Unusually slow OWC Mercury Accelsior

Discussion in 'Mac Pro' started by SimeoneSergio, Dec 31, 2014.

  1. SimeoneSergio, Dec 31, 2014
    Last edited: Dec 31, 2014

    SimeoneSergio macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
    #1
    Mac 5,1 Xeon 3,33Ghz
    24Gb Ram 1333Mhz
    Mercury Accelsior 240Gb
    Yosemite 10.10.1
    Latest OWC drivers installed properly

    The machine has hmm 2 years and a half. Software: BM Speed test.

    I remember having both read and write values way above 600 Mb/s (with Mountain Lion)...
    Now, read speed is always fine (600 and above) but write never goes above 300.

    What can it be?

    Thanks for your help!
     
  2. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
  3. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
  4. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
  5. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
  6. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
  7. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
  8. bxs, Jan 1, 2015
    Last edited: Jan 1, 2015

    bxs macrumors 6502

    Joined:
    Oct 20, 2007
    #8
    Doesn't BlackMagic write compressed data ? This maybe the reason. Did you measure write performance under Mountain Lion using same BlackMagic version as you're using today ?

    Try using other tests such as AJA and even use the dd command in Terminal.app such as

    dd bs=1000000 count=1000 if=/dev/zero of=~/Desktop/test_file

    I have a MP5,1 with the OWC's 960GB Mercury Accelsior PCIe SSD running Yosemite 10.10.1 and ran a test or two read/writing to the SSD. See attached pics.
     

    Attached Files:

  9. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
    #9
    Well i'm not sure of that, being completely honest.

    Went for AJA, and with 8Gb file size it dishes out 652.9 in write... And in read 629...

    Fine for me, just wondering why the write speed now is higher than the read one :)

    Thanks for your help folks!
     
  10. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
    #10
    Regarding your pics: i'm afraid they can't be displayed properly :confused:

    EDIT: one thing that now is different from yours and mine is that on the system information under SATA/SATA Express yours it is recognized as "OWC MERCURY ACCELSIOR" while on mine it simply is "MARVELL Raid VD0".

    Not sure what this does imply...

    And well, my rMBP goes up to 970Mb in write mode using the latest Blackmagic disk speed so i really dunno what to say :S
     
  11. bernuli macrumors 6502

    Joined:
    Oct 10, 2011
    #11
    How long has the Accelsior been in service? 2.5 years?
     
  12. bxs, Jan 1, 2015
    Last edited: Jan 1, 2015

    bxs macrumors 6502

    Joined:
    Oct 20, 2007
    #12
    Hmmmm.... "MARVELL" sounds like you possibly have the OWC 2x 240GB SATA SSDs setup as RAID-0.

    I'm sorry about the pic quality. I ran the test on a remote MP5,1 using LogMeIn and grabbed some screen shots. This is the reason they are poor quality, and there's not much I can do about that. However, I'll try a different approach and re-post them.

    Where did you purchased your PCIe SSD from ?
     

    Attached Files:

  13. SimeoneSergio thread starter macrumors regular

    Joined:
    Jun 27, 2012
    Location:
    London UK
    #13
    @Bernuli: Yes, it has been in service for like 2.5 yrs.

    @bxs: i've purchased it from an official european reseller, now i don't remember the name. Anyway, the AJA test displayed proper speeds, so i think i can be satisfied.

    Regarding the pics: i was meaning that i couldn't click & display them, not the low resolution.

    Other than this, thanks for your invaluable help folks! :)
     
  14. HHarm macrumors regular

    Joined:
    Mar 4, 2009
    #14
    Digging out an older thread.....

    My 480Gb Accelsior was serviced by OWC and now I got it back. I don't know the prior results but the Black Magic test (with 5Gb files) gives me 433 MB/s write and 490 MB/s read freshly formatted. If I understand correctly I should be getting more in the range of 600-700 MB/s. The card is in slot 2. Is there something abnormal about the result and if there is what should I be looking into?
     
  15. AidenShaw, Apr 12, 2016
    Last edited: Apr 12, 2016

    AidenShaw macrumors P6

    AidenShaw

    Joined:
    Feb 8, 2003
    Location:
    The Peninsula
    #15
    in ge
    How'd you try TRIM? Simply enabling TRIM won't help - you need some activity on the disk to get chunks marked for trimming.

    Enable trim. Delete unnecessary files. Run a disk utility that trims free space (or create some big scratch files and delete them). Let the disk sit powered on overnight for the garbage collector to do its thing. (Trim doesn't replace garbage collection - but it makes garbage collection much more efficient.)

    Then test.

    Also, don't lose sight of the fact that almost all SSDs have a pool of free space, and can write very quickly to that pool. When that pool runs out, writes have to wait for the garbage collector to create more free space. This can be tens or hundreds of times slower than writing to free space. Don't be surprised if your first tests on a drive are fast, but after repeated tests it markedly slows down. That's by design. (Trim helps this, and having plenty of reserved space helps. Having free space in a filesystem in general does not help unless you have trim.)
     
  16. transcendental333 macrumors newbie

    Joined:
    Apr 27, 2016
    #16
    i'm having the same slow issues with an accelsior... only in the the 300s read and write - i have another accelsior that gets 600s and seems to be fine, but the other is half speed for some reason... were there any conclusions on a fix for this?
     

Share This Page