Caviar Black RAID0 WTF

Discussion in 'Mac Pro' started by bbadalucco, Oct 2, 2009.

  1. bbadalucco macrumors 6502

    Jan 4, 2009
    I bought three WD Caviar Black 7200 HDDs. I've put all three in a stripped raid on my 2009 MP. The block size is at 32, only because thats what it defaulted to.

    Anyway, everything seems to be opening a bit slower than expected so I ran Xbench...below are the results.

    Results 193.15
    System Info
    Xbench Version 1.3
    System Version 10.6.1 (10B504)
    Physical RAM 3072 MB
    Model MacPro4,1
    Drive Type Mac OSX
    Disk Test 193.15
    Sequential 196.84
    Uncached Write 401.05 246.24 MB/sec [4K blocks]
    Uncached Write 373.99 211.60 MB/sec [256K blocks]
    Uncached Read 76.15 22.29 MB/sec [4K blocks]
    Uncached Read 494.63 248.60 MB/sec [256K blocks]
    Random 189.60
    Uncached Write 78.39 8.30 MB/sec [4K blocks]
    Uncached Write 713.67 228.47 MB/sec [256K blocks]
    Uncached Read 276.36 1.96 MB/sec [4K blocks]
    Uncached Read 301.07 55.87 MB/sec [256K blocks]

    I was expecting much more than this, did I do something wrong? I've used drive bays 1-3 for the raid.
  2. AZREOSpecialist macrumors 68020


    Mar 15, 2009
    Caviar Blacks are not designed for RAID. You need the RE2/RE3 version of this drive for a RAID array. The controller has the proper functions enabled to support RAID. Google this up, you'll find many folks with problems using the non-RAID version of this drive.
  3. UltraNEO* macrumors 601


    Jun 16, 2007
    Curious... Why did you choose to use RAID0 with three drives?
  4. UltraNEO* macrumors 601


    Jun 16, 2007
    Essentially those WDC Black are a little poor when it comes to timings, for simple soft-raid it's OK.
    But yeah, the RE3/RE4 are a much better choice! And they'll last longer!
    But maybe the OP doesn't want to spend that much? :rolleyes:
  5. bbadalucco thread starter macrumors 6502

    Jan 4, 2009
    Not it was a rushed purchase and I thought they'd work well because I've heard they were fast (as a single drive). I was looking for speed and size and thought this would be a good fit.

    Thats what happens when you don't do your homework.

  6. Tesselator macrumors 601


    Jan 9, 2008
    Xbench is crap! Use a different utility for a more robust result spread. Geekbench is crap all the more so. It's unfortunate but very true that the two most popular mac benchmarkers are indeed the two worst - in some areas not offering any semblance of accuracy at all.

    BTW, Black "consumer" drives are just fine for "consumer" RAID0! The raid class drives are only slightly more desirable and not at all faster. They're better only in that there may be less chance for error - and I stress maybe.

  7. nanofrog macrumors G4

    May 6, 2008
    Blacks would be fine for software RAID. Just change the stripe size to something larger (larger stripe size = faster throughputs, less capacity available). You may have to experiment around with what you do (not just benchmark utils) to see what best fits your usage.

    Just make sure you have a proper backup, no matter what you have for a drive setup. ;)
  8. Zerozal macrumors 6502

    Apr 3, 2009
    Nonsense. Blacks work just fine in RAID. The RE versions of the WD drives add additional vibration dampening for use in racks. For a home software RAID, the Blacks work just fine, and are no slower.
  9. nanofrog macrumors G4

    May 6, 2008
    The Blacks would be fine for software based arrays, but not necessarily on a true hardware RAID controller, as it can have problems. The reason is simple. It has to do with the recovery timings in the firmware, and this is especially the case if the controller is a SAS model (exremely picky about SATA drives, and there's extremely few, if any, consumer models that will actually work as is).

    WD has a utility out that can allow you to adjust the TLER values (recovery timing settings), but they're the only vendor I've ever found one for, as they guard them carefully.

