[ANN][BURST] Burst | Efficient HDD Mining | New 1.2.3 Fork block 92000

[enta2k]

lol, for example my last dadline is 707506293144
Does this tell me, that my plot file is fucked?

either that, or that the plot file has very very few nounces in it, and is super small. I've had very small plot files under generation generate crazy deadlines, but deadlines improve a lot as the file grow in size, something like perhaps the best seen deadline over some period gets half as big, with double the filesize (or something like that)

my pc with 3TB plots right now has a minimum deadline of 4721 over the last 6 hours of mining, so it's not super easy to hit the small ones.

btw. it would be overly super cool if the java wallet server could be set up to add to a file the best deadline seen for every block one is mining. Such a log would be fun to have. The current log has to be set to log *A LOT* of stuff before the wallet starts logging deadline numbers.

Haven´t seen a deadline smaller than 1 billion since i plottet my file with gpu.
Looks right gpu plotter is really really shitty and now i can start all over

[1714694048]

1714694048

[1714694048]

1714694048

[1714694048]

1714694048

[1714694048]

1714694048

[SpeedDemon13]

My plot file is 5,4TB and i plotted everything with gpu.
If this is causing a like 10mb performance, i really hope there will be a fix for the gpu plotter soon,
because, it takes almost a week to plot with cpu

You are the dude, if you say thats a very bad performance, i really should plot again with cpu,
as hard as it is, better plott forever than having like zero performance.

What gpu and driver did you use? The latest gpu plotter and AMD 14.x drivers plot fine, but older gpu plotters with AMD 14.x driver can produce incorrect plot data. If your using an older gpu plotter, then use the AMD 13.x drivers.

[enta2k]

i used the 2.0.1 mod with 14.5 drivers

Maybe it was the modded version that caused the issue.

[pinballdude]

My plot file is 5,4TB and i plotted everything with gpu.
If this is causing a like 10mb performance, i really hope there will be a fix for the gpu plotter soon,
because, it takes almost a week to plot with cpu

You are the dude, if you say thats a very bad performance, i really should plot again with cpu,
as hard as it is, better plott forever than having like zero performance.

I usually create a batch file with many calls to the generator, each generating like 128GB or some such. In that way, a bad file is not so much data, and i can just delete it and recreate. If You did that, you could cpu plot until a gpu plotter that works for you get along, and you could even plot with exactly the same specifications and probably compare the two files, or the results coming out of them, to verify the gpu plotter is working for you as it should.


right now i'm plotting a USB 1TB drive i found somewhere and it is done this way :

Code:

title  Drive:L 30000000 to 30500000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 30000000 500000 4000 4

title  Drive:L 30500000 to 31000000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 30500000 500000 4000 4

title  Drive:L 31000000 to 31500000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 31000000 500000 4000 4

title  Drive:L 31500000 to 32000000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 31500000 500000 4000 4

title  Drive:L 32000000 to 32500000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 32000000 500000 4000 4

title  Drive:L 32500000 to 33000000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 32500000 500000 4000 4

title  Drive:L 33000000 to 33500000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 33000000 500000 4000 4

title  Drive:L 33500000 to 34000000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx1200m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 33500000 500000 4000 4

title  Drive:L 34000000 to 34500000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx1200m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 34000000 500000 4000 4

title  Drive:L 34500000 to 35000000
"%PROGRAMFILES%\Java\jre7\bin\java.exe" -Xmx1200m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 34500000 500000 4000 4

means that the window always have a title what it's doing, and also means that the files show up on automatic as the script gets finished with the files.
I mine them while they are being created, that works perfectly fine

If the pc is sloggy bc of the CPU plotting, try to lower the priority of the java process that is doing your plotting to below average, that way other stuff gets preference. You can also select a number of cores that is less that what you have, to save some for other programs. I use 4 of 8 cores atm. and that's not really slowing anything down at all, but the process is also set with below average.

