burstcoin-jminer v0.4.10 - GPU assisted PoC-Miner (All Platforms)
-
@vExact said in burstcoin-jminer v0.4.10 - GPU assisted PoC-Miner (All Platforms):
Hey @luxe I am curious about this property (chunkPartNonces), what does it mean exactly and how it could impact mining depending on the plots we have?
thanks :)I ran a test on my miner using two different capacities. A small one (16TB) and a large one (88TB). Both showed the the lowest scan times when ChunkPartNonces (CPN) was near a value where CPN/TB~4096, I don't use the 320000 default but use powers of 2, so for each of these capacities, 16TB CPN=65.536 and 88TB CPN =524.288. Your mileage may vary possibly by GPU, CPU etc.
-
@rds interesting :) thanks for your input. But why particularly 4096? my drives are all 64kB formatted, could this also need to be taken into account or just the way plots are generated? I use Xplotter btw.
-
Well I should have had first read the description in the jminer.propertis lol
It is related to the stagger size of the plots :)# ----------------------------------------------------------------------------------- # chunkPartNonces - staggerSize defines number of nonces per chunk. # (default:320000) the miner will split chunks in smaller pieces called chunkParts. # this makes sense, to save memory and optimize speed. # in the best case chunkPart#1 will be checked before chunkPart#2 is # completely read ... depending on the power of your GPU. # if staggersize is smaller than chunkPartNonces, staggersize will be used. # e.g. play with +/- 160000 steps
-
@vExact , 4096 was derived empirically. I ran the miner on both sizes, 16TB and 88TB. They were running at the same time. So two instances of JMiner running at once. CPNs were varied for each run (3 rounds average scan time recorded) using CPN values:
16384
32768
65536
131072
262144
524288
1048576 and 2097152.Lowest average scan times using lowest CPN (memory) occurred near CPN/TB=4096.
Why 4096? Maybe because there are 4096 scoops in a plot file, maybe 64*4096 is the amount of memory used to construct one nonce, don't know for sure.
Also, the 2GB CPN crashed the system.
-
@rds so then following your reasoning and trials you made the optimal value for CPN would be the stagger size used for all your (optimized) plots where this value is a number with 2^n, meaning that the GB per plot is also a 2^n number. In any case let's wait for luxe to give his input here :)
-
@vExact said in burstcoin-jminer v0.4.10 - GPU assisted PoC-Miner (All Platforms):
@rds so then following your reasoning and trials you made the optimal value for CPN would be the stagger size used for all your (optimized) plots where this value is a number with 2^n. In any case let's wait for luxe to give his input here :)
No, CPN would be total size of all your plots in the scan directories, as reported by the JMiner at 100%, in TB*4096. I like to push this number to a power of 2, as I believe it is more efficient as memory is structured in binary.
-
@vExact said in burstcoin-jminer v0.4.10 - GPU assisted PoC-Miner (All Platforms):
Hey @luxe I am curious about this property (chunkPartNonces), what does it mean exactly and how it could impact mining depending on the plots we have?
thanks :)Well, the miner does not read one scoop and computes it.
It reads all scoops of staggersize/chunk into memory and than computes that data package.
On optimized plotfiles the miner would read all data of a plotfile first, and than compute it.
(without chunkPartNonces)Therefore there is the chunkPartNonces ... the miner divides the staggersize/chunk by chunkPartNonces to get a number of 'parts' he reads and than computes.
https://github.com/de-luxe/burstcoin-jminer/blob/master/src/main/java/burstcoin/jminer/core/reader/task/ReaderLoadDriveTask.java#L118
So on staggersize 8000 and chunkPartNonces 800 the miner will read and compute 10 parts, one after the other. It can compute part#1 while part#2 is still being loaded (different/independent thread pools) ...
So depending on your system memory / cpu and gpu ... the perfect size of chunkPartNonces may vary.
In general: lower = less memory usage, more cpu usage ... higher = more memory, less cpu usage.Currenty i'm using chunkPartNonces=920000 for my setup.
Edit: btw. if staggersize is smaller than chunkPartNonces ... chunkPartNonces has no effect at all. So it is for optimized plotfiles or plotfiles with high staggersize.
-
@luxe thanks for your reply. Let's say now I have a 2,56Tb mining capacity, consisting of 10 plots of 256GB each (1048576 nonces and staggersize for optimized plots). In this case which would be the optimal setup for chunkPartNonces?
-
@vExact for 2-3TB it does not matter much, just leave it like it is if you have no cpu or memory issue.
-
@luxe hey man... What is wrong with my setup? I underlined the lines with errors... I know it is mining and reading faster than my CPU but I can't see information regarding the Best DL or Last Winner, i tried to use the following addresses as walletServer:
- https://wallet.burst-team.us:8125
- https://wallet.burst-team.us:8125/
- http://wallet.burst-team.us:8125
- http://wallet.burst-team.us:8125/
- https://wallet.burst-team.us:8128
- https://79.168.208.213:8125
With all of them I always got this outcome... I should be doing something wrong, or this functions are not working properly @luxe ?
Here is my proprieties file:
# ----------------------------------------------------------------------------------- # - BURSTCOIN JMINER --------- THIS IS EXPERIMENTAL SOFTWARE, USE ON YOUR OWN RISK! - # ----------------------------------------------------------------------------------- # jminer is a PoC (Proof of Capacity) miner with GPU support for Burstcoin (BURST) # (openCL will also work with CPU in needed) # # Requirements: # - Java8 (64bit recommend to use more memory) # - openCL # # PLEASE DONATE # # jminer would not be possible without openCL kernels and java code provided by 'burstDev', # he really deserves some tips for that! # BURST-QHCJ-9HB5-PTGC-5Q8J9 # # Feel free to support future development of mining engine ... # BURST-LUXE-RED2-G6JW-H4HG5 # ----------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # NOTICE: your 'jminer.properties' hasn't got to contain all properties listed here # as long there is a default/fallback defined, mentioned in here. # ----------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # - PLOT-FILES ---------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # plotPaths - list of plot paths separated with , e.g. D:/,C:/,E:/plots,F:/plots (in one line) # (required) the miner will treat every path as 'physical' drive and use one thread for it # # scanPathsEveryRound - optional 'true' will check 'plotPaths' for changed plot files on every round # (default:true) 'false' will check only on start/restart # # listPlotFiles - optional ... list all plotFiles on start, If walletServer/soloServer is configured, # (default:false) it will show mined blocks and drive seeks/chunks of plotfile, too. # ----------------------------------------------------------------------------------- plotPaths=D:/plots/DONE,E:/plots/DONE scanPathsEveryRound=true listPlotFiles=true # ----------------------------------------------------------------------------------- # - MINING MODE --------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # poolMining - 'true' for pool mining, 'false' for solo mining. # (default:true) ensure to configure the chosen mining-mode below. # ----------------------------------------------------------------------------------- poolMining=true # ----------------------------------------------------------------------------------- # - MINING MODE - POOL ---------------------------------- ONLY NEEDED 4 POOL MINING - # ----------------------------------------------------------------------------------- # NOTICE: # ensure you already setup reward assignment # http://localhost:8125/rewardassignment.html # # numericAccountId - first number in all plot-files # (required for pool) # # poolServer - format is inclusive protocol and port e.g. 'http://pool.com:8125' # (required for pool) # # walletServer - define local or online wallet, to receive and show last winner! # (optional) if empty, winner feature will be just disabled. # format is inclusive protocol and port e.g. 'http://localhost:8125' # online use e.g. 'https://wallet.burst-team.us:8125' # # winnerRetriesOnAsync - number of retries to get winner from walletServer # (default:4) # winnerRetryIntervalInMs - time to wait until next retry to get winner from walletServer # (default:500) # ----------------------------------------------------------------------------------- numericAccountId=9732741443391337776 poolServer=http://pool.ccminer.net:8080 # Winner walletServer=https://79.168.208.213:8125 winnerRetriesOnAsync=10000 winnerRetryIntervalInMs=500 # ----------------------------------------------------------------------------------- # - MINING MODE - SOLO ---------------------------------- ONLY NEEDED 4 SOLO MINING - # ----------------------------------------------------------------------------------- # soloServer - WARN! soloServer should be http://localhost:8125 or http://127.0.0.1:8125 # (default: Solo means you send your PASS on commit results! # http://localhost:8125) DO NOT try to use a online wallet or pool as Server! # # passPhrase - secretPhrase/password of solo mining burst-account # (required for solo) # # targetDeadline - min. deadline to be committed. e.g. 750000 # (optinal) # # triggerServer - emulates open wallet, to prevent it from # (default: false) falling asleep (not sure if needed at all) # # recommitDeadlines - recommits deadlines below 1200 3 times in a interval of 5 sec. # (default: false) experimental feature to increase chance that deadline gets propagated # e.g. on short disconnect or connected peers busy etc. # ----------------------------------------------------------------------------------- soloServer=http://localhost:8125 passPhrase=xxxxxxxxxxxxxx targetDeadline= triggerServer= recommitDeadlines= # ----------------------------------------------------------------------------------- # - OpenCL ----------------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # The miner uses openCL for most of the mining calculations, ensure it is setup correctly. # Instructions can be found e.g. here (thanks cryo): # https://github.com/bhamon/gpuPlotGenerator/blob/master/README.md # You could also use that instruction to find your platformId and deviceId if needed. # # platformId - id of openCL platform on your system. one platform may have multiple # (default:0) devices, the miner currently uses just one (in general not the bottleneck) # # deviceId - specifies the device used by OCLCecker, can be your first GPU, # (default:0) in most cases it