Fusion drive speed tests?

Discussion in 'MacBook Pro' started by toke lahti, Mar 5, 2014.

  1. toke lahti, Mar 5, 2014
    Last edited: Mar 5, 2014

    toke lahti macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #1
  2. benwiggy macrumors 68020

    Joined:
    Jun 15, 2012
    #2
    A Fusion drive read and write times are going to vary considerably depending on what you are doing. I get much faster speeds for a 1Gb and 4Gb test than I do for a 2Gb or 3Gb test, which is odd.

    However, a 60Gb SSD portion is quite small.
     
  3. Mike Boreham macrumors 65816

    Joined:
    Aug 10, 2006
    Location:
    UK
    #3
    I have home brew Fusion Drives on both my MBPs, and I just did a Black Magic test on the one with 960Gb SSD and 1Tb HDD.

    The results were 449 and 537 peak read/write.

    The size of the SSD should not affect the test. Fusion Drives work by keeping 4Gb free on the SSD at all times, and everything goes to the SSD first. So for example when you copy an 8Gb file to a Fusion Drive, the first 4Gb copies at SSD speed and then it drops to HDD speed because the SSD is full and the write switches to the HDD. When the whole 8Gb has copied, the SSD will unload 4Gb to the HDD in the background to recreate the 4Gb space on the SSD. You can see all this happening with iStat Menus which shows disk activity separately for the SSD and HDD parts of the Fusion Drive. It is one of the tests I use to check that a Fusion Drive is working.

    So the speed test should read and write to the SSD and show the SSD speeds.

    When I watch iStat menus disk activity during the Black Magic test I can see all the activity is on the SSD and none on the HDD, as expected.

    Your result looks slow unless it is a very slow SSD or the SSD has degraded and needs trimming.
     
  4. toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #4
    Just checked s.m.a.r.t. with SMARTreporter and it tells:
    ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE
    9 Power_On_Hours 0x0032 100 100 000 Old_age Always - 136846247985247
    !
    I bought this from Amazon.de 2 weeks ago...
     
  5. Mike Boreham macrumors 65816

    Joined:
    Aug 10, 2006
    Location:
    UK
    #5
    Smart reporting of SSD is sometimes erroneous but suspicious in view of surprising low result.

    Drive DX is a more comprehensive Smart tool and probably handles SSD better.

    Sorry not to be more help.
     
  6. p3ntyne, Mar 5, 2014
    Last edited: Mar 5, 2014

    p3ntyne macrumors 6502

    p3ntyne

    Joined:
    Jan 10, 2014
    Location:
    Sydney, Australia
    #6
    That is perfectly normal. The SSD you have uses a SandForce controller which works better with compressed data (Blackmagic disk speed test uses incompressible.)

    i have a 60Gb OWC SSD in my fusion drive and get the same speeds in blackmagic. However, if I calculate the speeds manually while transferring a file, I get ~360 Write and ~520 Read. The speed will be the same in or out of a fusion drive.

    P.S. I get 60 write in 151 read in blacmagic.
     
  7. toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #7
    Wow, didn't think researching how well ssd controllers handle compressed (video) data vs. uncompressed would be necessary...
    Are you saying that BMspeedtest is showing wrong numbers with uncompressed files and calculating manually gives right numbers?
     
  8. Mike Boreham macrumors 65816

    Joined:
    Aug 10, 2006
    Location:
    UK
    #8
    The SSD in my Fusion Drive (results quoted earlier) is a Crucial M500.
     
  9. p3ntyne macrumors 6502

    p3ntyne

    Joined:
    Jan 10, 2014
    Location:
    Sydney, Australia
    #9
    Yes. To test the write speed manually, open terminal and type 'iostat -c 2000'. This will give you a list of the disk activity. Next, create a .dmg of less than 4 Gb and watch the actual read speed of you drive.

    In the screenshot you can see actual write speeds of which I get ~280-360. It is probably on the low side since the ssd portion is pretty full.

    EDIT: Just did the test again after closing all other apps and got 354.
     

    Attached Files:

  10. toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #10
    Thanks for the info.
    Just got similar answer from Sandisk's support, but your info is, let's say, more credible.

    Also thanks, Mike!
    Checking that Drive DX right now...
     
  11. p3ntyne macrumors 6502

    p3ntyne

    Joined:
    Jan 10, 2014
    Location:
    Sydney, Australia
    #11
    I'm glad I could help!
     
  12. toke lahti, Mar 9, 2014
    Last edited: Mar 10, 2014

    toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #12
    "iostat -c 2000" with AJA's test running shows ~96 MB/s write and ~124 MB/s read.
    Same numbers in terminal as in AJA's test.
    Making an image shows ~99 MB/s in iostat.

    I'm wondering if Fusion is working backwards (data goes first to 5K1000 hdd...)?

    EDIT: Weird, I run AJA a bit more times and now speed is showing 106/136.
    Better than last time, but still suspiciously slow...?
    Anyway iostat shows traffic in disk0, which is ssd and 5K1000 shouldn't be able over 100MB/s speeds, although it's still almost empty...
     
  13. toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #13
    Noticed weird thing in System Information:
    NVidia MCP79 AHCI:

    Vendor: NVidia
    Product: MCP79 AHCI
    Link Speed: 3 Gigabit
    Negotiated Link Speed: 1.5 Gigabit

    Anybody else have this issue with Kingston's V300?
     
  14. toke lahti thread starter macrumors 68000

    Joined:
    Apr 23, 2007
    Location:
    Helsinki, Finland
    #14
    Now Kingston tells, that V300 has a problem with Santa Rosa chipset and I should send the ssd to somewhere for tinkering to get promised sataII 3Gb/s...
     
  15. p3ntyne macrumors 6502

    p3ntyne

    Joined:
    Jan 10, 2014
    Location:
    Sydney, Australia
    #15
    Well, that's not normal. I would return the SSD. A fusion drive will automatically determine which drive is faster and use that as the primary drive no matter which order you put the drives in or which bay they are in.

    The only change that data is going to the 5k1000 first is if the SSD is faulty and really. You should be able to test this by having iostat open in the corner while performing some basic tasks (e.g. opening Safari, commonly used documents) and see which drive has activity.
     

Share This Page