Many plots or a big one?



  • It's not clear to me if there might be a difference when mining in terms of efficiency? Is it better to have one big plot (e.g. 1TB) or several plots (e.g. x5 200 GB)?



  • @vExact i wonder that too cause i have several issues of the plotting getting stopped for some silly reason



  • For me, mining throughput (MB/s read) is exactly the same for

    • a 7 disk stripe with 128 GiB files on them
    • a 7 disk stripe with 839 GiB files (don't ask)
    • a single disk with 1 TiB files

    An optimized 128 GiB file

     size: 128 * 2^30 == 137438953472
    

    has a continous length of

     128 * 2^30 / 4096 == 33554432  (~32 MB or 32 MiB )
    

    that is read on every block.

    If you plotted with a memory size of, say, 2 GiB, and did not optimize the plotfile
    afterwards, the situation is different, and that is totally unrelated to file size.

     2 * 2^30 / 4096 == 524288  ( 512 KiB )
    

    So mining any file with stagger 8192 ( from 512 KiB / 64 Bytes == 8192 ) is dead slow because of head movement;
    Your disk is busy looking up metadata (where is block xy of plotfilefile), move there, read a miniscule 512 KiB, repeat.

    I'd say, as long as your plotfiles are optimized (and laid out sequentially, physically, on the disk) and your stagger size is
    at least 16 MB (*_262144), all is good. That may be different when using a super-fast GPU-assisted miner, but that
    is very hard to generalize. Bigger can be better.



  • Huck says multiple files is the same as 1 file
    Watch around 13:30 onward



  • @vaxman thanks for the input. Yes, all my plots are optimized and I am using 256 GB as a size for each of them.



  • @mrgoldy thanks for sharing the video :)