    You can also get into other details, such as the UBE. Consumer models typically have a rating of 1E14, and enterprise uses 1E15. That order of magnitude can make a difference.

    In the end, it always depends on the specifics. But in the OP's case, the Blacks would suffice, as they're attached to the logic board's SATA controller. :)
  10. AZREOSpecialist macrumors 68020


    Mar 15, 2009
    It is not "nonsense" that a basic Caviar Black will not work well in a hardware RAID setup. I did extensive research on this subject when I was purchasing several hard drives for use with my RocketRAID 4320 controller. A few basic google searches, even reading customer reviews on Newegg, revealed the issues involved. Based on what others have already said, the Caviar Black may not have issues in a software RAID configuration, but the original post to which I replied did not indicate whether it was a software or hardware RAID setup.

    There are additional differences with the RE versions of the drive relating specifically to the controller, not just vibration dampening.
  11. nanofrog macrumors G4

    May 6, 2008
    When getting into RAID, every single detail matters.

    There are drastic differences with ease of locating usable drives between software, Fake RAID, and a true hardware RAID controller. Not only the drive maker & model, but even the drive's firmware revision can make a difference. Other issues as well, and it can be even more complicated on Macs, particularly if you want to boot from the array, or run multiple OS's that have access to it.

    So specifiying the details is sort of needed, as both of you were correct from a certain POV, but the statements were too general to be applicable to any siutuation. ;)
  12. hyram macrumors regular

    Jun 15, 2009
    While the "nonsense" comment is a little much, so is your contention that they should not be used in a hardware RAID. Yes, there are lots of folks out there that have had issues... mostly because they don't want to take the time to learn how to set up a RAID and how to configure these particular drives. I'd give a newegg “review” about a weighting of a 2 on a scale of 1 to 10.

    And yes the RE2's and 3's will work much better as that is what they are designed for. But in the area of the globe where I live the RE’s are special order and come at a $75 premium over the US RE3 price. I can get the Blacks at about the same US price.

    I've been using 8 of the 1TB Blacks connected to a rocketraid 4322 controller (RAID5) for about 3 months now with zero problems. And I do mean zero. I use it mainly for editing video and can sustain 3 or 4 HD streams at the same time with no bogging. The biggest secret to these drives is to turn off the TLER when used in hw RAID, I’m not so sure about a sw RAID but I if I read the WD whitepaper correctly you should leave it on. And I'd suggest stress testing any RAID before taking it live... Disktester 2.0 works well for this, but there are others.

  13. nanofrog macrumors G4

    May 6, 2008
    RAID with hardware controllers is much more tricky than software based arrays on the system's SATA controller chip/s.

    The WD's are unique in the sense you can adjust the TLER values, and it's a really handy thing to be able to do. The consumer drives are set at 0,0, while the enterprise counterparts are 7,0 (read, write, in seconds). This is particularly useful on the Green drives they made before they came out with the RE4 versions of them (enterprise grade Greens are in these). It's also less expensive, but you do loose a few things (additional feedback circuits on the drive's controller). This has to be weighed in during the planning process.

    Testing out an array prior to placing any data on it is a necessity, not really an option IMO, as it can easily bite you in the @ss if you don't. :eek: :p
  14. seisend macrumors 6502a


    Feb 20, 2009
    Switzerland, ZG
    I have a MP 2009 with 3 x 1TB Caviar Black's in RAID0 , too.

    @ Tesselator, which benchmark programm do you recommend?
  15. bbadalucco thread starter macrumors 6502

    Jan 4, 2009
    Would you just try xbench...I'm just curious as to what speeds you get. It may not be the best program, but I'd be nice to compare our results with the same software.

