Results 1 to 7 of 7
  1. #1
    Joined
    Aug 2003
    Posts
    181

    Question How many Gigahertz is a teraflop?

    I was just wondering how many gigahertz is teraflop? I was trying to get an idea at how many combined gigahertz power supercomputers have and it seems like I will have to covert teralops to gigahertz to figure that out.

  2. #2
    Joined
    Oct 2001
    Location
    Jax,Fla,USA
    Age
    48
    Posts
    1,561
    Not enough information to complete the equation.Seriously,
    a TeraFLOP is 1 trillion Floating Point Opperations,it is a way
    to measure the computing power of supercomputers.More
    specifically,the number of teraFLOPs in a given amount of time,
    usually 1 second.
    Quote Originally Posted by JC Denton
    Quote Originally Posted by Bob Page
    I will burn like the brightest star...
    You're gonna burn all right.

  3. #3
    Joined
    Aug 2003
    Posts
    181
    I already know that stuff you told me, so it doesn't really help. If you need specific cpu to caculate it, how about many FLOP per sec can the top CPU around today perform?
    Like the Opteron 2 ghz, Athlon 64 FX, Athlon 64, Athlon 3000, Itanium, P4 EE, P4 3.2 ghz, etc....

  4. #4
    Joined
    Oct 2001
    Location
    Jax,Fla,USA
    Age
    48
    Posts
    1,561
    Then you should have asked that,I believe Itanium 2s produce
    the most FLOPS.
    Quote Originally Posted by JC Denton
    Quote Originally Posted by Bob Page
    I will burn like the brightest star...
    You're gonna burn all right.

  5. #5
    Joined
    Aug 2003
    Posts
    181
    hmm think intel or amd site will give specific #s? Guess I will have to go dig it out of some place......

  6. #6
    Joined
    Oct 2001
    Location
    Jax,Fla,USA
    Age
    48
    Posts
    1,561
    They usually only give you the FLOP per $ standard,the FLOP
    performance specifics are with complete systems,here ,that shuold give you an idea.
    Quote Originally Posted by JC Denton
    Quote Originally Posted by Bob Page
    I will burn like the brightest star...
    You're gonna burn all right.

  7. #7
    Joined
    Aug 2002
    Location
    Orangeburg, South Carolina (summer)
    Posts
    1,320
    FLOPs are an inherently non-deterministic statistic, and so tend to be misleading. To measure a computer's FLOPs, you code up a benchmark with xxx floating point operations and then see how long thed computer takes to run it. BUT--what about overhead? How was the code compiled? Are all operations equivalent on a processor?

    Also, different processor are capable of completing more or less instructions per cycle. Athlons, for instance, tend to complete more instructions per cycle than P4s.

    So, my point is a) FLOPs depend on the program and techniques you use to benchmark b) you can't make a meaningful relation between GHz and FLOPs cross-platform, or even on the platform, if you don't know how the FLOPs were measured
    "To follow knowledge like a sinking star / Beyond the utmost bound of human thought ... To strive, to seek, to find, and not to yield."
    --from "Ulysses", Tennyson

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •