1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Poor benchmark results for a RAID 0 setup??

Discussion in 'Storage' started by KBD, Sep 30, 2008.

  1. KBD New Member

    Joined:
    Feb 23, 2007
    Messages:
    2,477 (0.91/day)
    Thanks Received:
    279
    Location:
    The Rotten Big Apple
    I decided to run some HDD benchmarks on 2x 320GB Barracudas 7200.11 in RAID 0 on a new systems i built for someone. First off, the strange thing is that HD Tach and HD Tune give totally diffrent burst rates for the RAID 0 setup. Furthermore, the Random Access Times and Burst Rates are just horrible, worse than my single non-RAID 150GB Raptor and Barracuda 320GB 7200.10. Here are the results for 2x RAID 0 drives:

    HD Tach

    Burst Rate: 436 MB/s
    Randon Access Time: 17.4ms
    Average Read: 187.8 MB/s

    HD Tune

    Burst Rate: 89.9 MB/s
    Randon Access Time: 16.6ms
    Average Read: 178.3 MB/s

    As you can see there is some discrepancy between the 2 benchmarks in case of Random Access Times & Avg Read, but the discrepancy between Burst Rates is humongous! 436 MB/s for HD Tach and 89 for HD Tune. I think that HD Tach is way off, 89 looks like it could be closer to the truth, but i it still very low comparing to my single drives, here are their results:


    HD Tach

    WD Raptor 150

    Burst Rate: 133.7 MB/s
    Randon Access Time: 8.9ms
    Average Read: 75.3 MB/s

    Barracuda 320 7200.10

    Burst Rate: 134 MB/s
    Randon Access Time: 13.7ms
    Average Read: 59.7 MB/s

    HD Tune

    WD Raptor 150

    Burst Rate: 116 MB/s
    Randon Access Time: 8.2ms
    Average Read: 67.8 MB/s

    Barracuda 320 7200.10

    Burst Rate: 111 MB/s
    Randon Access Time: 13.5ms
    Average Read: 59 MB/s


    So judging by these results my single drives beat the 2 drives in RAID 0 in every test except for Avg Read, how could this be? Are these normal results?
  2. DanTheBanjoman SeƱor Moderator

    Joined:
    May 20, 2004
    Messages:
    10,553 (2.82/day)
    Thanks Received:
    1,383
    Burst tends to be screwed up, I have a far lower burst than average read. Really makes no sense.
    The Raptor is a 10K disk, its sole purpose is to have lower access times.
    KBD says thanks.
  3. KBD New Member

    Joined:
    Feb 23, 2007
    Messages:
    2,477 (0.91/day)
    Thanks Received:
    279
    Location:
    The Rotten Big Apple
    So you are saying these are normal numbers? Is there a way to measure burst rate accurately? And what about access times on the RAID 0 array, are they supposed to increase? As one can gather from the results, the access times on the single 7200.10 Barracuda are lower than on the 2 RAID 0 drives.
  4. Sasqui

    Sasqui

    Joined:
    Dec 6, 2005
    Messages:
    7,542 (2.37/day)
    Thanks Received:
    1,332
    Location:
    Manchester, NH
    Perhaps a difference in block size between the two programs?
    KBD says thanks.
  5. KBD New Member

    Joined:
    Feb 23, 2007
    Messages:
    2,477 (0.91/day)
    Thanks Received:
    279
    Location:
    The Rotten Big Apple
    Good point Sasqui. i was actually gonna ask about that. In HD Tune the block size defaults to 64kb. I dont think i can change it in HD Tach. But does anyone one know which block size should the drives be tested at? Is it fine to leave it at default 64KB for HD Tune?
  6. twicksisted

    twicksisted

    Joined:
    Oct 4, 2007
    Messages:
    2,436 (0.97/day)
    Thanks Received:
    350
    it all depends on what you're going to use the computer for... large file storage, small file storage, database etc.. small stripes with large data is pointless and large stripes with small data etc..

    Im using 128k stripe on my raid 0 setup and its pretty decent.
  7. KBD New Member

    Joined:
    Feb 23, 2007
    Messages:
    2,477 (0.91/day)
    Thanks Received:
    279
    Location:
    The Rotten Big Apple
    yes, i'm using 128k stripe for this RAID 0 array also. I used that only because the Intel RAID setup utility recomended that for this array. I also read an article on the subject beforehand and it seems that there is a lot of debate in regards to stripe sizes for arrays. Anyway, i think i should've tested the RAID array with 128 block size with HD Tune as i can change it in that program, i think i'm gonna retest and see what happens.
  8. Sasqui

    Sasqui

    Joined:
    Dec 6, 2005
    Messages:
    7,542 (2.37/day)
    Thanks Received:
    1,332
    Location:
    Manchester, NH
    Yea, makes sense. If HD Tune defaults to a larger block size, the results would definitely be lower. If it's larger than your allocation unit size, then the numbers would likely go *way* down - presumably would have to read at least two units ber block!

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page