RE: AMBER: PMEMD: dual XEON vs SGI cluster

From: Chris Moth <chris.moth.vanderbilt.edu>
Date: Mon, 15 Sep 2003 20:55:46 -0500

>===== Original Message From "Robert Duke" <rduke.email.unc.edu> =====
>Chris -
>Some quick comments about timings on sander6, sander7, and pmemd....

    Case closed - Thank you very much. It all makes sense.

    As you predicted, here is the logfile from an 8 SGI CPU PME/MD:

                   Parallel Profiling Results

                              A
                      N n
              F o g
              X n B S R O T
              d b n h u t o
              i o d a n h t
      P s n D k m e a
      E t d i e d r l
    ---------------------------------------------------------------------
      0 6102.1 73130.3 9809.0 466.8 606.7 327.2 90442.0
      1 6471.5 82723.9 239.2 373.4 716.0 16.0 90540.0
      2 6506.9 83048.6 1.4 409.0 688.6 17.3 90671.8
      3 6499.1 83075.0 1.3 410.1 680.2 16.8 90682.5
      4 6497.5 83073.2 1.3 400.2 686.4 16.7 90675.4
      5 6504.6 83074.5 1.4 401.4 681.4 14.8 90678.2
      6 6496.2 83061.4 1.3 371.0 703.3 15.4 90648.5
      7 6472.2 83014.8 1.4 350.3 732.4 14.6 90585.7
     av 6443.8 81775.2 1257.0 397.8 686.9
    std 129.7 3269.4 3233.3 32.8 34.8
    min 6102.1 73130.3 1.2 350.3 606.7
    max 6506.9 83075.0 9809.0 466.8 732.4

    and here is a job of the same size on my dual Xeon:
                              A
                      N n
              F o g
              X n B S R O T
              d b n h u t o
              i o d a n h t
      P s n D k m e a
      E t d i e d r l
    ----------- ----------------------------------------------------------
      0 1628.2 103843.7 2862.6 452.4 1169.8 119.5110076.3
      1 1734.1 107054.3 0.2 361.6 1213.7 60.1110424.0
     av 1681.1 105449.0 1431.4 407.0 1191.8
    std 53.0 1605.3 1431.2 45.4 22.0
    min 1628.2 103843.7 0.2 361.6 1169.8
    max 1734.1 107054.3 2862.6 452.4 1213.7

    Thanks very much

    Chris

Chris Moth
chris.moth.vanderbilt.edu


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Sep 16 2003 - 03:53:00 PDT
Custom Search