How do you know if you can GPU plot/mine
-
@rds The first error was from you needing to run jminer from the directory in which the JAR resides, which you rectified on your second attempt. The second error is jminer looking for a 64-bit version of java (since you specified -d64), which you might not have installed. I install the 64-bit version separately, and then point jminer to it:
"C:\Program Files\Java\jdk1.8.0_60\jre\bin\java.exe" -jar -d64 -XX:+UseG1GC burstcoin-jminer-0.4.8-RELEASE.jar
-
Please check out my post in this thread: https://forums.burst-team.us/topic/24/burstcoin-jminer-v0-4-8-gpu-assisted-poc-miner-all-platforms/61. That solved my java problems earlier so that I could run Jminer. It is a while ago though, so maybe the version numbers are different now.
-
Thanks guys, just got back from golf, I will play with it soon.
-
Thanks to all the input I received from you guys, I was successfully able to start the jminer. I had to install the 64 bit Java, never realized I was using the 32 bit before. I went through the properties file and configured everything to the best I could. Here is a screenshot of the miner:
So, it appears the jMiner is about the same speed as the Blago CPU miner.
I thought the GPU was a lot faster. At least with plotting I see speeds on the forum like 10k nonces/min where I CPU plot at about 3900 n/m. Possibly because my GpU is integrated and only 1GB, but it does have 24 threads/channels.
Are there parameters I need to tweak to get more speed?
for instance,
chunkPartNonces=
readerThreads=I left them at default. 320,000 and 0.
Also my
platformId=0
deviceId=0It looks like the deviceid=0 is the GPU and the deviceid=1 is the cpu. So I left it on GPU.
I says the GPU is 1GB and 24 threads and the workgroup is 256. The CPU is 17GB and 4 threads and the workgroup is 8192.
Thanks again to all.
-
@rds Glad to hear you got it working. You could always try the other device, to compare speeds. Back when I used to use my integrated GPU I noticed a few seconds difference, but not much. I had about 7 TB then. The difference becomes much bigger when you have larger plot capacity! I left my Jminer settings on default, but maybe @luxe knows a trick or two?
If you haven't optimized your plot files yet, you should do that. That will definitely give you better read times.
-
Thanks, I did try to use the device 1 which is my CpU and it seemed to crap out. I closed the instance, did want to mess anything up.
All my files are optimized.
How about the default chunckpartnonces=320000? I played with that but didn't notice much change.
Anyone know why 320000 is default? I thought every parameter associated with burst plots was better as a power of 2, e.g. 262144 or 524288.
-
@rds @Propagandalf Here are some settings I tweak:
scanPathsEveryRound=false
(no need to re-scan for plots unless you are plotting while mining)targetDeadline=86400
(no need to report 24+ hour long deadlines)As a comparison, I'm running an old AM3+ system with an FX-4300 and a GTX 950 with about 70TB of optimized plots and jminer reads them in about 27 seconds.
-
@sevencardz said in How do you know if you can GPU plot/mine:
@rds @Propagandalf Here are some settings I tweak:
scanPathsEveryRound=false
(no need to re-scan for plots unless you are plotting while mining)targetDeadline=86400
(no need to report 24+ hour long deadlines)As a comparison, I'm running an old AM3+ system with an FX-4300 and a GTX 950 with about 70TB of optimized plots and jminer reads them in about 27 seconds.
I set my targetDeadline to 10000. I think that FX-4300 must have more power than an integrated Intel HD 5000 with 1GB of RAM. My 70+ TB takes 55-65 sec to scan.
I think after running it for awhile now, it is slightly faster than CPU. But when scanning it makes the computer laggy. CPU mining did not do that. Nothing terrible, just an observation.
I like the idea of having both methods available for use. I also run two wallets on different ports for even more diversity.
-
@rds Sounds like you're pushing that machine for all it's worth. :) And yep, the GPU miner and plotter can put a big hit on your response time. I don't feel it with the GTX 950 much while mining, but it will crash the driver and sometimes the entire system if I try to plot with it and run graphics at the same time.
Another thing you could try would be to turn down Windows graphical settings as low as they will go. I doubt you'll see a huge performance difference though. http://www.thewindowsclub.com/disable-visual-effects-windows
-
@sevencardz said in How do you know if you can GPU plot/mine:
@rds Sounds like you're pushing that machine for all it's worth. :) And yep, the GPU miner and plotter can put a big hit on your response time. I don't feel it with the GTX 950 much while mining, but it will crash the driver and sometimes the entire system if I try to plot with it and run graphics at the same time.
Another thing you could try would be to turn down Windows graphical settings as low as they will go. I doubt you'll see a huge performance difference though. http://www.thewindowsclub.com/disable-visual-effects-windows
This little laptop is working like a dog. I also am plotting with the Xplotter. I have about 16TB more and I will be done, for awhile :).
BTW, any insight on what "blocks" =N/A or 1 or 2 means? This is the list of files when the miner starts up.
-
@rds I think that's showing which of your plots have forged blocks so far and N/A just means none were found there yet.
-
Thanks, interesting did not know that.
-
@rds exactly what @sevencardz said. If not OPTIMIZED! ... it will show the number of needed drive seeks ...
gpu or cpu ... if the computing is not the bottleneck but your read speed due drives not attached to use their full speed, it will limit you ... e.g. if you attach 10 drives to one USB3, USB3 will limit your speed ... can be lot of things. I always try to find the current bottleneck and than improve/fix it :-)
not sure if you noticed the setting 'showDriveInfo=true' already, it could help you find slow drives.
