CreepMiner very slow



  • Re: creepMiner - C++ Burst Miner (based on Uray's Miner)

    Hallo!
    My Problem ist, that the creepMiner is very slow.
    0_1506063933819_7bb7a871-c514-45cc-92fb-0a4789a6a36c-image.png

    Here some Info about my version:
    0_1506063988505_6488c7d3-72dc-40c8-94c2-1bc8adf74292-image.png

    Here some hdparm drive speeds :
    0_1506064106252_ea85725d-9550-4083-9b92-68823138e43d-image.png

    Drive h is connected via SATA3 and Drive n is connected via USB 3.0

    System is a i7 2600k with 16GB RAM and Ubuntu 16.04

    I honestly have no idea, why the hack it is so slow.



  • I never used that version, but you could consider upgrading - the latest version is 1.6.3, so you're a couple of versions behind. The release notes for 1.6.0 do say that that plot reader was optimized - not sure by how much though.

    Your read speed does seem incredibly slow though, could you have too many drives attached to one USB controller? I'm using a similar setup, and currently getting a plot read speeds of between 50 and 120 MB/s according to my log file.

    [EDIT] Oh wait, you're getting the same speed on a SATA drive too? Buggered if I know, sorry.


  • Mod

    This post is deleted!


  • @Zaziki

    I'd suggest turning down your max plot readers # to test. Each is a thread and unless you have the bandwidth and CPU to actually scan 14 devices at once AND verify with three threads you're going to be going nowhere fast.

    I have a mix of SATA and USB drives on my machines and I get the best overall processing speed (read + verify) with a plot reader number of 2-3. If you go too high you'll get seemingly faster reads but verification will lag and add overall time. In your case I think it's WAY too high.

    Also Increase your buffer size a little if you have the ram 1024MB for 60TB would be good, 128MB is choking your read times I think, as the CPU maxes to fill the tiny buffer, then goes to verify, and then has to wait to fill the buffer again. It's completely overloading your CPU as my first guess, in a terribly inefficient way.

    Experiment!

    Also as Tinycoins said, get the newest version (1.6.3), it has a scan bar that differentiates between read data and verified data that makes it easier to visualize and optimize the process.

    Good luck!



  • @captinkid

    Thanks!

    I changed it to about 8gb of ram and 4 threads - this seems to work quite good.


Log in to reply
 

Looks like your connection to Burst - Efficient HDD Mining was lost, please wait while we try to reconnect.