I should probably have gone GPU plotting, but my GPUs were already mining skynet and indocoin, and no reason to interfere with that ;-)

[SpeedDemon13]

i used the 2.0.1 mod with 14.5 drivers

Maybe it was the modded version that caused the issue.

AMD 14.5 would work ok on GPU plotter 2.1.1. Just redo the plots again.

[pinballdude]

can someone give me an example of a "too long deadline"?
I really dont know if mine are long or not

This is of course when doing solo
The system is set up to to average 4 minute blocks, so 240 seconds should be average. That means that half the time a 240 second deadline will win a block, and half the time it is too large. at 120 seconds it might be 1/4 risk you don't get a block, and at 60 it might be 1/8 risk someone is faster, etc. etc. conversely at 480 you will only win a block 1/4 of the time, at 960 only 1/8 of the time... i think.. not sure these calculations work out, but it ought to be something like that. i had one at 180 that did not win a block, and one at 114 that did win a block. Patience is pretty vital, I had to wait almost a week before i got a block, and i was cpu plotting like crazy on 3 pcs, 14 cores,  24/7 towards 4TB, so on average i must have had somewhat over 1TB plottable

Okay i have to say i don´t get it, you mean your system is giving you a result every 4 Minutes average?
Isn´t the average time set by the amount of space you are using?
I mined solo some time too and sometimes theres a result after 3 seconds and sometimes it takes ages.
Can you post a screenshot of what you mean, because i think i missunderstand it.

The thing is i plottet 6GB gpu and everyone says it´s worse than cpu plots,
i really would like to see if my average whatever time is good or bad.

It works this way - everyone mining at all finds a deadline in their plot files... the larger plot file, the more tickets they have in the plotfiles, so the better the chance of having a ticket with a really low number. (ticket is a seed that gets converted to a number, using data from the network, so basically the resulting deadline is your lucky number, which you cannot fake, and anyone else can easily verify). So every miner has a best deadline, and the very very best deadline is the time it takes before a block is generated. The miners simply wait as many seconds as is their best deadline, and if the block is still not created, they do it.

The numbers and difficulty and stuff is arranged such that on average  someone will have a deadline around 240 seconds.  If we stray away from that, the entire systen recalibrates. Still, someone might be lucky and get a real real low deadline, then a block is solved very very fast, and the next time perhaps everyne is quite with bad luck, and the smallest deadline anywhere is like 800 and we have to wait and wait until the 800 seconds passes and the lucky dude with a 800 deadline gets his block.

so the amount of space you have plotted will get you smaller deadlines with more space - simply because you have more random number tickets to begin with (more random numbers, more of them will randomly tend to be smallish) a block can be solved instantly by anyone, but the network lets whoever do it, who can prove he has the code for the lowest deadline, and these codes we have in our plot files.


I don't know all this in enough detail to really be certain about how stuff works, but above is how i think it sort of works.

[enta2k]

Hey pinnballdude thats really very usefull, i will do it like that.
Damn that makes me sad, i thought after 2 weeks of trying around now everything is fine.
I was able to plott the whole 5,4tb file in 15h, now it takes a week or i have to find a gpu plotter version that works correctly.
Didn´t knew, that it could be so bad.

I have one more question, so for example i go start plot 0 and add plots 1000 like this

-Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 0 1000 4000 4 the next cmd would be
-Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 1000 1000 4000 4 i don´t have to go 1001 + 1000, because i read that in this thread
and it made it impossible to plot even files, because it couldn´t be evenly divided.

Thats why i startet to plot 1 big file with gpu.

Hard and very time intensive learning process

[burstcoin]

lol, for example my last dadline is 707506293144
Does this tell me, that my plot file is fucked?

either that, or that the plot file has very very few nounces in it, and is super small. I've had very small plot files under generation generate crazy deadlines, but deadlines improve a lot as the file grow in size, something like perhaps the best seen deadline over some period gets half as big, with double the filesize (or something like that)

my pc with 3TB plots right now has a minimum deadline of 4721 over the last 6 hours of mining, so it's not super easy to hit the small ones.

btw. it would be overly super cool if the java wallet server could be set up to add to a file the best deadline seen for every block one is mining. Such a log would be fun to have. The current log has to be set to log *A LOT* of stuff before the wallet starts logging deadline numbers.