  16. johnsosn macrumors member

    Apr 24, 2009
    I have 3 Re3's in Raid 0 and get better results in Xbench.
  17. racer748 macrumors newbie

    Apr 29, 2009
    I've just set-up my new Mac Pro and used 3x'1.0TB Hitachi Deskstar 7K1000.B 7200RPM SATA I/II HD with 16MB Cache' in Raid 0, the stock 640GB WD as a standalone and a Crucial 128GB as Boot. I really wanted the enterprise 2GB Hitachi disks, but they didn't have any in stock when I ordered and compared to my old G5 2.7, I figured I was upgrading anyway. :)

    X-Bench results:

    Results 210.33
    CPU Test 154.12
    Thread Test 721.38
    Memory Test 344.06
    Quartz Graphics Test 202.45
    OpenGL Graphics Test 116.27
    User Interface Test 292.88
    Disk Test 180.34
    Sequential 272.11
    Uncached Write 560.97 344.43 MB/sec [4K blocks]
    Uncached Write 568.91 321.89 MB/sec [256K blocks]
    Uncached Read 107.10 31.34 MB/sec [4K blocks]
    Uncached Read 548.69 275.77 MB/sec [256K blocks]
    Random 134.86
    Uncached Write 73.27 7.76 MB/sec [4K blocks]
    Uncached Write 411.26 131.66 MB/sec [256K blocks]
    Uncached Read 128.39 0.91 MB/sec [4K blocks]
    Uncached Read 172.67 32.04 MB/sec [256K blocks]

    Results 274.73 <-- don't know how it gets this number!
    CPU Test 153.80
    Thread Test 684.39
    Memory Test 338.22
    Quartz Graphics Test 202.68
    OpenGL Graphics Test 116.13
    User Interface Test 305.98
    Disk Test 84.53
    Sequential 164.94
    Uncached Write 170.74 104.83 MB/sec [4K blocks]
    Uncached Write 168.42 95.29 MB/sec [256K blocks]
    Uncached Read 129.99 38.04 MB/sec [4K blocks]
    Uncached Read 209.93 105.51 MB/sec [256K blocks]
    Random 56.82
    Uncached Write 19.77 2.09 MB/sec [4K blocks]
    Uncached Write 232.51 74.43 MB/sec [256K blocks]
    Uncached Read 102.06 0.72 MB/sec [4K blocks]
    Uncached Read 174.97 32.47 MB/sec [256K blocks]

    Results 228.95
    CPU Test 154.27
    Thread Test 745.38
    Memory Test 337.80
    Quartz Graphics Test 203.66
    OpenGL Graphics Test 117.29
    User Interface Test 314.98
    Disk Test 314.57
    Sequential 292.53
    Uncached Write 416.34 255.62 MB/sec [4K blocks]
    Uncached Write 355.93 201.38 MB/sec [256K blocks]
    Uncached Read 159.36 46.64 MB/sec [4K blocks]
    Uncached Read 457.22 229.80 MB/sec [256K blocks]
    Random 340.20
    Uncached Write 114.93 12.17 MB/sec [4K blocks]
    Uncached Write 574.01 183.76 MB/sec [256K blocks]
    Uncached Read 3636.57 25.77 MB/sec [4K blocks]
    Uncached Read 961.55 178.42 MB/sec [256K blocks]
  18. wpc33 macrumors 6502


    Jul 2, 2006
    Vancouver, BC
    Why do you ask? Is that not such a hot idea? I have my brother's G4 running a RAID0 on 3 disks.

    I'm still waiting on a suggestion, Tess...
  19. seisend macrumors 6502a


    Feb 20, 2009
    Switzerland, ZG
    3 X 1TB Caviar Black's in RAID 0:

    Results 214.72
    System Info
    Xbench Version 1.3
    System Version 10.6.1 (10B504)
    Physical RAM 12288 MB
    Model MacPro4,1
    Drive Type Macintosh HD
    Disk Test 214.72
    Sequential 210.41
    Uncached Write 347.02 213.07 MB/sec [4K blocks]
    Uncached Write 448.72 253.89 MB/sec [256K blocks]
    Uncached Read 82.55 24.16 MB/sec [4K blocks]
    Uncached Read 559.92 281.41 MB/sec [256K blocks]
    Random 219.21
    Uncached Write 101.31 10.72 MB/sec [4K blocks]
    Uncached Write 787.11 251.98 MB/sec [256K blocks]
    Uncached Read 271.09 1.92 MB/sec [4K blocks]
    Uncached Read 292.63 54.30 MB/sec [256K blocks]
  20. nanofrog macrumors G4

    May 6, 2008
    My guess, it's to do with the fact it's a stripe set (RAID 0), and the issues it has when it fails. And that's not getting into the specific needs, such as system uptime (high availability for servers that run 24/7 from the day it goes live).

