Cold-cache Sequential IO Benchmark

Discussion in 'backup, imaging & disk mgmt' started by angstrom, Jan 8, 2017.

  1. WildByDesign

    WildByDesign Registered Member

    Joined:
    Sep 24, 2013
    Posts:
    2,587
    Location:
    Toronto, Canada
    This new 64-bit build is working great here in my testing so far. :thumb:
     
  2. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,945
    Location:
    The Pond - USA
    @angstrom - the 64-bit version works just fine in a PE10 configuration, should work without issue in PE3.1 or PE5.

    Thanks much for the effort!
     
  3. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,945
    Location:
    The Pond - USA
    @angstrom - quick question. Does this benchmark run under Windows XPx86? I've tried it a coupla times and it fails with msg in the LOG... which I can't seem to find anywhere :confused:

    Any help, greatly appreciated...
     
  4. guest

    guest Guest

    I don't think so.
    website: "Requires: Windows Vista or newer"
     
  5. TheRollbackFrog

    TheRollbackFrog Imaging Specialist

    Joined:
    Mar 1, 2011
    Posts:
    4,945
    Location:
    The Pond - USA
    Thanx, @mood , I didn't notice that.
     
  6. angstrom

    angstrom Registered Member

    Joined:
    Oct 27, 2013
    Posts:
    126
    Location:
    Switzerland
    Never tested it on XP to be honest. Tried just now and it fails in NtSetSystemInformation() call when trying to clear the standby list (you can see the log by starting the program with "-d" switch). This effectively means that there's no reliable way to purge the file system cache before testing, so the results will be skewed and the whole thing will be largely pointless.

    That said, you should be able to switch off cache purging by changing test type to Custom, and then setting "Cache control" to "Leave cache alone". That's in theory. In practice however you need to set it to the exact opposite - "Purge cache before each test" - because I messed up and put options there in the reverse order. I'll patch it tomorrow.

    Edit: regarding this -
    To be more precise, any tests of the "buffered / cached" variety will be pointless. The "unbuffered" bunch should be OK... at least in theory.
     
    Last edited: Apr 13, 2019
  7. treehouse786

    treehouse786 Registered Member

    Joined:
    Jun 6, 2010
    Posts:
    1,411
    Location:
    Lancashire
  8. plat

    plat Registered Member

    Joined:
    Dec 19, 2018
    Posts:
    2,233
    Location:
    Brooklyn, NY
    ccio42019.PNG
    Thank you for making this tool available. On a two year old 250 GB M.2 drive w/98% drive life remaining, in Good health via CrystalDiskInfo. Results are somewhat compatible with Passmark's https://imgur.com/QWgX2XM Disk Mark, read status, though not write. The driver and firmware are the latest by Samsung but I believe the write speed is rather below what is advertised (1800). Things are snappy regardless so I think it's got a ways to go.

    Edited for clarity. Edit: searched for below-advertised write speed for the Samsung EVO 960 but haven't come up with anything useful (yet).
     
    Last edited: Apr 14, 2019
  9. angstrom

    angstrom Registered Member

    Joined:
    Oct 27, 2013
    Posts:
    126
    Location:
    Switzerland
    CCSIO Benchmark 1.4.3 is out

    Changes:
    • Fixes an issue with "Cache control" options in "Custom test" being in reverse order.
    • Volume serial number is now masked out from the screenshots.
    • Also, this update pulls in all the changes in libraries it shares with bvckup2 that accumulated since August last year.
    32-bit version is on the homepage - https://ccsiobench.com
    64-bit version is over at https://ccsiobench.com/files/ccsio-bench-1.4.3-x64.exe
     
  10. Zero3K

    Zero3K Registered Member

    Joined:
    Mar 28, 2004
    Posts:
    380
    Location:
    Louisville, KY
    Try using the method suggested at https://stackoverflow.com/a/7113153/8067355 instead of using NtSetSystemInformation.
     
  11. angstrom

    angstrom Registered Member

    Joined:
    Oct 27, 2013
    Posts:
    126
    Location:
    Switzerland
    You can already do that by setting cache control to "leave the cache alone" and then running unbuffered tests only.
     
  12. Zero3K

    Zero3K Registered Member

    Joined:
    Mar 28, 2004
    Posts:
    380
    Location:
    Louisville, KY
    That method will flush the cache.
     
  13. Zero3K

    Zero3K Registered Member

    Joined:
    Mar 28, 2004
    Posts:
    380
    Location:
    Louisville, KY
    The build you linked to me in an email is crashing under Windows Server 2003.
     
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.