Just change in the logging config nxt.level = INFO to nxt.level = FINE and all the deadlines will be in the log file. Then just do something like cat nxt.log | grep "started mining, deadline" >deadlines.log to pull just the deadline lines.

[enta2k]

Thanks so much for your help guys, i really apreciate.

Im gonna redo it with 2.1.1 and 14.4 drivers on my 7970
As i see it works with ~ 22500nonces/min, so i should be ready in 16h
but funny sidefact, even with a tiny file after plotting for now 3min. or what i found a deadline with 1mio. before nothing under a billion! Smaller than before with 6tb drive full.
I asked 20 pages before if my deadline are to high and no one said anything

This explains why i never found anything, it´s so complicated if you have absolutely no glue about this mining thing lol
Hopefully in a few month im better

[pinballdude]

Hey pinnballdude thats really very usefull, i will do it like that.
Damn that makes me sad, i thought after 2 weeks of trying around now everything is fine.
I was able to plott the whole 5,4tb file in 15h, now it takes a week or i have to find a gpu plotter version that works correctly.
Didn´t knew, that it could be so bad.

I have one more question, so for example i go start plot 0 and add plots 1000 like this

-Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 0 1000 4000 4 the next cmd would be
-Xmx2000m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner generate  blabla 1000 1000 4000 4 i don´t have to go 1001 + 1000, because i read that in this thread
and it made it impossible to plot even files, because it couldn´t be evenly divided.

Thats why i startet to plot 1 big file with gpu.

Hard and very time intensive learning process

I haven't totally mastered the numbers reg. what the next last number must be (something with number of nonces or some such)

I just went ahead and created plots of size 500000 or 1000000  (that's 256*500000 bytes = 128 gig) or 256 gig for 1000000  my expectation is that at most, the last 4000 bit of the file is fubar, and i can live with that, if the rest of the half million is fine. I understand that if the 4000 number is larger, then mining will have to seek less times on disk, which should make looking for good deadlines faster (you don't want to spend 10 seconds finding your 5 second deadline only to see some guy with a 8 second deadline being first to announce). Anyways, the files i have created with 5000000 or 1000000 sizes and 4000 as the last number they are working fine fine.

if you do like 1.000.000 500.000 then the last generated nounce ought to be 1.499.999 (bc the first was 1.000.000) so it shoulld be correct to start the next file at 1.500.000 and have no waste. i would guess that it is important that the 3rd number divides the 2nd number evenly or else we probably have a bad last block or something, however, i am qute sure that if the very last bit of a file is fubar, you only loose tickets from that very last partial block of say 4000 tickets(nonces or whatever they are called).


oh,also - remember to install 64 bit java, i had all sorts of memory problems with the default 32 bit java that seemingly is only installed on windows 64 bit. You can get the 64 bit one from sun, just follow the links. When you have both, be sure to start the correct one

cpu plotting is monster slow, takes days to fill a decent sized drive, even with fair gear.

[vaxman]

I asked 20 pages before if my deadline are to high and no one said anything

sorry, might have been lost in the noise, 537 pages and counting.
std disclaimer: use the forum: http://burstforum.com

[pinballdude]

lol, for example my last dadline is 707506293144
Does this tell me, that my plot file is fucked?

either that, or that the plot file has very very few nounces in it, and is super small. I've had very small plot files under generation generate crazy deadlines, but deadlines improve a lot as the file grow in size, something like perhaps the best seen deadline over some period gets half as big, with double the filesize (or something like that)

my pc with 3TB plots right now has a minimum deadline of 4721 over the last 6 hours of mining, so it's not super easy to hit the small ones.

btw. it would be overly super cool if the java wallet server could be set up to add to a file the best deadline seen for every block one is mining. Such a log would be fun to have. The current log has to be set to log *A LOT* of stuff before the wallet starts logging deadline numbers.

