Skip to content

burstforum.com thread recovery

Jump to bottom Edit New page
luxe edited this page Sep 4, 2015 · 3 revisions

Welcome to the burstcoin-jminer wiki! I tried to recover the discussion from burstforum.com (witch is actually down) via google cache. Guess we lost a few of the last comments ... but 100 recovered is better than nothing.

luxe Moderator New java (all platforms) poc-miner with GPU support.

Requirements:

  • Java8
  • openCL driver/sdk (if you have BURST Wallet and gpuPlotter running, it should work out of the box)

[stable release] burstcoin-jminer-0.3.6-RELEASE

  • solo and all pools supported
  • AMD and nVIDIA supported
  • Windows, Linux and Mac supported
  • for a QUICKSTART check README.txt

https://github.com/de-luxe/burstcoin-jminer/releases

About: I was just for fun developing a own poc miner, i got it working ... but the miner was ~10% slower than e.g. blago miner, therefore i was asking in the community for openCL implementation of SHABAL, to optimize the mining performance ... after @cryo gave me some kernel files from burstDev, i contacted burstDev and he provided the code, that enables this miner to use GPU for mining. PLEASE DONATE TO HIM: BurstDev: BURST-QHCJ-9HB5-PTGC-5Q8J9

To understand the importance, look at the following CPU and I/O usage, 2 Blocks mined with CPU and than 2 Blocks mined with GPU-Support. Sorry there is no GPU graph ... it was ca. 50% i guess. (Please login here to see full size pictures ...)

poc-mining: CPU vs. GPU

I give you some numbers to decide, if its worth following this thread.

PC-1: 85TB plots: ~80sec with CPU (12 cores intel@4.1GHz) bottleneck: CPU ~16sec with GPU (280x@~50%) bottleneck: HDD read speed

PC-2: 56TB plots: ~70sec with CPU (8 cores amd@4.5GHz) bottleneck: CPU ~24sec with GPU (amd HD 7970@~50%) bottleneck: HDD read speed ~48sec with GPU (amd HD 5450@~100%) bottleneck: GPU

As you can see we have some benefits, finish rounds faster means: commit all deadlines, even on fast blocks and less hardware usage, sure this will save energy :)

Here is another screen, take a look a GPU Load (280x@85TB) GPU & CPU & IO @ 85TB

Sounds to good to be true, indeed ... there is still a lot of work and a lot of issues to fix ...

known issues / todo:

  • optimized devPool support
  • strange deadline results from some plot-files ... (i replotted ca. 10% to reduce/fix that, seams to work, hell maybe i fund some corrupt ones)
  • writing FAQ / Docs for miner, to keep all infos at one place.
  • optimize number of parts per chunk calculation [thx blago]

burstcoin-jminer-0.3.4-SNAPSHOT: screenshot of 0.3.4 version.

http://www.howtogeek.com/howto/19982/how-to-make-the-windows-command-prompt-wider

to be continued soon ...

  Attached Files: jminer_cmd_0.3.1.gif jminer_cmd_0.3.1.gif File size: 139.9 KB Views: 190 mining_engine_restart.gif mining_engine_restart.gif File size: 65.4 KB Views: 177 Last edited: Aug 26, 2015 luxe, Jul 18, 2015 #1 TruBadun, crowetic, Merick and 3 others like this. luxe luxe Moderator THIS MINER IS EXPERIMENTAL SOFTWARE, USE AT YOUR OWN RISK!

  • tested on AMD/nVIDIA GPU, please give me feedback if it works on CPU or other GPUs
  • tested on burst.ninja, pool.burscoin.de and devV2Pool.

NOTICE: on updating you can re-use your jminer.properties (if not mentioned)

[release] burstcoin-jminer-0.3.6-RELEASE

[snapshots] burstcoin-jminer-0.3.6-SNAPSHOT.zip 8.4 MB

burstcoin-jminer-0.3.5-SNAPSHOT.zip 8.4 MB

burstcoin-jminer-0.3.4-SNAPSHOT.zip 8.4 MB

  • lot of refactoring and cleaning up code (no line of code untoched...)
  • complete event-driven miner core, CommandLineRunner just 'listening'.
  • added additional information while mining (progress, send, confirm, read speed, baseTarget etc.)
  • harden the whole jminer.properties code to print useful errors on missing/invalid config.
  • 'winner' in pool mode disabled if no walletServer specified
  • some miner bugfixes, additional info in jminer.properties to help config the miner ... and more ...
  • prepared a bunch of new settings that will be adjustable in next version ...
  • for a QUICKSTART check README.txt https://mega.co.nz/#!EgxWHCxI!PigGwnrNwcdbKBSL8BfBSLM7gEvNioqWv1DdeRNrWZQ

burstcoin-jminer-0.3.3-SNAPSHOT.zip 8.4 MB

  • NEW setting in jminer.properties to tune memory usage / mining speed (NOTICE: requires updating jminer.properties!!!!)
  • in solo-mode walletServer is not required anymore
  • miner will wait for round to finish before it restarts itself
  • winner in pool-mode is no longer out of sync (miner will retry on wrong blockNumber) https://mega.co.nz/#!A5JxlaDS!Mi8j1JiKsV_-v9XfQrd1LmI-1bkyEjZQCPSBzVS74qg

burstcoin-jminer-0.3.2-SNAPSHOT.zip 8.4 MB

burstcoin-jminer-0.3.1-SNAPSHOT.zip 8.4 MB https://mega.co.nz/#!AsoTTbJI!uKDT3BXKJynUQlsljf69wYDPTj2XnLjN63R6t4r5Rk8   Last edited: Aug 26, 2015 luxe, Jul 18, 2015 #2 M3ta, crowetic, daWallet and 2 others like this. luxe luxe Moderator [UPDATE] first test version available!   luxe, Jul 19, 2015 #3

Haitch Haitch Gigabyte Luxe, I'm getting a file is no longer available error when hitting the mega link.

H.   Haitch, Jul 19, 2015 #4

luxe luxe Moderator sorry, re-uploaded and replaced url, should work again ...   luxe, Jul 19, 2015 #5

luxe luxe Moderator is it running for you guys? damn, has no one tested this yet?! any problems setting it up? nothing?!

btw. this page is awesome: http://smartcontr.actor/   Last edited: Jul 21, 2015 luxe, Jul 20, 2015 #6

Irontiga Irontiga Moderator Staff Member luxe said: ↑ btw. this page is awesome: http://smartcontr.actor/ Click to expand... Be careful, those at's are NOT tested, but they should work.

