Attaching Firewire800 drive speeds up SSD

Discussion in 'MacBook Pro' started by altecXP, Sep 4, 2010.

  1. altecXP macrumors 65816

    Joined:
    Aug 3, 2009
    #1
    I have this odd problem on my i5 MBP. When I have no external drives attached AJA System test reports my SSD speed at about 227mb/s read and write. When I plug in a USB external drive the SSD reports the same speed. When I plug in a FW800 drive the speeds jumo to 234 - 245mb/s. When I unplug the FW800 drive they go back down.


    Anyone else see this problem?
     
  2. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #2
    Hahaha. Interesting. I'm curious to see how this thread goes.
     
  3. Inconsequential macrumors 68000

    Joined:
    Sep 12, 2007
    #3
    I'll try it on my *Mac Pro* soon :)

    Sounds weird tho :p
     
  4. gianly1985 macrumors 6502a

    gianly1985

    Joined:
    May 30, 2008
    #4
    @OP

    Problem? What problem? :)

    1) it's just a benchmark

    2) the difference you're reporting is minimal

    3) and it's just linear speed. Random speed is what matters most for OS real life speed.

    So: this phenomenon should be just a "curiosity". Not a problem, no practical consequences. Don't expect your Mac to be actually slower if the FW800 drive is unplugged.
     
  5. altecXP thread starter macrumors 65816

    Joined:
    Aug 3, 2009
    #5
    I have seen another site report seeing this problem, and yes, if I am loosing almost 20mb/s by not having a FW drive plugged in I see that as a problem. It's an about 8% lose in its possible performance.

    Why should I not expect to get the full capabilities of my SSD all the time?


    I tested this on my desktop PC with an SSD (same brand) and it does not have this issue.
     
  6. gianly1985 macrumors 6502a

    gianly1985

    Joined:
    May 30, 2008
    #6
    Get to know how SSDs work, what are linear speeds and random speeds, what matters most for real world performances, etc and maybe then you will be able to tell what's a "problem" and what's not. A good start is to read everything Anandtech has ever published about SSDs.

    7-17mb/s of difference in LINEAR speeds is a non-issue, there's no way they can have any consequence in your real life usage.

    And you don't know about the reliability of that benchmark.

    You're talking about nothing, feel free to worry about it if you want. Just wanted to let you know that calling it a "problem" or blabbering about "the full capabilities of my SSD" is plain-dumb-wannabe-tech-savvy-but-not-one-actually speaking.
     
  7. altecXP thread starter macrumors 65816

    Joined:
    Aug 3, 2009
    #7
    I know the different, and I know what the numbers mean. Are you honestly telling me that even thought the difference is minimal I should accept ANY performance drop from a $2,600+ laptop and $700 drive?

    I have seen many people use AJA for their benchmark and recommend if over xBench.

    Hay look here is a guy getting a near 50% jump from plugin in a FW drive: http://forums.macrumors.com/showthread.php?t=940805

    ANd here are a few sites showing the problem:
    http://macperformanceguide.com/Reviews-MacBookProCore_i7-SATA-Bizarre.html
    http://www.markc.me.uk/MarkC/Blog/Entries/2010/6/17_2010_Macbook_Pro_i7_-_Weird_SSD_Performance.html
     
  8. gianly1985 macrumors 6502a

    gianly1985

    Joined:
    May 30, 2008
    #8
    I'll teach you a couple of new expressions: significant and non-significant. There's no "ANY" when you're talking about NUMBERS in SCIENCE/ENGINEERING.

    Your drop is non-significant, no matter how much you paid the machine. By your reasoning, you should worry about a 0.1 mb/s drop :rolleyes:. A drop from 240mb/s to 227mb/s in linear speeds is non-significant realworld-performance-wise.

    About the links, 2 of them are RAID setups and the performance drop is way higher than in your case.
     
  9. Eddyisgreat macrumors 601

    Joined:
    Oct 24, 2007
    #9
    I kind of have to agree. It's admirable to want every last bit of performance but benchies are too finicky.
     
  10. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #10

    +1

    Benchmarks are good for somethings, but in a case like this it's worthless. You can go ahead and complain about this until your blue in the face, but I doubt you'll find the cause. It could also be a glitch in the benchmark software, since anyone with reasonable computer knowledge will know that simply plugging in a Firewire device will not speed up another drive, on another type of bus.

    Here's an idea, go back to your mac and actually clock it with the drive plugged and unplugged, and if you can find a real slowdown in a real app, then it might be a reason to worry... otherwise...
     
  11. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #11
    Someone with more than 'reasonable computer knowledge' would not define that as law. There is a driver problem here... OSX is riddled with random bugs that appear under a small set of circumstances.
    They are, however, good for revealing issues like this.
     
  12. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #12
    So, you can argue that plugging in a FW800 should actually speed up something? How. It could be a driver problem, but one disk on one bus should not affect another disk on another. Yea, I'll define that as law.
     
  13. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #13
    It will not actually speed something else up... unless there is a problem that resolves itself when the firewire is plugged into. If you try to tell yourself that, as a rule, it's impossible for anything to speed up under any circumstance as a result of plugging into FW then you're ruling out a software issue by default.

    This issue is repeatable and demonstrated by multiple people across multiple programs.
     
  14. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #14
    No, I'm not. I know exactly what I'm saying.

