XPlotter for optimized plots (CPU)
-
@Blago said in XPlotter for optimized plots (CPU):
@Focus 8)
44 threads, 200 Rb RAM = 50600 nonces/min = 4Tb by ~5 hrs
it's speed of 10 CPU FX-63002 hours in, it is at 37%
-
@Focus When plotting done, can you try again and send to me screenshot with crashes (-mem 250G) ?
-
@Blago said in XPlotter for optimized plots (CPU):
@Focus When plotting done, can you try again and send to me screenshot with crashes (-mem 250G) ?
It wasn't really a crsah but more of the program being stuck.
It would do this:
Plot in RAM the first 250GB, start writing to disk but not fully finish, next block would start writing to RAM before first one finished and get locked up.I know it's supposed to work that way but for some reason it would just get stuck
Settings: 500000 stagger, 40 threads
-
@blago I am 7 hours in and it's at 77%
-
Completed 4TB in 12.01 Hours
-
This post is deleted!
-
2 e5 seriers xeons with a total of 48 threads, I wanted the 20 core cpus but couldn't get them !
-
reading your post, I'm plotting 1 file 797Gb and so far I'm at 59% and I'm 19 hours later :D
-
If it was my system it would be done. I had 4 threads and 4096 memory, CPU about 2.5 GHz.
-
This post is deleted!
-
@Blago what is the best allocation unit size when formatting a drive for later plotting? I chose the default one, but I am afraid it performs much more slower than choosing the maximum allocation unit size. Any input on that?
-
-
@Blago thanks! That's why it has taken an eternity for me to plot only 2 TB, I don't know if it deserves the pain to start over and do it again... Can it have an impact on mining after if I keep plotting with an small unit size allocation?
-
@Blago I am doing 10.5k nonces/min, but the HDD writing scoops is taking an eternity!
-
@vExact said in XPlotter for optimized plots (CPU):
Can it have an impact on mining after if I keep plotting with an small unit size allocation?
No, see results https://ejrh.wordpress.com/2012/10/26/cluster-size-experiment
I am doing 10.5k nonces/min, but the HDD writing scoops is taking an eternity!
start plotting with Admin's rights?
for me writing 6*1024 nonces takes ~30 seconds
-
@Blago yes with admin rights.
-
@vExact said in XPlotter for optimized plots (CPU):
@Blago what is the best allocation unit size when formatting a drive for later plotting? I chose the default one, but I am afraid it performs much more slower than choosing the maximum allocation unit size. Any input on that?
I have read exactly the opposite in another thread https://forums.burst-team.us/topic/1900/large-scale-burst-mining-operation. Here, they put forward that 64k cluster size may make the write speed slower, and that using the default cluster has fixed it, according to @nixxda. I am currently plotting 8TB drives using gpu plotter in buffer mode with 64k cluster, and it started fast, but has since dropped almost to a halt (10 MB/s). This is a different plotter than xplotter of course. I have not been able to test with default cluster size yet.
@luxe is it possible to run gpuplotgenerator.exe with admin rights through a batch file, and will it make any difference?
2Create a unique Burst account ID with Luxe's vanity generator: https://github.com/de-luxe/burstcoin-address-generator/releases
BURST-E2NW-8UXH-92YR-BALLS
one_reply_to_this_post
-
@Propagandalf thanks for your input. I think the answer was given before with the link @Blago provided https://ejrh.files.wordpress.com/2012/10/copying1.png
If you have big files to copy it is recommended then to use 64kb as allocation unit size for the drive, and the contrary applies to moderate or small files. Although I am not sure now how does this apply here, as each nonce has a small size... According to @Blago it is recomended for Xplotter.exe to run it as administrator, as it goes faster. Just right clicking the batch file and run it as administrator :)
-
The other discussion was about SMR disks specifically. And as I remember my tests about cluster size were not conclusive! SMR's are using the periphery of the disk as an fast buffer (non smr) and everything inside is smr (overlapping). I read somewhere that all of this (and the controller) is very likely optimized for the default cluster size!
But we didn't have Xploter at the time! I might replot one of these disks yust to see how it "behaves"!-)
-
@Blago Wow 1.0 is way better than the GPU plotter finally I can have optimized plots good work.
Edit: Just finished plotting 6TB in one big plot file took 35 hours.