Will give the GPU miner a shot some time...haven't been mining lately....will have to find that usb hdd to give a test.   Irontiga, Jul 20, 2015 #7

Merick Merick Moderator Staff Member Looks good Luxe, now I have a reason to play around with my minner again, for the last 5 months or so it has just been crunching away. I will try this shortly. Keep up the good work. I put a link to this thread in the Stickied tools list - https://burstforum.com/index.php?threads/tool-collection-and-info.777/   Merick, Jul 21, 2015 #8

luxe luxe Moderator Merick said: ↑ ... I put a link to this thread in the Stickied tools list - https://burstforum.com/index.php?threads/tool-collection-and-info.777/ Click to expand... thanks merick, maybe you also want to add this one: https://burstforum.com/index.php?threads/burst-address-generator.759/#post-5930   luxe, Jul 21, 2015 #9

Blago Blago Gigabyte luxe, i have some issues, pm to you. this.workgroupSize[0] always = 64L ?   Blago, Jul 21, 2015 #10 luxe likes this. luxe luxe Moderator Blago said: ↑ luxe, i have some issues, pm to you. this.workgroupSize[0] always = 64L ? Click to expand... no.

Like i said before the openCL code was provided by burstDev, but as far as i can see, there is everything ok. The miner uses jocl lib, and the method 'clGetKernelWorkGroupInfo(...)' is called, to detect workgroup size. (e.g. for me it is 256) javadoc: http://www.jocl.org/doc/org/jocl/CL.html (search for method name) the doc says: 'Returns information about the kernel object that may be specific to a device.' So i would say the miner will auto-detect your workgroup size.

But you provided your plot-file names, and said some are working some are not. Please go step by step: Remove every drive from config, that get the exception: 'CL-Exception Dir: D:/plots' for example. Not sure if you completely did that already?! The miner will only use plot-files with your 'numeric accountId' in it, so after you identified the drives, you can find the problematic plot files, if you rename the plot-files to exclude. (e.g. remove first number)

I also had that issue, for me it where just some plot files, mainly smaller ones, i just deleted and replotted them. (I forgot about that, must be some weeks ago.) Thanks for reporting! The miner should be able to find 'not supported' plot-files and list them to the user. (or at least log the problematic file and not the whole drive xd) The error handling is not that clever at this point. Its on the todo list now ...

This may be useful: http://www.howtogeek.com/howto/19982/how-to-make-the-windows-command-prompt-wider

Edit: i will look deeper into it, to find the reason and not only solve the consequences ... but that can take a while. If you found a not working stagger / plot file name, please let me know, i will than plot a test-one to debug into it :p

Edit: problem seams solved with new version.   Last edited: Jul 21, 2015 luxe, Jul 21, 2015 #11 Blago likes this. Blago Blago Gigabyte luxe said: ↑ no.

Like i said before the openCL code was provided by burstDev, but as far as i can see, there is everything ok. The miner uses jocl lib, and the method 'clGetKernelWorkGroupInfo(...)' is called, to detect workgroup size. (e.g. for me it is 256) javadoc: http://www.jocl.org/doc/org/jocl/CL.html (search for method name) the doc says: 'Returns information about the kernel object that may be specific to a device.' So i would say the miner will auto-detect your workgroup size.

But you provided your plot-file names, and said some are working some are not. Please go step by step: Remove every drive from config, that get the exception: 'CL-Exception Dir: D:/plots' for example. Not sure if you completely did that already?! The miner will only use plot-files with your 'numeric accountId' in it, so after you identified the drives, you can find the problematic plot files, if you rename the plot-files to exclude. (e.g. remove first number)

I also had that issue, for me it where just some plot files, mainly smaller ones, i just deleted and replotted them. (I forgot about that, must be some weeks ago.) Thanks for reporting! The miner should be able to find 'not supported' plot-files and list them to the user. The error handling is not that clever at this point. Click to expand... ok, maybe try to change in findLowest() and findTarget() try{ long numNonces = data.length / 64; calcWorkgroups = numNonces / this.workgroupSize[0]; if (numNonces % 64L != 0L) { calcWorkgroups += 1L; }

to if (numNonces % this.workgroupSize[0] != 0L) { ........................ catch (Exception e) { long calcWorkgroups; // not used   Blago, Jul 21, 2015 #12 luxe likes this. kappafree kappafree Megabyte luxe said: ↑ THIS MINER IS EXPERIMENTAL SOFTWARE, USE AT YOUR OWN RISK! I will provide sources in later versions (soon) or on trusted people request :p

  • only tested on AMD yet, please give me feedback if it works on nVIDIA
  • only tested on burst.ninja pool yet, please give me feedback about issues on other pools
  • no optimized dev v2 pool support yet, that will follow. Click to expand...

hello luxe, it works for me, 12 hours of test In the pool pool.burscoin.de AMD Radeon R9 37GB in 17 sec. with network drives

But I have problems on files larger than 256GB   kappafree, Jul 21, 2015 #13 luxe likes this. luxe luxe Moderator Blago said: ↑ ok, maybe try to change in findLowest() and findTarget() try{ long numNonces = data.length / 64; calcWorkgroups = numNonces / this.workgroupSize[0]; if (numNonces % 64L != 0L) { calcWorkgroups += 1L; }

to if (numNonces % this.workgroupSize[0] != 0L) { ........................ catch (Exception e) { long calcWorkgroups; // not used Click to expand... you have pm (here) with that changes made ... hope that can fix the issue ... i can provide sources to you if you like, its just not cleaned up yet, therefore not public yet.   luxe, Jul 21, 2015 #14

luxe luxe Moderator kappafree said: ↑ hello luxe, it works for me, 12 hours of test In the pool pool.burscoin.de AMD Radeon R9 37GB in 17 sec. with network drives

But I have problems on files larger than 256GB Click to expand... first of all, thanks for that testing :) nice to hear, that you got it running so far. (was very important for me to know, if it works with other pools! good to know that network-drives are not a issue!) the file size in general should not be a issue, i have 4tb files without issues ... please provide the exception (just make a screenshot or something^^) and the file-name/stagger size, and i can look into it. (sonds like you got the same bug like blago)   Last edited: Jul 21, 2015 luxe, Jul 21, 2015 #15

Blago Blago Gigabyte luxe said: ↑ you have pm (here) with that changes made ... hope that can fix the issue ... i can provide sources to you if you like, its just not cleaned up yet, therefore not public yet. Click to expand... WORKS fine! 11 discs, 17482 Gb @ 23 sec (Blago's miner @ 42 sec :) )   Blago, Jul 21, 2015 #16 luxe likes this. luxe luxe Moderator Blago said: ↑ WORKS fine! 11 discs, 17482 Gb @ 23 sec (Blago's miner @ 42 sec :) ) Click to expand... wow!! thanks for that, i will just test it on my machines and than release a new version :) could you post your pool/solo and GPU, would be interesting.

