plotting speeds vs ram used

Feel free to check out our Forums by clicking the link at the top menu of the site, we have over 10,000 Active members in the community and growing every single day! We also have the most FEATURE RICH live chat room we call "The Trollbox" where you can jump right in and talk with other Cryptocurrency enthusiasts who are also online.

  • Ok so i have been testing plotting speeds vs amount of ram used to plot. I have found on my systems that plotting at around 1/4 total memory yields higher speeds. I tested using identical computers one had 5gb ram and the other has only 2gb. so i had to plot with lower ram speed on the 2nd computer only to find that it plots 2-3 times faster. So I tested in the one that has more ram using a lower stagger and got the same results. It plots faster also using less ram! Has anyone else had this happen?

  • Samaus,

    I have to agree with with rds on this one. Xplotter may need a little more finesse to get to work on a Linux box vs Windows.

    But in the end, it all comes down to how quick can your CPU do the calculations vs how fast it can get it written to a drive and here size does matter.

    My current stats:

    Core i5 4690 desktop PC

    OS: Windoze 8.1 (x64)

    OS drive: OCZ 128GB SSD

    Chip: i5 4690, 256k L1; 6MB L3 cache (Haswell), 3.5GHz, 4 cores / 4 threads

    RAM: 16GB

    GFX card: ATI Radeon R280

    SATA ports: on-board

    Plotter: XPlotter

    Plot settings: 4 cores, 10GB RAM

    Plot summary: Local drive averages around 8000/min

    Time for 4TB: roughly 2.5 days.

    Time for 1TB chunks: 6-8 hours per chunk - 32 hours total

    ML310e Gen8v2

    OS: Ubuntu Server Core, 16.04.3 LTS (x64)

    OS drive: 1TB Seagate server drive on raid controller

    Chip: Xeon E3-1220 v3, 128k L1; 8MB smart cache (Haswell), 3.1GHz, 4 cores / 4 threads

    RAM: 16GB

    GFX card: N/A

    SATA: HP P222 RAID controller with 512MB Cache BBWC

    Plotter: omdxxt/plotavx2

    Plot settings: 4 cores, 4/12GB RAM (Calculated to match nonce count)

    Plot summary: Local drive averages around 3600/min

    Time for 4TB: roughly 3.5 days.

    Time for 1TB chunks: 6-8 hours per chunk - 32 hours total

    What is weird but kind of obvious is the fact that although the Xeon processor is by far superior to the Core i5, and has little to no OS overhead (Linux vs Windoze), I have to investigate a little further but I believe it may be swap file handling related. Although Linux uses as much RAM as possible before it swaps; WIndoze uses as little as possible RAM and swaps almost everything. This in my mind will add strain to the already laden windows kernel. And, above this, Windows uses both a swap file and a page file. As the one computer stores these files on SSD; this may very well be where the boost comes in.

    Hope it helps a bit!

    If you like what you read, remember to like.

    If you don't like what you read, just to bad - we all entitled to agree to disagree.

    If you really like what advise I have given please feel free to share:

    BURST-9QN4-Q3UZ-MC38-AN255 - Any size, any time. Big or small, all are welcome!