Results 1 to 7 of 7

Thread: AMD faster than Intel, but only under Linux!!!

  1. #1
    Join Date
    Aug 2007
    Location
    Europe
    Posts
    401

    Default AMD faster than Intel, but only under Linux!!!

    AMD taster than Intel, but only under Linux!!!

    http://it.anandtech.com/IT/showdoc.aspx?i=3784&p=7

    Hmmm, one more reason to go Linux.

  2. #2
    Join Date
    Aug 2007
    Location
    Europe
    Posts
    401

    Default

    While the article mentions Amdahl's Law, it is outdated, as described at http://www.cis.temple.edu/~shi/docs/amdahl/amdahl.html

    and

    http://www.cis.temple.edu/~shi/docs/amdahl.ps


    .

  3. #3
    Join Date
    Jun 2009
    Posts
    1,145

    Default

    Quote Originally Posted by sabriah View Post
    AMD taster than Intel, but only under Linux!!!

    http://it.anandtech.com/IT/showdoc.aspx?i=3784&p=7

    Hmmm, one more reason to go Linux.
    well i realized that long time ago, but it make sense after all. Windows have been always very sicky in the way it handle threads(beside the horrible gut killing pain of sync them in the win32 thread api :* ). on the other hand posix threads have always seem extremely logical and stable in linux and other unix like oses too.

    my bet here is that the poor performance in windows is due to hypertransport and the way cores are interconnected internally but is not AMD faults at the end is intel's/microsoft faults.

    my bet is that somehow microsoft have some internal optimization in their thread api that force it to assume or add latency to compensate the heat produced for the old fsb system from intel, so that could ovbiuosly kill the benefit of Athlon/phenom/opteron hypertransport and inner memory manager due to the self generated latency.

    now on the other hand could simple be that microsoft and intel have some hidden deal in there to slowdown amd chips internally so they make sure amd wont kill they processor in benchys by much, pretty much like the famous microsft/foxxcon bios "misunderstanding" from the g33 chipsets

  4. #4
    Join Date
    Sep 2006
    Posts
    210

    Default

    Quote Originally Posted by jrch2k8 View Post
    ...
    Or maybe MS' compiler just produces worse 64bit code than GCC
    (although I don't know which toolchain built the Blender Windows
    binaries in the AT test linked to above).

    Here's an interesting evaluation of compilers. It's 3 years old
    but still highly entertaining (at least to me):
    http://www.fefe.de/know-your-compiler.pdf

  5. #5
    Join Date
    Sep 2006
    Posts
    210

    Default

    [stupid time-limit to edit posts]

    The OpenLDAP guys also noticed the Opteron beating the
    pants off a faster Xeon in 64bit mode:

    http://www.connexitor.com/blog/pivot/entry.php?id=191

  6. #6
    Join Date
    Aug 2007
    Posts
    437

    Default

    AMD invented the x86-64 mode.

    Conroe (core 2)'s 64-bit implementation is still a little bit hackish.

    Try adding Nehalem Xeons to the comparison, it truces AMD in no matter how many bits computing mode.

  7. #7
    Join Date
    Apr 2010
    Posts
    48

    Default

    Quote Originally Posted by FunkyRider View Post
    Try adding Nehalem Xeons to the comparison, it truces AMD in no matter how many bits computing mode.
    X5570 and X5670 are both Nehalem based.

Posting Permissions

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