@kappafree maybe you just wait for that ... (give me 1h, need to eat something ... xd)   Last edited: Jul 21, 2015 luxe, Jul 21, 2015 #17

luxe luxe Moderator [UPDATE] new version added   luxe, Jul 21, 2015 #18 Irontiga likes this. kappafree kappafree Megabyte I confirm that the new version 0.3.2 works fine, with all my plots files.

very fast   kappafree, Jul 21, 2015 #19 luxe likes this. kappafree kappafree Megabyte I tested with an old NVIDIA Quadro NVS 290. I installed the latest graphics drivers. It works without errors !!! performances are not exciting however, it goes from 90 sec to 72 sec with 20TB in an old PC   kappafree, Jul 22, 2015 #20 luxe likes this.

luxe luxe Moderator kappafree said: ↑ I tested with an old NVIDIA Quadro NVS 290. I installed the latest graphics drivers. It works without errors !!! performances are not exciting however, it goes from 90 sec to 72 sec with 20TB in an old PC Click to expand... Good news, so NVIDIA is supported, nice! Once again thank you so much for testing and reporting ... that's a big help!

Everyone - please - report about issues or success ...   luxe, Jul 22, 2015 #21

kappafree kappafree Megabyte Yes... nice! It also works with a NVIDIA GeForce 8400 48 sec instead of 72 sec to 12TB in a very old PC with 4 HDD EXT USB 2.0!!! windows 8.1 and too little memory

Important latest version of java.   kappafree, Jul 22, 2015 #22

luxe luxe Moderator kappafree said: ↑ Important latest version of java. Click to expand... Yes, the miner requires Java8 ... did you have issues with early Java8 version or just with e.g. Java7 ?   luxe, Jul 22, 2015 #23

kappafree kappafree Megabyte I had the old version Java7. And obviously not working.   kappafree, Jul 22, 2015 #24 luxe likes this. luxe luxe Moderator kappafree said: ↑ I had the old version Java7. And obviously not working. Click to expand... Ok, its in the README.txt that the miner needs Java8, i will add that info to the jminer.properties, to ensure everybody knows.

Some info about memory usage and more ... The miner will read everything into memory without waiting for the checker! In the theoretical worst case ~0,025% of total capacity mined. But the reader splits your plot-file chunks (one chunk = staggersize) into smaller parts and delivers this chunk-parts to checker. So in theoretical best case, the first chunk-part is already checked (and deadline can be committed) before the second is read into memory. That enables the miner to read even a optimized 4tb chunk with low memory usage. (If GPU is fast enough) The size of those chunk-parts will be adjustable in next version, that will allow a kind of tuning memory usage and/or mining speed.

Summary: Fast GPU and small chunk-parts lead to low memory usage, but if chunk parts are to small, it will slow down the miner. Therefore this will be adjustable.   Last edited: Jul 23, 2015 luxe, Jul 22, 2015 #25

daWallet daWallet Gigabyte Does this miner work (fast) with a simple Intel HD 4000 graphic card? Probably not, right?   daWallet, Jul 22, 2015 #26 luxe likes this. luxe luxe Moderator daWallet said: ↑ Does this miner work (fast) with a simple Intel HD 4000 graphic card? Probably not, right? Click to expand... I tested a passive AMD HD 5450 ... 1153603.jpg and it was faster than a new AMD CPU with 8 cores@4,5GHz, just give it a try and report :)

You could even test the miner on a CPU, just point the diviceId+platformId to it. One thing i didn't test yet :rolleyes:   Last edited: Jul 22, 2015 luxe, Jul 22, 2015 #27 daWallet likes this. luxe luxe Moderator [UPDATE] new version added   luxe, Jul 22, 2015 #28

Blago Blago Gigabyte [​IMG] Win7 Ult., Java 8, RAM 16 Gb, 11 HDDs. problem with Windows memory manager (pre-reading files). EnableSuperFetch=0 EnablePrefetcher=0   Blago, Jul 23, 2015 #29 luxe likes this. luxe luxe Moderator Blago said: ↑ Win7 Ult., Java 8, RAM 16 Gb, 11 HDDs. problem with Windows memory manager (pre-reading files). EnableSuperFetch=0 EnablePrefetcher=0 Click to expand... Not sure if i got you right, you have a problem with that or is this the solution ?!

However, java has (as far as i know) no influence on how windows handles its caches ... i did not look deeper in it, but using java vm limits influence on guest-system i guess. Maybe miner could force java vm to allocate nearly all system memory before round starts, and free the memory after that o_O So there will be no memory that win could waste for caching ...

