burstcoin-jminer v0.4.10 - GPU assisted PoC-Miner (All Platforms)
-
After much trail and error, I have finally managed to run jminer on a headless linux machine that have 2 GPU's. I tried to change the parameters of the config file, that the miner will use more GPU power (currently only one GPU is used at 13% capacity). Any advice on how to maximized the power of the CPU and GPU combined ? Platform ID and device ID ?
-
@CoinBuster Currently only one openCL device is possible. In most cases GPU is not the bottleneck, so this is not a big issue. Not sure if crossfire linked GPUs are used/show up as one device?! However you could always split up your drives to multiple miner instances and let every single instance use a dedicated device. Also, if there is need for that feature, the code should be adjustable quite simple, never thought about.
Do you currently plan a rig, where you expect 100% GPU load ?! :-)I created a issue on github, will look into it as soon i find some time ...
https://github.com/de-luxe/burstcoin-jminer/issues/27
-
@luxe I really don't know if the crossfire linked GPUs show up as one device....that's an interesting question. The GPUs on the headless machine work separately, and I need to look into the documentation if it is possible to crossfire them. I have plotted 10TB on the cloud, and I am testing the speeds in which I can read them.
At the moment, I am running a headless machine with 10GB network connection, 16 vCPU 22 RAM and 2 GPU. Reading speeds are good at 100mb/s, but I have reached higher speeds (machine with 32 vCPU, 128 RAM and 10GB connection).The speed was at about 160 Mb/s and all 32 vCPU were maxed to the 100%.
I am trying to reach the read speed limit. Do you have any advice on how to play with the config file to achieve this ?
-
@CoinBuster What read speed limit?! Of your internet connection?! Speeds may be limited by cloud server ... on 10TB i'm quite sure that not GPU/CPU or your System is the bottleneck.
Check this post https://forums.burst-team.us/topic/2411/everybody-post-their-mining-rigs-show-photos/86 i'm just using one 280x ...
-
@luxe Exactly, I want to test this out. I am not sure where is the bottle neck if 32 vCPU went all the 100%while reading the plots. With the GPU assistance they only use 40% and the GPU 13% but I'm sure I can maximize it more to reach much higher reading speeds.
If I ran the same setup and reach the limit at 160mb/s I'll know that this is the cloud provider bottleneck.
-
@CoinBuster Speed provided by your cloud provider may vary depending on his current server load ... so may have nothing to do with jminer at all.
-
@luxe Maybe there was extra load at that time. I have to check again. Is there any chance to change the CPU GPU config with jminer ?
-
@CoinBuster Yes, even i see no big need for it :-) Like i told above, i already created a issue for it, to remind myself ... https://github.com/de-luxe/burstcoin-jminer/issues/27
But could take a week or two.
-
@luxe Thanks buddy I'll sure send some burst in your direction soon. Keep an eye out ;-)
-
@CoinBuster It's more difficult than i thought to add multiple openCL devices with same miner instance ... the code is just not prepared for that, yet. So i fear there will be no quick fix.
-
@luxe Ok. Thanks for looking into that. Is it make sense to you that I mine faster on a VPS with 16vCPU then with a VPS with Nvidia K80 GPU ? it's seems that i have a GPU bottleneck ?
-
@CoinBuster Not sure that Nvidia K80 seams to be a beast ... maybe the openCL drives for it are not very well. Guess must who have such hardware use cuda ... like discussed above, i do not think that cpu or your GPU is bottleneck, have you monitored it?! Is it really on 100% while mining?!
-
I moved my plot files to a NAS and started getting this error:
"ERROR 2304 --- [ readerPool-2] b.j.c.reader.task.ReaderLoadDriveTask : IOException: Bad file descriptor"Any ideas on how to troubleshoot or fix this error?
Thanks in advance!
-
@Rev Was there a stacktrace/more details on that error? I searched for that error, the only thing i found, that it may be related to file permissions ... so ensure the user executing the miner has needed permissions on your NAS, but i have no idea how to configure that, maybe you can figure out.
ps: sorry for late response, i simply missed your question, maybe use @luxe next time, so i get messaged :-)
-
@luxe Thanks for response. I had to increase the "chunkPartNonces" to 960000 or above to solve the problem with wifi connection, cabled 1G ethernet to NAS was no problem.
-
@Rev Well, glad you found a solution ... even if i do not understand why changing that makes a difference in this case :-)
-
burstcoin-jminer-0.4.6-SNAPSHOT
Download:
https://github.com/de-luxe/burstcoin-jminer/releases
Updates:
- 2nd best deadline will get queued and committed, in case that best deadline does not get confirmed. Or in other words, best not confirmed deadline will no longer block next best from getting committed. This should ensure to get as much shares on pool mining as possible.
- Updated libs to current versions (including latest jocl lib, hopefully not causing older gpu's not working 'again')
- Some minor changes to prevent useless / obfuscated error messages
-
Ok I heard about this J-Miner and I do have a Gpu so, that leads me to ask myself "why am I using the Cpu-Miner"? Maybe because this is what I heard to use when first learning. I heard that the 'Blago' Cpu-Miner was the best. But I would like to try out J-Miner as I do have a Gpu. I have a Nivida 1080, should this work fine and is it as easy to set-up as the Blago?
This is what made me want to try out the J-Miner - "GPU-MINER "J-Miner' > sCANS faster than a CPU-miner WHICH gives you an advantage to report to the 'pool' quicker". I read this somewhere here I believe, can't remember.
So, is 4.6 the latest release and do we really need to download the CL-Driver - http://jocl.org/ and of course the latest Java, correct?
Thx y0u!
-
@tradz GPU jminer works faster then CPU miner
-
@Blago Thx f0r the response but do you think you 0r anyone reading this with the knowledge, can answer my questions above. Starting with "is it as easy to set-up as Blago"?
I guess instructions will be given inside the downloaded file, correct? And is 4.6 the lastest release?
Thx

