dont try to teach a geek if ur a newbie".

I'm Australian, and I wouldn't react the same way. I mean I'd say much the same thing but I'd make the piss-take a bit more subtle. kano's lack of subtlety does make me cringe at times. I prefer the insults I use to be slow burners.

Well - he did tell me to f-off first so I just returned the favour

So his comment:

I wouldn't want to help someone with such a negative/attacking attitude.

I don't even care if my comments are false, as you can clearly see, I am new here, and being so harsh, and attacking with new people is a bit much dude.

Was purely hypocritical

...

Here:

Ya know what.. dude...  since you seem to imply that there are tons of options for cuda miners....

Link me to some windows CUDA miners that don't require me to switch to linux or switch pools.

Otherwise, kindly f-off.

Edit: though I did reply with help after that anyway
i.e. cabin's post above includes info that I've already said
https://bitcointalk.org/index.php?topic=18313.msg1334217#msg1334217