Windows7 files system caching is a known problem while mining BURST (you know that), not sure if other miners solved it?!, cause i first changed to Linux cause of that and later to Windows8, where this issue does not exist.   Last edited: Jul 23, 2015 luxe, Jul 23, 2015 #30

Blago Blago Gigabyte Yes, this is a problem Win 7, Win 8 is better to use or Win 10. People are asked to make a miner, which does not require the wallet (pool-mode), you can do as an option? Also asked to tell more about the staggerSize chunkPartNonces- finding the optimal algorithm of this value.   Last edited: Jul 23, 2015 Blago, Jul 23, 2015 #31 luxe likes this. luxe luxe Moderator Blago said: ↑ People are asked to make a miner, which does not require the wallet (pool-mode), you can do as an option? Click to expand... sure, will make this optional, in next version you will be able, to just leave it empty, and feature will be disabled. Blago said: ↑ Also asked to tell more about the staggerSize - finding the optimal algorithm of this value. Click to expand... staggerSize, in general? optimized plot-files are best

or are we talking about the 'chunkPartNonces' thing?! said something about that a few posts above. (Some info about memory usage and more ...) here is the code that uses it ... Code: private int calculateNumberOfParts(long staggeramt) { long targetNoncesPerPart = chunkPartNonces != null ? chunkPartNonces : 320000;

// calculate numberOfParts based on target
int suggestedNumberOfParts = (int) (staggeramt / targetNoncesPerPart) + 1;

// ensure stagger is dividable by numberOfParts, if not adjust numberOfParts
while(staggeramt % suggestedNumberOfParts != 0)
{
  suggestedNumberOfParts += 1;
}
return suggestedNumberOfParts;

} (i'm not happy with the 'while', should solve that better e.g. adjust targetNoncesPerPart )   Last edited: Jul 23, 2015 luxe, Jul 23, 2015 #32

Blago Blago Gigabyte luxe said: ↑ sure, will make this optional, in next version you will be able, to just leave it empty, and feature will be disabled.

staggerSize, in general? optimized plot-files are best

or are we talking about the 'chunkPartNonces' thing?! said something about that a few posts above. (Some info about memory usage and more ...) here is the code that uses it ... Code: private int calculateNumberOfParts(long staggeramt) { long targetNoncesPerPart = chunkPartNonces != null ? chunkPartNonces : 320000;

// calculate numberOfParts based on target
int suggestedNumberOfParts = (int) (staggeramt / targetNoncesPerPart) + 1;

// ensure stagger is dividable by numberOfParts, if not adjust numberOfParts
while(staggeramt % suggestedNumberOfParts != 0)
{
  suggestedNumberOfParts += 1;
}
return suggestedNumberOfParts;

} (i'm not happy with the 'while', should solve that better e.g. adjust targetNoncesPerPart ) Click to expand... let me offer the variant... if stagger <= chunkPartNonces, suggestedNumberOfParts must be ==1, maybe better

chunkPartNonces != null ? chunkPartNonces : 320000; // calculate numberOfParts based on target int suggestedNumberOfParts = (int) (staggeramt / chunkPartNonces); if ((staggeramt % chunkPartNonces) != 0) suggestedNumberOfParts += 1;

(chunkPartNonces - it's const per session, so targetNoncesPerPart not needed)   Blago, Jul 23, 2015 #33 luxe likes this. luxe luxe Moderator Blago said: ↑ let me offer the variant... if stagger <= chunkPartNonces, suggestedNumberOfParts must be ==1, maybe better

chunkPartNonces != null ? chunkPartNonces : 320000; // calculate numberOfParts based on target int suggestedNumberOfParts = (int) (staggeramt / chunkPartNonces); if ((staggeramt % chunkPartNonces) != 0) suggestedNumberOfParts += 1;

(chunkPartNonces - it's const per session, so targetNoncesPerPart not needed) Click to expand... thanks for your suggestion ... i will push sources to github in 1-2 weeks, please give me some time, to clean up code before. Code: int suggestedNumberOfParts = (int) (staggeramt / targetNoncesPerPart) + 1; suggestedNumberOfParts is 1 if chunkPartNonces bigger than staggeramt, there is a +1 setting chunkPartNonces bigger than highest stagger will cause miner to always use one part, and prevent splitting chunks. Blago said: ↑ (chunkPartNonces - it's const per session, so targetNoncesPerPart not needed) Click to expand... true, its just temp but at this state i want to focus on features and major bugs, and maybe not on such a detail. Blago said: ↑ if ((staggeramt % chunkPartNonces) != 0) suggestedNumberOfParts += 1; Click to expand... this will only work if i change the code in reader, cause it currently expects stagger is dividable by numberOfParts ... i will keep that in mind, cause it would be better this way. (no risk of to much parts)   Last edited: Jul 23, 2015 luxe, Jul 23, 2015 #34

M3ta M3ta Gigabyte Sort of works, but most of the output is cluttered with:

java.lang.Thread.Run(Unknown Source)

Solo mode.   M3ta, Jul 26, 2015 #35

luxe luxe Moderator M3ta said: ↑ Sort of works, but most of the output is cluttered with:

java.lang.Thread.Run(Unknown Source)

Solo mode. Click to expand... Thanks for testing ... If you get exceptions while running the miner, please give me some more info (whole stacktrace/settings/screenshot), that one line does not help much ...

If you just criticise, that sources are not published yet, please be patient, that will follow.   Last edited: Jul 26, 2015 luxe, Jul 26, 2015 #36

M3ta M3ta Gigabyte Heh, sorry about that. I keep forgetting Java can throw a zillion different errors, and then it's meaningless, if you don't look at the 30 other errors before the last one.. :)

Anyway.. screenshot:

[​IMG]   M3ta, Jul 26, 2015 #37 luxe likes this. luxe luxe Moderator does that happen in every round or just from time to time? i harden that one for coming version, guess it does not stop or stuck the miner, its just about calculating capacity for 'done' TB / %   luxe, Jul 26, 2015 #38

M3ta M3ta Gigabyte luxe said: ↑ does that happen in every round or just from time to time? i harden that one for coming version, guess it does not stop or stuck the miner, its just about calculating capacity for 'done' TB / % Click to expand... It happens on every round. For debugging, I tried running it with only the C: plots. No errors. Then gradually added 1 drive at a time, to see which ones were triggering the errors.

I found out that including some(and only some) drives on the jminer.properties throws these errors. The other drives.. no problems.

All my plots are optimised and have been "healthy" for some time now.. I have no idea why jminer only "likes" 8 out of 14 drives. 😕   M3ta, Jul 26, 2015 #39

luxe luxe Moderator Hmm, are u using 0.3.3 version or 0.3.1 ?! (if u use 0.3.1. updating should solve that) If u are using 0.3.3 you could try to find the plot-files that cause the error, just rename files (e.g. remove first number of numeric account id) to exclude them from mining. Maybe we can find a not supported stagger setting ...   luxe, Jul 26, 2015 #40

  1. M3ta M3ta Gigabyte 0.3.3.

Well, all my staggers are the same, and since the last version of Blago's miner, wrong staggers show a warning, so I'd know.

