600 Hours to Plot
-
I am just getting into burstcoin mining.
I tried to get gpuPlotGenerator (https://github.com/bhamon/gpuPlotGenerator) working on linux with an Nvidia GTX770 but I always get '[ERROR][-36][CL_INVALID_COMMAND_QUEUE] Error in step2 kernel finish' in short order (even after multiple driver/distro changes).
I have moved to mdcct (https://github.com/Mirkic7/mdcct) and started plotting successfully, however it indicates it will take 600 hours (25 days) to finish plotting my 64TB raid0 array.
I calculate that my raid array should be able to be fully written in around 9 hours in an ideal scenario so there is some serious CPU bottlenecking going on. Note: I am using a brand new i7 7700k cpu and am utilizing all 8 threads (all 8 virtual CPUs show as 100% usage).
I was wondering if it would be worthwhile to to mount my raid array as a network drive and use my Windows 10 gaming PC to crunch through the hashes.
Does anyone know roughly how much of an improvement I could expect to get from using a GPU plotter?
-
@nathanowen42 First, don't do it as a single big RAID0 array - the miners are multi threaded and will mine 8 * 8TB drives simultaneously much faster than a single 64TB raid array.
Given your CPU, I'd look at installing Windows on a temporary boot drive, use xplotter to plot them, then boot back into Linux and mount the NTFS drives using jMiner, or just leave it running Windows and use either jMiner or Blago.
-
Is there an advantage to using xplotter over mdcct?
-
already optimized plot. doesn't make your cpu like a stove
-
@nathanowen42 What @HiDevin said - right from the get go all plots are optimized with none of the overhead that the GPU Plotter requires.
-
Okay, I have pulled the raid card, switched to Windows 10 and set everything up. Do you think it is faster to run xplotter with 8 threads one drive at a time, or would it be faster to run 8 separate xplotter instances (1 per drive) with one thread each.
-
depends, does one drive slow down/on the writing scoops part, it goes gray?
-
I went ahead and tested plotting with 8 separate instances. It does seem to be a little faster overall (about a 7% improvement). Think I'll leave it to plot. Thanks for the help guys, looks like we have gotten it down to around 12 days which is a reasonable improvement.