Just change in the logging config nxt.level = INFO to nxt.level = FINE and all the deadlines will be in the log file. Then just do something like cat nxt.log | grep "started mining, deadline" >deadlines.log to pull just the deadline lines.

Thanks, I am already doing something like that, but on windows these things are not as easy as on unix, unless i install a decent shell , etc.
If the deadline message could have its nxt.level raised to INFO that would be quite cool, then we could not have to have giant log files if we want to see what deadlines we've been having.  with level FINE the log file grows almost faster than posts to this forum ;-)

When i get more time, i'll try to build the jars in windows (no cygwin) and if i succeed, i'll probably make a fork and later a PR with various small changes - but work is really tough at the moment. Easy stuff like making the miner batch files find the installed java like the wallet , etc. I think without much work on my end, i can make it somewhat easier for windows users to get involved without knowing a lot of tech. Great coin, btw. very tech innovative and a dev that is still alive.. keep up the good work!

[enta2k]

Hey PBD, i have a 6 core i980x @ 4Ghz and still super slow.
I now try with gpu plotter 2.1.1, looks good so far, will see tomorrow.

But i think you are right, they say it should not overlap, not that it is a big deal if there is 1 space.
I had super fucked up ram issues the last couple days, now i installed a new 12gb kit (only had 6gb before), deleted java 32 (already had 64),
priorised the java application to high and set the xms to 4096MB, now it works absolute fine, but still java uses up to 3gb ram.

I think i didn´t solved the problem, but i fixed it by just adding more ram and giving java more ram
Not the best option, but as long as it works its fine with me

[burstcoin]

lol, for example my last dadline is 707506293144
Does this tell me, that my plot file is fucked?

either that, or that the plot file has very very few nounces in it, and is super small. I've had very small plot files under generation generate crazy deadlines, but deadlines improve a lot as the file grow in size, something like perhaps the best seen deadline over some period gets half as big, with double the filesize (or something like that)

my pc with 3TB plots right now has a minimum deadline of 4721 over the last 6 hours of mining, so it's not super easy to hit the small ones.

btw. it would be overly super cool if the java wallet server could be set up to add to a file the best deadline seen for every block one is mining. Such a log would be fun to have. The current log has to be set to log *A LOT* of stuff before the wallet starts logging deadline numbers.

Just change in the logging config nxt.level = INFO to nxt.level = FINE and all the deadlines will be in the log file. Then just do something like cat nxt.log | grep "started mining, deadline" >deadlines.log to pull just the deadline lines.

Thanks, I am already doing something like that, but on windows these things are not as easy as on unix, unless i install a decent shell , etc.
If the deadline message could have its nxt.level raised to INFO that would be quite cool, then we could not have to have giant log files if we want to see what deadlines we've been having.  with level FINE the log file grows almost faster than posts to this forum ;-)

When i get more time, i'll try to build the jars in windows (no cygwin) and if i succeed, i'll probably make a fork and later a PR with various small changes - but work is really tough at the moment. Easy stuff like making the miner batch files find the installed java like the wallet , etc. I think without much work on my end, i can make it somewhat easier for windows users to get involved without knowing a lot of tech. Great coin, btw. very tech innovative and a dev that is still alive.. keep up the good work!

http://gnuwin32.sourceforge.net/ install in windows, add the bin directory to system path, and then a lot of linux commands work fine in windows command prompt.

Deadline spam rapidly adds to the log size if you have a few miners going, so it really doesn't belong enabled by default IMO.

[boliu]

another simple question, hope someone can resolve.