It's going to be nearly impossible to manually find the specific plots jminer is complaining about. More than 500 plots total.. :)   M3ta, Jul 26, 2015 #41

  1. luxe luxe Moderator well, another thing that could cause that issue is a duplicated plot-file, did you use https://bchain.info/BURST/tools/overlap to check all your used plot files?! sorry but that's currently the last reason i could think of. i use the nonce-numbers as id, so overlapping nonces could lead to your error.

however, i will look deeper into that code, and try to make the next version working for you.   luxe, Jul 26, 2015 #42

  1. M3ta M3ta Gigabyte I'll check for overlaps.

Thanks for the quick replies, jminer is looking very good! I'm sure this is a problem on MY plots, not on the plotter. :)

Maybe add a debug mode that throws more details in a log file to find these kinds of problems? Just an idea!   M3ta, Jul 26, 2015 #43

  1. crowetic crowetic Moderator Staff Member Great job @luxe! Now I have to replot my drives that I've been too lazy to for a long time so that I can test it with correctly plotted drives.

But I'm very stoked to see this kind of development! Great work!

Sent from my iPhone using Tapatalk   crowetic, Jul 26, 2015 #44

  1. luxe luxe Moderator crowetic said: ↑ Great job @luxe! Now I have to replot my drives that I've been too lazy to for a long time so that I can test it with correctly plotted drives. Click to expand... What is wrong with your plot-files?! Overlapping of few nonces or something should not be a problem in general.

There is only a problem if 2 parts, the miner extracts from your chunks start with same nonceNumber. e.g. xxx_100_100_10 -> 110 (not unique startNonce) xxx_110_10_10 -> 110 (not unique startNonce)   luxe, Jul 26, 2015 #45

  1. crowetic crowetic Moderator Staff Member Hmm, maybe I'll try it out and see.

Sent from my iPhone using Tapatalk   crowetic, Jul 26, 2015 #46

  1. crowetic crowetic Moderator Staff Member The plots just aren't optimal, the stagger could be a lil more... Divisible by 256. Lol

Sent from my iPhone using Tapatalk   crowetic, Jul 26, 2015 #47

  1. luxe luxe Moderator deadline_0.GIF Lowest deadline ever ... :D! Found while testing ...   luxe, Jul 27, 2015 #48 crowetic and Blago like this.
  2. luxe luxe Moderator [UPDATE] new version added https://burstforum.com/index.php?threads/ann-burstcoin-jminer-beta.842/#post-5949   Last edited: Jul 28, 2015 luxe, Jul 28, 2015 #49 Blago likes this.
  3. Blago Blago Gigabyte at start doubled strings (networkPool-1) [​IMG] poolServer=http://195.209.126.166:8080 walletServer=http://localhost:8125

next blocks - all ok

[​IMG]   Last edited: Jul 28, 2015 Blago, Jul 28, 2015 #50 luxe likes this. 11. luxe luxe Moderator thanks for reporting ... those lines are printed on events from network, the timeout is bigger than the refresh interval, so it can happen ... no big deal .. guess it is not happening all the time?!

you can now leave walletServer empty, and winner feature will be disabled.   Last edited: Jul 28, 2015 luxe, Jul 28, 2015 #51

  1. luxe luxe Moderator @TruBadun reported problems with miner on devPoool (V1+V2) in short, devPools are not working with this miner yet, working on that issue, may take some time, cause they have their own api.

Edit: devPools are working since 0.3.6   Last edited: Aug 4, 2015 luxe, Jul 28, 2015 #52

  1. crowetic crowetic Moderator Staff Member Tested and working for me! Great read times on my 280x. Awesome work Luxe! gotta love it.

The only thing I would suggest, is possibly being able to run two account numbers on the single miner instance, right now I have to run two because I'm running my asset miner and my personal miner on the same rig.

But, it actually may be faster running two, so it doesn't much matter to me. GREAT!

I love it.   crowetic, Jul 30, 2015 #53 luxe likes this. 14. luxe luxe Moderator crowetic said: ↑ Tested and working for me! Great read times on my 280x. Awesome work Luxe! gotta love it.

The only thing I would suggest, is possibly being able to run two account numbers on the single miner instance, right now I have to run two because I'm running my asset miner and my personal miner on the same rig.

But, it actually may be faster running two, so it doesn't much matter to me. GREAT!

I love it. Click to expand... Thanks for feedback and suggestion. That could be a future feature ... just continue running multiple instances at the moment.   luxe, Jul 30, 2015 #54

  1. M3ta M3ta Gigabyte After fixing some overlapped plots shame almost everything works flawlessly.

I say "almost" because randomly (every 10 blocks or so) I get a java.lang.OutOfMemoryError: Direct buffer memory. This is with 32GB RAM, and according to system info, 43% is unused.

But as of 3.4, it really "feels" like a finished product now. eff.read tops 500MB/s occasionally. Major kudos, luxe!   M3ta, Jul 31, 2015 #55 luxe likes this. 16. luxe luxe Moderator M3ta said: ↑ After fixing some overlapped plots shame almost everything works flawlessly.

I say "almost" because randomly (every 10 blocks or so) I get a java.lang.OutOfMemoryError: Direct buffer memory. This is with 32GB RAM, and according to system info, 43% is unused.

But as of 3.4, it really "feels" like a finished product now. eff.read tops 500MB/s occasionally. Major kudos, luxe! Click to expand... Nice that it works for you!

About memory issues: Memory usage depends on your GPU speed. If you read faster than you GPU can handle that, memory usage will grow! (cause the reader will not wait for GPU) On memory issues, you can try 3 things ...

  1. allow java to allocate more memory e.g.: Code: // -Xmx20G - will allow java to use up to 20GB of memory // -XX:+UseG1GC - force java to free some not used memory // (on garbage collection after each round) java -jar -Xmx20G -XX:+UseG1GC burstcoin-jminer-0.3.4-SNAPSHOT.jar to find the right value, you can use Java Mission Control to monitor memory usage, e.g. on win C:\Program Files\Java\jdk1.8.xxx\bin\jmc.exe
  2. reduce 'chunkPartNonces' (will only help, if GPU is fast enough, will increase CPU usage)
  3. get a faster GPU (e.g. if you have 75TB of plots and a 70€ GPU, maybe buy a faster one, 280x can handle ~150-200TB within ~15sec)   luxe, Jul 31, 2015 #56 M3ta likes this.
  4. luxe luxe Moderator [UPDATE] new version with devPool support added. https://burstforum.com/index.php?threads/ann-burstcoin-jminer-beta.842/#post-5949   luxe, Aug 1, 2015 #57 M3ta likes this.
  5. M3ta M3ta Gigabyte luxe said: ↑ Nice that it works for you!