will not be 100% used. (depends on capacity) # ----------------------------------------------------------------------------------- platformId= deviceId= # ----------------------------------------------------------------------------------- # - MINING ENGINE ------------------------------------------------------------------- # ----------------------------------------------------------------------------------- # refreshInterval - interval of asking wallet/pool for mining info (in ms), # (default:2000) to check for new block # # connectionTimeout - increase the 'connectionTimeout' on network problems. # (default:12000) this timeout is used for all network requests. # if you use pool or online-wallet, the 12 sec. default may # cause timeout on committing nonces or getting mining info etc. # # debug - setting 'debug' to true will log additional information of the mining process, # (default:false) that are not related to mining, but to miner internals. # # writeLogFile - setting 'writeLogFile' to 'true' will write all logs from console to a file, too. # (default:false) the name of that file can be specified by 'logFilePath'. # # logFilePath - path (filename and optional directory, relative to miner location) # (default:log/jminer.log.txt) # ----------------------------------------------------------------------------------- refreshInterval=1000 connectionTimeout=60000 debug=true writeLogFile=true logFilePath=C:\Users\Goncalo\Desktop\BURST\software\jminer # ----------------------------------------------------------------------------------- # - MINING ENGINE - APPEARANCE / BEHAVIOR ------------------------------------------- # ----------------------------------------------------------------------------------- # readProgressPerRound - defines how often the mining progress is shown per round # (default:9) thats the 'xx% done ...' info. # # byteUnitDecimal - switch between decimal units (true): TB/GB/MB (divided by 1000), # (default:true) or binary units (false) TiB/GiB/MiB (divided by 1024) - https://en.wikipedia.org/wiki/Byte # # showDriveInfo - set this to 'true' to show info about every drive on finish reading it, # (default:false) this is useful to find the slow ones ... can help to optimize your setup. # # showSkippedDeadlines - set this to 'true' to show found deadlines below targetDeadline from # (default:true) this config or provided by pool (overwriting the targetDeadline specified in here) # ----------------------------------------------------------------------------------- readProgressPerRound=5 byteUnitDecimal= showDriveInfo=true showSkippedDeadlines=true # ----------------------------------------------------------------------------------- # - MINING ENGINE - MEMORY USAGE ---------------------------------------------------- # ----------------------------------------------------------------------------------- # chunkPartNonces - staggerSize defines number of nonces per chunk. # (default:320000) the miner will split chunks in smaller pieces called chunkParts. # this makes sense, to save memory and optimize speed. # in the best case chunkPart#1 will be checked before chunkPart#2 is # completely read ... depending on the power of your GPU. # if staggersize is smaller than chunkPartNonces, staggersize will be used. # e.g. play with +/- 160000 steps # # readerThreads - normally '0' means, the miner takes one thread per drive (plotPath) this is recommend. # (default:0) choosing a other number of 'readerThreads' can be useful on memory issues. # For example, if you mine on 4 drives (plotPaths), you can reduce the memory usage # by setting 'readerThreads=2', this will reduce mining speed but save memory. # ----------------------------------------------------------------------------------- chunkPartNonces= readerThreads=
-
if you got local wallet running try pointing to it (http://localhost:1825)
-
@LithStud Yeah I have not a local one in this machine... ;D
I may install it tho. But it is saying that the burst-team wallet is out of sync when in fact it isn't lol
-
@gpedro the IP one didnt work for me at all so no wonder there. Not sure why others doesnt work tho (thus recommendation to try local and see what happens :D )
-
I'm running a AMD r9 280x with 16.11.5 and AMD SDK 3.0
I tried 0.4.9 and 0.4.10 - without any luck - my Card never gets detected at the first plattform overwiew. What is wrong here? Any suggestions?
Windows 10 x64
i7 4770
r9 280x
-
@gpedro you can leave 'walletServer=' empty (disable infos) or try 'walletServer=http://burstcoin.cc:8889'
not sure, but https may be a issue choose wallet without if thats the case ... at burstcoin.cc you will find some.
-
@Zaziki If you know platform and device just specify it manualy ...
e.g.platformId=2 deviceId=1The platform overview is just a info and maybe not detects exotic setups?! not sure ...
May also be a driver issue. Have you checked opencl-z if it shows up there?
-
@luxe that worked hehehe
thanks bro!
-
@gpedro Well guess i have to check why https does not work ... i expected miners to use localhost :-) Never tested something else ...
-
@luxe yeah I thought so but I still didn't got the local wallet running on this PC hehe
-
got it Running, but for now pretty slow.
I have about 92TB, Read times ~2min 34s
With Blago i need about 52s in i7 4770.there must be sth wrong.
All my plots are optimized and One File per Drive.
I Read that r9 280x only Support opencl 1.2, is there maybe the Problem?