    There you go. Maybe you were confusing yourself. All I said was "anyone with reasonable computer knowledge will know that simply plugging in a Firewire device will not speed up another drive, on another type of bus", yet you said this wouldn't be defined as law :confused:

    Edit: Looks like you were confused, as you changed the wording of your post :p
     
  15. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #15
    Ok, yes, reasonable computer knowledge will say that. Now, are you ruling out there being an issue because they apparently can't effect each other?
     
  16. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #16
    That isn't what I said though. It's obviously a software glitch.
     
  17. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #17
    Yeah, which I agree with. The benchmark program being the issue was ruled out... so we have some weird OSX bug happening. :eek:
     
  18. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #18
    Read my above post. But I also don't think a 20 MB increase in linear transfer speed at speeds above 200 MB/s will yield a noticeable, real world difference.
     
  19. RedReplicant macrumors 6502a

    Joined:
    Mar 31, 2010
    #19
    I wonder if it effects normal HDDs too, since they are on the same bus... I don't have a FW anything to test with. :(
     
  20. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #20
    I have a couple FireWire devices laying around. Which benchmark are people using? The issue here is that normal HDDs (especially laptop drives) don't usually have such a high transfer rate to begin with. I'd hypothesize and say it would only be a few MB difference, if any. Also, I think this problem is only affecting Macs on the new Arrandale platform (i5 & i7).
     
  21. Tortri macrumors 6502a

    Tortri

    Joined:
    Aug 30, 2010
    #21
    After reading this post I ran geek bench 64 bit, then promptly turned on my fw800 with a hitachi 500gb hd in it and the score was basically the same. To be sure I also rebooted my mbp with out the fw800 cable connected, same speeds, then connected fw800, no major difference, fw800 turned on, again little difference. There were times where w/o fw800 score inched out ahead, but I'm talking about like tops 10 point difference.

    I have a mbp 2010 15" 4gb ram with a realssd c300 128gb. Ssd is in HD spot in mbp.

    If there are any benchmarks programs you want me to try name them and I'll see if I get different results.
     
  22. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #22
    Geekbench has nothing to do with hard drive speed. It's only a measure of CPU and RAM performance... which isn't relevant to this thread.
     
  23. Tortri macrumors 6502a

    Tortri

    Joined:
    Aug 30, 2010
    #23
    D'oh, that would explain. Honestly I just seen someone mention it and downloaded it and gave it a try. I'll give xbench a try. I know that has a HD test. I'll post the results today.
     
  24. Tortri macrumors 6502a

    Tortri

    Joined:
    Aug 30, 2010
    #24
    #1 FW800 connected and turned on

    Disk Test 346.80
    Sequential 217.86
    Uncached Write 235.73 144.73 MB/sec [4K blocks]
    Uncached Write 230.69 130.52 MB/sec [256K blocks]
    Uncached Read 136.20 39.86 MB/sec [4K blocks]
    Uncached Read 409.59 205.86 MB/sec [256K blocks]
    Random 849.70
    Uncached Write 953.98 100.99 MB/sec [4K blocks]
    Uncached Write 424.90 136.03 MB/sec [256K blocks]
    Uncached Read 2968.58 21.04 MB/sec [4K blocks]
    Uncached Read 1032.03 191.50 MB/sec [256K blocks]



    #2 FW800 completely disconnected

    Disk Test 327.53
    Sequential 202.58
    Uncached Write 238.57 146.48 MB/sec [4K blocks]
    Uncached Write 230.40 130.36 MB/sec [256K blocks]
    Uncached Read 113.53 33.23 MB/sec [4K blocks]
    Uncached Read 415.67 208.91 MB/sec [256K blocks]
    Random 854.78
    Uncached Write 999.74 105.83 MB/sec [4K blocks]
    Uncached Write 424.20 135.80 MB/sec [256K blocks]
    Uncached Read 2772.32 19.65 MB/sec [4K blocks]
    Uncached Read 1040.35 193.04 MB/sec [256K blocks]



    #3 FW800 connected and turned on then turned off

    Disk Test 322.09
    Sequential 199.17
    Uncached Write 233.55 143.39 MB/sec [4K blocks]
    Uncached Write 223.78 126.61 MB/sec [256K blocks]
    Uncached Read 112.86 33.03 MB/sec [4K blocks]
    Uncached Read 404.47 203.28 MB/sec [256K blocks]
    Random 841.31
    Uncached Write 1014.78 107.43 MB/sec [4K blocks]
    Uncached Write 411.11 131.61 MB/sec [256K blocks]
    Uncached Read 2745.92 19.46 MB/sec [4K blocks]
    Uncached Read 1028.36 190.82 MB/sec [256K blocks]



    The only thing I'm seeing in a big jump is Sequential Uncached Read [4K blocks] and Random Uncached Read / Uncached Write [4K blocks]. But honestly after looking at the numbers and comparing #3 and #2 then looking at #1 the only thing that FW800 turned on has going for it so far is just Sequential Uncached Read [4k blocks]. The rest is just minut couple MB here and there ahead of turned off compared to turned on. So yes FW800 connected and turned on is faster, but i'd be surprised if you could really see the difference with it off. My 2 cents
     
  25. green86 macrumors 6502

    green86

    Joined:
    Sep 27, 2007
    Location:
    North Carolina
    #25
    Haha I love it. Random writes and reads went up with it disconnected, which is where most of the performance comes in. So... I guess attaching a FW800 drive doesn't actually speed up performance then?:rolleyes:
     

Share This Page