About memory issues: Memory usage depends on your GPU speed. If you read faster than you GPU can handle that, memory usage will grow! (cause the reader will not wait for GPU) On memory issues, you can try 3 things ...

  1. allow java to allocate more memory e.g.: Code: // -Xmx20G - will allow java to use up to 20GB of memory // -XX:+UseG1GC - force java to free some not used memory // (on garbage collection after each round) java -jar -Xmx20G -XX:+UseG1GC burstcoin-jminer-0.3.4-SNAPSHOT.jar to find the right value, you can use Java Mission Control to monitor memory usage, e.g. on win C:\Program Files\Java\jdk1.8.xxx\bin\jmc.exe
  2. reduce 'chunkPartNonces' (will only help, if GPU is fast enough, will increase CPU usage)
  3. get a faster GPU (e.g. if you have 75TB of plots and a 70€ GPU, maybe buy a faster one, 280x can handle ~150-200TB within ~15sec) Click to expand...

Thanks for your reply and tips. GPU is not a problem (I think), it's a XFX R9 290.

Increased java memory allocation, and 0.3.5 is working flawlessly! Thank you Thank you Thank you Thank you :)   M3ta, Aug 1, 2015 #58 luxe likes this. 19. luxe luxe Moderator M3ta said: ↑ Thanks for your reply and tips. GPU is not a problem (I think), it's a XFX R9 290.

Increased java memory allocation, and 0.3.5 is working flawlessly! Thank you Thank you Thank you Thank you :) Click to expand... You are welcome!   luxe, Aug 1, 2015 #59

  1. luxe luxe Moderator [UPDATE] new version added, devPool should work now. https://burstforum.com/index.php?threads/ann-burstcoin-jminer-beta.842/#post-5949   Last edited: Aug 4, 2015 luxe, Aug 3, 2015 #60 daWallet and TruBadun like this.

TruBadun TruBadun Megabyte luxe said: ↑ [UPDATE] new version added, devPool should work now. https://burstforum.com/index.php?threads/ann-burstcoin-jminer-beta.842/#post-5949 Click to expand... Thanks a lot! Working Great!   TruBadun, Aug 4, 2015 #61 luxe likes this. luxe luxe Gigabyte TruBadun said: ↑ Thanks a lot! Working Great! Click to expand... Puh finally :) Would not be there without your feedback, thanks for that!   luxe, Aug 4, 2015 #62

Burstage Burstage Megabyte Hi Luxe, I followed the setup readme file and set everything up but when I start the miner it says "unable to get mining info from wallet: invalid protocol pool.burstcoin.de "? I'm mining on burst.de with blagos miner without problem but wanted to test your miner to cut read times. Any help? Thanks   Burstage, Aug 12, 2015 #63

luxe luxe Gigabyte @Burstage sounds like you did not add 'http://' in front of 'poolServer' property, try using 'poolServer=http://pool.burstcoin.de:8080'

If that does not work, you need to provide some more information, e.g. post your jminer.properties   Last edited: Aug 12, 2015 luxe, Aug 12, 2015 #64

Burstage Burstage Megabyte Thats what I missed:oops: I was getting max 19-20mb/s on usb3 5tb now with your miner it reads at 56mb/s basically in half the time before 75-80sec now 30-40sec. Hopefully now I don't miss as many blocks. Highly recommended on usb great performance boost. Anyone using internal sata3 with speed comparisons? Thanks Luxe   Burstage, Aug 12, 2015 #65 luxe likes this. luxe luxe Gigabyte Burstage said: ↑ Thats what I missed:oops: I was getting max 19-20mb/s on usb3 5tb now with your miner it reads at 56mb/s basically in half the time before 75-80sec now 30-40sec. Hopefully now I don't miss as many blocks. Highly recommended on usb great performance boost. Anyone using internal sata3 with speed comparisons? Thanks Luxe Click to expand... Your plots are not optimized right? Whats your staggersize? (last number in plot-file names) With optimized plots you should need below 20sec. if read speed of your drive(s) is the bottleneck, and not CPU/GPU or your hardware in general. USB3 or SATA3 should make no difference, both can transfer data faster, than a drive can read. I have both USB3 and SATA, and no 4TB drive takes longer than 14-16sec. (Not all of my plots are optimized, but i have a quite high staggersize -> less drive seeks) Try: https://bchain.info/BURST/tools/overlap If you paste your plotfile-name(s) there, you will get info about drive seeks etc.   luxe, Aug 12, 2015 #66

Haitch Haitch Gigabyte Burstage,

I have 8 SATA3 drives (36 TB Total) via 2 * multiplexed eSATA links, mining time is 62-64 seconds, previously it was around 85 secs.

Another miner with 10 internal SATA3 drives (40TB total), direct SAS/SATA links, mining time is 18-21 secs, previously 44-48 seconds.

H.   Haitch, Aug 13, 2015 #67

Burstage Burstage Megabyte Not optimised stagger 16gb as that's the max ram I have, I'm running this on my laptop 2760qm 16gb ddr3 1600mhz win 10 pro on 500gb ssd, laptop is about 4 years old but can plot 4000-5000 nonce per min. Drive is connected to usb3 4 bay dock orico, planning to add 3 more drives when I have some spare funds available. I have another USB3 port free and an external esata port which I can add another 4 bay dock or single, what is the max drives on each port can this handle? My 5tb drive I need to replot as that was my first attempt and plotted 1 single plot but read the tutorial to leave 10% free, so I left 500gb free seems a bit of a waste of free space. Once I order more drives I'm planning to replot 200-300gb plots on the drives, read somewhere get better performance but can't find benchmarks for comparison.   Burstage, Aug 13, 2015 #68

Haitch Haitch Gigabyte If plotting/replotting use the gpuplotter in buffered mode - you can create plots with optimal stagger size.   Haitch, Aug 13, 2015 #69

crowetic crowetic Moderator Staff Member I have most of my drives plotted with a single plot file, and my miner with gpu miner reads plots in 26s or less with a 280x. Fantastic performance!

