Lies, Damn Lies and File System Benchmarks

Discussion in 'other software & services' started by Pedro, Aug 12, 2009.

Thread Status:
Not open for further replies.
  1. Pedro

    Pedro Registered Member

    Joined:
    Nov 2, 2006
    Posts:
    3,502
    http://www.linux-mag.com/cache/7464/1.html
     
  2. dw426

    dw426 Registered Member

    Joined:
    Jan 3, 2007
    Posts:
    5,543
    I'm glad this came out, I get annoyed seeing the almost monthly "Which Browser is Faster?" articles plastered all over computing and technology websites. I know that's only one part of the issue, but it's a major one. I can understand a gamers obsession with benchmarking video cards, but, over the years, even I as a gamer have known these benchmark tests do more to separate teenagers and adults alike from their hard-earned 5-600 dollars than they do to provide explanations of the benefits of one card over another.

    I've said it before and I'll say it again, humans aren't even able to discern the speeds that these benchmark tests put out. Hardware, software, these tests simply don't matter now. I'm not sure they ever really did.
     
  3. Sully

    Sully Registered Member

    Joined:
    Dec 23, 2005
    Posts:
    3,719
    I agree. Testing hardware or software, there are so many variables that come into play it is hard to tell exactly how your results will be until you plug it in or install it. I like hardware benchmarks to get an idea of just how large a margin of difference there might be. But when hardware benchmarks are pretty close, or when looking at software performance, I don't feel you can 'normally' take them as accurate. That is, I have tried a lot of hardware/software that is supposed to be 'faster, better, etc' but that on my machine, with my hardware and my configuration, either it is not as claimed or the difference is imperceptible.

    I am a fan of looking at these tests to get a feel of what is out there that I have never heard of or used, then trying them out. Often I find new useful tools this way, but normally I don't find a clear winner in performance over what I already have been using. There are exceptions to this as always though.

    Sul.

    EDIT: Here is an experiment I did a couple years ago, it is a cluster/stripe size experiment for raid arrays (raid 0).
    http://mrwoojoo.com/
     
    Last edited: Aug 12, 2009
Loading...
Thread Status:
Not open for further replies.