    Keep in mind, it's not if, but when, no matter the array type used. ;)

    For "home use"/non high availibility systems, a RAID0 can work fine, so long as the backups are kept up with properly to reduce data loss to an absolute minimum (important for any drives, single or RAID, no matter the array type used). It also assumes you can deal with the down time involved for rebuilding the array when a failure occurs ("the compromise" involved with the low cost to implement it).
  21. Abidubi macrumors 6502

    Feb 13, 2009
    This is my 2X WD 640 blacks raid 0. Seems to be just fine to me.

    And FYI the 640GB drives are slightly better than the 750 and 1TB if you look at the data sheet page 2.

    Results 153.72
    System Info
    Xbench Version 1.3
    System Version 10.6.1 (10B504)
    Physical RAM 6144 MB
    Model MacPro4,1
    Drive Type System RAID
    Disk Test 153.72
    Sequential 205.28
    Uncached Write 366.81 225.22 MB/sec [4K blocks]
    Uncached Write 334.32 189.16 MB/sec [256K blocks]
    Uncached Read 87.69 25.66 MB/sec [4K blocks]
    Uncached Read 422.88 212.54 MB/sec [256K blocks]
    Random 122.86
    Uncached Write 46.89 4.96 MB/sec [4K blocks]
    Uncached Write 784.20 251.05 MB/sec [256K blocks]
    Uncached Read 177.12 1.26 MB/sec [4K blocks]
    Uncached Read 232.03 43.06 MB/sec [256K blocks]
  22. UltraNEO* macrumors 601


    Jun 16, 2007
    Correct me if I'm wrong but isn't RAID0 a stripe disk? If that's so, then using three drives would mean there's triple the chances of something failing and triple the chances of losing your data - lets hope you don't lose anything valuable, and the drives data is restricted to domestic (non-commercial) uses. If I recall, RAID level 0 disc partitions are almost impossible to recover mainly because the drives don't keep any parity data.

    OK, speaking from experience. RAID0 are wonderful provided nothing goes wrong on any of the drives involved, all it takes is for one drive to develop a fault and that could cause the entire array to fail, making the system unstable; leading to data lost!! This could be something simple such as a back sector or track, error cause by power outtage or system failure/crash/panic.

    When I first got my MP i built a simple array using two WDC blacks too! Now I use my machine for commercial work and securing my data has been a major task. The bigger my project are, the more I have to waste time creating backups. However back then, I was inexperienced and took the tasks of creating backups lightly... unfortunately within a year something terrible happened.. one of the drives start stuttering, system bootup times were still fast, but as soon as I opened any applications it all came crashing down!!

    I've since upgrade to hardware based RAID5 (local) with four disks and hardware based RAID6 (remote). Sure, it's not as fast as RAID0 and despite what other rattle on about, I don't seem to have issues with editing video online plus I have added redundancy and my data is safer - but I still create backups! Can't seem to escape creating backups...
  23. bbadalucco thread starter macrumors 6502

    Jan 4, 2009
    Isn't that what the time machine application is for??? Thats my plan...
  24. UltraNEO* macrumors 601


    Jun 16, 2007
    TM isn't the best option out there! It's not fool proof! And it's still backing up on mechanical devices which can still fail. So don't rely on it completely. There are many users here, who've had a bad experience with TM.. go search the forums.
  25. bbadalucco thread starter macrumors 6502

    Jan 4, 2009
    Will do, thanks for the heads up

Share This Page