Luxe, I commend you on the awesome work.

I'm going to put a 270 in my rackserver when I can spend the extra cash to get one, with hopes that the 40TB there will read much faster. Most drives on the rackserver are connected via USB3 controller cards and currently take over 1min50sec to fully read plots, but I'm sure the bottleneck is the old XEONs that are in the thing. My thoughts are performance will drastically increase with a gpu and luxe's badass miner. :)

Sent from my iPhone using Tapatalk   crowetic, Aug 13, 2015 #70

Burstage Burstage Megabyte I'll give that a go, I only have a nvidia gt555m so not sure if the performance is going to be that great but worth a try. What would be my max stagger if using gpu plotter for my current system?   Burstage, Aug 13, 2015 #71

crowetic crowetic Moderator Staff Member Think it still depends on memory of the computer, if I'm not mistaken. I used 12000 for my stagger on most drives, but that was before I really knew exactly what I should do.

I would say go with 12800 or anything divisible by 64 multiplied up till you use you max mem.

Sent from my iPhone using Tapatalk   crowetic, Aug 13, 2015 #72

Burstage Burstage Megabyte I have stagger of 16384 right now with no problems. I have now been running this miner for 24hours and is very stable. Only thing is I get low memory warning even though I still have 40%-50% free, I just close and ignore the warning and miner keeps mining away. On blagos latest v15.09 I found I had to reset miner at least once a day due to it crashing, tried a loop .bat script but bugged out alot because the crashed miner window would not close and if left all day I would come back to 20+ instances or until windows crashed so had to check up few times a day making sure everything running smoothly.   Burstage, Aug 13, 2015 #73

luxe luxe Gigabyte Burstage said: ↑ I have stagger of 16384 right now with no problems. I have now been running this miner for 24hours and is very stable. Only thing is I get low memory warning even though I still have 40%-50% free, I just close and ignore the warning and miner keeps mining away. On blagos latest v15.09 I found I had to reset miner at least once a day due to it crashing, tried a loop .bat script but bugged out alot because the crashed miner window would not close and if left all day I would come back to 20+ instances or until windows crashed so had to check up few times a day making sure everything running smoothly. Click to expand... For memory issues, please read: https://burstforum.com/index.php?threads/ann-burstcoin-jminer-beta.842/page-3#post-6036

try the following: Code: java -jar -Xmx12G -XX:+UseG1GC burstcoin-jminer-0.3.6-SNAPSHOT.jar if that does not work, in jminer.properties: 'chunkPartNonces=8192' That will force the miner to handle your chunks with 16384 nonces in 2 parts, that can save memory, if you gpu is fast enough.   Last edited: Aug 13, 2015 luxe, Aug 13, 2015 #74

Burstage Burstage Megabyte Testing the memory allocation to see how that runs. Thanks   Burstage, Aug 14, 2015 #75

YepOnlyOne YepOnlyOne Megabyte Hi guys,

thanks for your work on this :)

The program ran fine on Win7 with 32gb ram but had the known memory issues, having it crash eventually. So today I updated to Win10 but the miner won't start (it did start without problem in Win7), I am getting loads of errors looking like in the attached txt file.

I have updated java to the latest version, and the wallet runs without problems on that java version. It is pointed at Dev's v2 pool with about 30TB. I have changed the chunkPartNonces value and tried the above tip without success.

Any idea what can be causing this? The error log mension the deviceID etc but those are correct, I think. I might add that Blago's miner runs fine after the update with no memory problems.

Yup   Attached Files: jminer output.txt File size: 17.2 KB Views: 3 Last edited: Aug 14, 2015 YepOnlyOne, Aug 14, 2015 #76 luxe likes this. luxe luxe Gigabyte The error happens on parsing a plot-file name ... the miner takes your accountId (18446744073709551615) as startnonce, as far as i can see ... Did you rename a plot-file to: 'rnd_accountId_startnonce_numberOfNonces_staggersize' or '_accountId_startnonce_numberOfNonces_staggersize' instead of 'accountId_startnonce_numberOfNonces_staggersize' ?! If you want a plotfile to be ignored, just remove first digit of accountId.

Did not harden parsing the filename yet, following code is executed and may lead to exceptions on bad filename format. Code: String[] parts = filename.split("_"); address = Convert.parseUnsignedLong(parts[0]); startnonce = Long.valueOf(parts[1]); plots = Long.valueOf(parts[2]); staggeramt = Long.valueOf(parts[3]); I will harden that code, to give a speaking exception in future versions ...   luxe, Aug 14, 2015 #77

YepOnlyOne YepOnlyOne Megabyte Hey, thanks for looking :) However, my account is not "18446744073709551615" - I do not know how that figure ended up in the log. I will check if all the plots are named correctly.

Thanks :)   YepOnlyOne, Aug 14, 2015 #78

YepOnlyOne YepOnlyOne Megabyte I checked the disks now and no plot is mis-named. In my "jminer.properties", I found that a disk had gotten a different name (K instead of J), but that did not change anything.

The weird thing is that is was mining fine under win7 (but only with memory issues) but wont mine at all after a upgrade win10 that kept everything intact.   Last edited: Aug 14, 2015 YepOnlyOne, Aug 14, 2015 #79

YepOnlyOne YepOnlyOne Megabyte I found the error. It was a weird plot causing it, but it wasn't the account number. Somehow "18446744073709551615" was the start nonce in an 64GB sized plot. Why I would start plotting with that number, I can't understand, perhaps it was something else from the beginning and I edited somehow. And why it accepted18446744073709551615 when running win7 and not win10 might be one of those MS secrets.

I now get reading speeds over 400MB/s with nvidia 750TI 2GB, and everything seems stable :)

Thanks!   Last edited: Aug 14, 2015 YepOnlyOne, Aug 14, 2015 #80

luxe luxe Moderator YepOnlyOne said: ↑ It almost most be something with win10...

If I download the program fresh on a win7 box and edit jminer.properties, the program runs.

If I make a rar-file out of that whole folder with all files and place it on a win10 machine, I get the errors I uploaded when I try to run it. Click to expand... Thanks for reporting! Will send you a PM.