1. For plots, does it have to be sequential, as long as I don't overlap the nonce, I should be ok right?
2. I have one computer to mine, and another to generate the plots files, once new plots files are created, can I just copy them to the mining plot directory, would I need to restart the miner every time new plot files are copied?

thanks

[CryptoSurfer]

another simple question, hope someone can resolve.

1. For plots, does it have to be sequential, as long as I don't overlap the nonce, I should be ok right?
2. I have one computer to mine, and another to generate the plots files, once new plots files are created, can I just copy them to the mining plot directory, would I need to restart the miner every time new plot files are copied?

thanks

1: Correct, they do not need to be sequential
2: Correct, no need to restart the miner. It will read all files in the plot directory.

[pinballdude]

Deadline spam rapidly adds to the log size if you have a few miners going, so it really doesn't belong enabled by default IMO.

I agree, that's why i was thinking along the lines of some other place where the deadlines could be saved, to a seperate file, perhaps optionally. But this is of course quite low priority, it is mostly for the curious and perhaps for figuring if something is not working as expected somewhere. The deadline saving could also be done by the miners, so that the wallet was not involved at all. That would potentially open up for stats about what kind of deadline you get from what plot files and mining processes, perhaps identifying bad or broken ones.

Another cool (but rather useless except for the fun factor) feature could be if I could see somehow in the web client, what the deadline was, for each of the blocks i've forged. Perhaps in block details if it is at all possible to calculate after the fact. (i guess an approximation could be done by looking at the timestamp of the prior block)

Anyway, like other people suggested, coin features are more important than miner bells and whistles. As long as it is quite easy to start mining, then i guess the most value to the community can be generated by enhancing coin/currency features, documentation, marketing etc. Dudes like me with little time could sponsor work by putting up prizes for tasks. I might do that, once i have found more than the one block i have right now. Won't take long i hope, with 5TB and growing.

[boliu]

another simple question, hope someone can resolve.

1. For plots, does it have to be sequential, as long as I don't overlap the nonce, I should be ok right?
2. I have one computer to mine, and another to generate the plots files, once new plots files are created, can I just copy them to the mining plot directory, would I need to restart the miner every time new plot files are copied?

thanks

1: Correct, they do not need to be sequential
2: Correct, no need to restart the miner. It will read all files in the plot directory.

Thank you.

[paradigmflux]

So far only 200 assets have sold for BMPOOL.
If the community does not want the idea of community owned hardware to continually create buy pressure, then I will not press it.  God knows I am busy enough already with the hashrate NXT community pool.

The asset # on the Burst AE is 11375670541237055652

 
Unless there are some more sizable investors soon, the idea will be abandoned and I will issue buy orders to buy back every share that was sold.

(the pool would not be investing in hashlets or the same Scrypt/SHA ASIC hardware that anyone else can buy and get hosted - we would be purchasing additional Scrypt-N FPGA hosted equipment the same way the Hashrate pool has recently invested (as it has an absolutely amazing ROI - less than 6 weeks).
no, I won't post the FPGA provider or manufacturer either.

[theprofileth]

So far only 200 assets have sold for BMPOOL.
If the community does not want the idea of community owned hardware to continually create buy pressure, then I will not press it.  God knows I am busy enough already with the hashrate NXT community pool.

The asset # on the Burst AE is 11375670541237055652

 
Unless there are some more sizable investors soon, the idea will be abandoned and I will issue buy orders to buy back every share that was sold.

(the pool would not be investing in hashlets or the same Scrypt/SHA ASIC hardware that anyone else can buy and get hosted - we would be purchasing additional Scrypt-N FPGA hosted equipment the same way the Hashrate pool has recently invested (as it has an absolutely amazing ROI - less than 6 weeks).
no, I won't post the FPGA provider or manufacturer either.

You should probably make a separate thread that is NOT in the burst coin ann thread that way people have a way to easily ask questions and you can update asset holders plus it is easier to market that way
Also consider taking non burst currencies as not that many people actually own burst