If someone else has Win10 ... with or without issues, please let us know. Not sure why it should not work on Win10, but if ... it will be solved.   luxe, Aug 14, 2015 #81 Blago likes this. YepOnlyOne YepOnlyOne Megabyte You wrote your reply when I edited my post ;) Here is it:

I found the error. It was a weird plot causing it, but it wasn't the account number. Somehow "18446744073709551615" was the start nonce in an 64GB sized plot. Why I would start plotting with that number, I can't understand, perhaps it was something else from the beginning and I edited somehow. And why it accepted18446744073709551615 when running win7 and not win10 might be one of those MS secrets.

I now get reading speeds over 400MB/s with nvidia 750TI 2GB, and everything seems stable

Thanks!   YepOnlyOne, Aug 14, 2015 #82

Blago Blago Gigabyte luxe said: ↑ Thanks for reporting! If someone else has Win10 ... with or without issues, please let us know. Not sure why it should not work on Win10, but if ... it will be solved. Click to expand... Win10. java 0.3.6 works fine   Blago, Aug 14, 2015 #83 luxe likes this. luxe luxe Moderator YepOnlyOne said: ↑ I found the error. Click to expand... Good to hear! Blago said: ↑ Win10. java 0.3.6 works fine Click to expand... Thanks for sharing that info, i updated, too. No issues with Win10!   luxe, Aug 14, 2015 #84

Burstage Burstage Megabyte win 10 here working. I adjusted the mem down to 8gb on miner and have not encountered the low mem warning and it hasn't impacted on performance maybe 1-2 sec read times but thats fine with me. working great now.   Burstage, Aug 14, 2015 #85

YepOnlyOne YepOnlyOne Megabyte Also works fine on both am old nvidia laptop card (460m) and on Intel's internal graphics cards (tried HD3000 and HD4000). The latter gives reading speeds of 150MB/s over USB - not too bad :)

Newer nvidia cards as 750ti and 980 is also supported.

Good work! :)   YepOnlyOne, Aug 15, 2015 #86

Haitch Haitch Gigabyte I'm running the 0.3.3 snapshot on a small plot on Win10 - I did have to edit .properties though, the platform/device ID's changed when going from Win7 to win10.

H.   Haitch, Aug 15, 2015 #87

Arrow Arrow Megabyte Is it ok to run on Ubuntu 14.10 ?   Arrow, Aug 20, 2015 #88

Irontiga Irontiga Moderator Staff Member Arrow said: ↑ Is it ok to run on Ubuntu 14.10 ? Click to expand... Yes, just make sure you have openCL and java 8 installed, then you'lll be good to go :D   Irontiga, Aug 20, 2015 #89 luxe likes this. luxe luxe Moderator Just want to say, that i will continue development as soon as i find some time, have some nice things in mind ...   luxe, Aug 20, 2015 #90

bsom bsom Megabyte I sometimes get the error "Error: Failed to submit nonce to pool: null" Am I'm not getting credit for some possible submissions because they're not going through?   bsom, Aug 25, 2015 at 7:20 PM #91

bsom bsom Megabyte I think I might be getting these errors because I have some pretty strict garbage collection parameters set. It almost seems like the GC is removing something from memory that it shouldn't be.   bsom, Aug 25, 2015 at 7:23 PM #92

luxe luxe Moderator bsom said: ↑ I think I might be getting these errors because I have some pretty strict garbage collection parameters set. It almost seems like the GC is removing something from memory that it shouldn't be. Click to expand... I don't think it has something to do with GC. bsom said: ↑ I sometimes get the error "Error: Failed to submit nonce to pool: null" Am I'm not getting credit for some possible submissions because they're not going through? Click to expand... Looked it up, i should be able to display some more information on why it failed in future version ... The 'null' just means, that there was no message provided by thrown exception, it may just be a timeout! You could try setting 'connectionTimeout=10000' in jminer.properties. Maybe your commit was just refused by server cause of too much load ...

Edit: re-commit on error is on my todo list, but not implemented yet.   luxe, Aug 25, 2015 at 7:50 PM #93

luxe luxe Moderator [update] pushed sources to github and build a 0.3.6-RELEASE https://github.com/de-luxe/burstcoin-jminer   luxe, Aug 26, 2015 at 11:35 PM #94

Burstage Burstage Megabyte Hi Luxe I encountered a problem after I updated to java 8 60. for some reason my 5 tb running on its own miner will not start, when I click run it opens and closes instantly when I uninstall the java update back to java 8 51 everything is working again. Anyone else having problems with latest java update?   Burstage, Aug 30, 2015 at 3:30 PM #95 luxe likes this. luxe luxe Moderator @Burstage i just updated (x64 version on win8/win10) and i have no issues ...

If miner window closes instantly, there was an error ... you could run it from commandline instead of clicking *.bat, to see that error and provide it here if you like.   luxe, Aug 30, 2015 at 5:15 PM #96

Burstage Burstage Megabyte What do I need to put in the commandline? I'm in the miner dir but not sure what to do next?   Burstage, Aug 31, 2015 at 2:06 AM #97

Burstage Burstage Megabyte I added pause in the .bat and can see the error "

D:\Burst\G burstcoin-jminer-0.3.6-SNAPSHOT>java -jar -Xmx8G -XX:+UseG1GC burstcoin-jminer-0.3.6-SNAPSHOT.jar Invalid maximum heap size: -Xmx8G The specified size exceeds the maximum representable size. Error: Could not create the Java Virtual Machine. Error: A fatal exception has occurred. Program will exit.

D:\Burst\G burstcoin-jminer-0.3.6-SNAPSHOT>pause Press any key to continue . . ."

I have tried different ram size and still get this error but only on java 8 60, on java 8 51 I do not encounter this problem when I run the miner at 4gb.   Burstage, Aug 31, 2015 at 2:13 AM #98

Burstage Burstage Megabyte I have 3 other drives total 1.5tb on 1 miner and with the java 8 60 it runs fine on 1gb ram but only getting problem with my 5tb drive with latest update. I'll just revert back to java 8 51 for now because I can run that for days without needing reset.   Burstage, Aug 31, 2015 at 2:18 AM #99

Burstage Burstage Megabyte Thanks Blago deleted the -Xmx8G in .bat file fixed the error and working on win 10 java 8 60. Thanks   Burstage, Aug 31, 2015 at 4:36 AM #101

Add a custom footer
Add a custom sidebar
Clone this wiki locally