Personal tools
You are here: Home Codes ZEUS 3D SGI Power Challenge (32 R10000 CPUs)
Document Actions

SGI Power Challenge (32 R10000 CPUs)

by streeter last modified 2007-03-30 04:42
  • These tests were run under a light system load.

  • All ZEUS-MP routines were compiled with: f77 -c -O3 -g3 -64 -r10000

  • Most ZEUS-3D routines were compiled with: f77 -c -O3 -w -g3 -64 -mips4 -r10000 -pfa list -WK,-ro=3,-so=3,-o=5,-as=l -OPT:roundoff=3,IEEE_arithmetic=3

  • Note that ZEUS-3D cannot be run on more than one POWERnode because the system's shared memory does not extend across POWERnodes.

  • WORK IS SCALED WITH THE NUMBER OF PROCESSORS



 
GRID: 32 x 32 x 32 per processor(tile) 
ZEUS-MP (10 steps) 

                                                                                  Speedup/
 Processors  Layout  Wall Clock  tused(s) Zone-Cycles/sec  MFLOPS  C90s   Speedup Processor 
      1       1x1x1     3.36       3.32        97936       89.67   0.82    1.00     1.00 
      2       2x1x1     3.59       3.55       183423      167.94   1.53    1.87     0.94 
      2       1x2x1     3.45       3.41       191093      174.96   1.59    1.95     0.98 
      2       1x1x2     3.41       3.37       193044      176.74   1.61    1.97     0.99 
      4       2x2x1     3.71       3.67       354998      325.02   2.95    3.62     0.91 
      4       2x1x2     3.69       3.64       357374      327.20   2.97    3.65     0.91 
      4       1x2x2     3.56       3.52       370471      339.19   3.08    3.78     0.95 
      8       2x2x2     4.55       4.49       580416      531.41   4.83    5.93     0.74 
     16       4x2x2     8.14       8.02       649581      594.74   5.41    6.63     0.41 
     16       2x4x2     8.11       7.94       652211      597.14   5.43    6.66     0.42 
     16       2x2x4     7.93       7.74       666355      610.09   5.55    6.80     0.43 
     32       4x2x4    16.81      16.27       632822      579.39   5.27    6.46     0.20 
 
ZEUS-3D (20 steps) (same layout)
                                                                          Speedup/
 Processors  Wall Clock  tused(s) Zone-Cycles/sec MFLOPS   C90s   Speedup Processor 
      1         4.00       4.58        71508       62.50   0.33     1.00    1.00 
      2         5.00       4.88       134384      114.63   0.42     1.88    0.94 
      4         7.00       6.16       212621      177.75   0.62     2.97    0.74 
      8        11.00      10.58       247758      202.75   0.69     3.46    0.43 
     16        23.00      20.74       252845      204.74   0.57     3.54    0.22 
     32        52.00      46.46       225701      182.76   0.51     3.16     0.10 



 
GRID: 64 x 64 x 64 per processor(tile) 
ZEUS-MP (10 steps) 

                                                                                   Speedup/
 Processors  Layout  Wall Clock  tused(s) Zone-Cycles/sec  MFLOPS   C90s   Speedup Processor 
      1       1x1x1     29.80      29.47        88395       70.24   0.46     1.00    1.00 
      2       2x1x1     30.76      30.44       171170      136.02   0.89     1.94    0.97 
      2       1x2x1     30.35      30.03       173502      137.87   0.91     1.96    0.98 
      2       1x1x2     30.16      29.85       174559      138.71   0.91     1.97    0.99 
      4       2x2x1     32.12      31.78       327817      260.50   1.71     3.71    0.93 
      4       2x1x2     32.08      31.74       328324      260.90   1.72     3.71    0.93 
      4       1x2x2     31.52      31.19       334149      265.53   1.75     3.78    0.95 
      8       2x2x2     40.94      40.29       514782      409.07   2.69     5.82    0.73 
     16       4x2x2     72.80      71.92       579394      460.42   3.03     6.55    0.41 
     16       2x4x2     71.50      70.14       588763      467.86   3.08     6.66    0.42 
     16       2x2x4     72.27      71.37       583558      463.72   3.05     6.60    0.41 
     32       4x4x2    144.80     142.46       584651      464.59   3.06     6.61    0.21 
     32       2x4x4    144.09     141.67       587548      466.90   3.07     6.65    0.21 
 
ZEUS-3D (10 steps) (same layout)

                                                                           Speedup/
 Processors  Wall Clock  tused(s) Zone-Cycles/sec  MFLOPS   C90s   Speedup Processor 
      1         48.00      47.70        54957       44.97   0.15     1.00    1.00 
      2         49.00      48.52       108064       87.50   0.24     1.97    0.98 
      4         57.00      54.91       190972      154.63   0.43     3.47    0.87 
      8        100.00      96.56       217198      175.87   0.49     3.95    0.49 
     16        213.00     202.22       207408      167.94   0.47     3.77    0.24 



 
GRID: 128 x 64 x 64 per processor(tile) 
ZEUS-MP (10 steps) 
 Processors  Layout  Wall Clock  tused(s) Zone-Cycles/sec MFLOPS   C90s   Speedup Processor 
      1       1x1x1    52.46      51.92       100340       77.96   0.48     1.00    1.00 
      2       2x1x1    53.70      53.14       196088      152.35   0.93     1.95    0.98 
      2       1x2x1    52.93      52.38       198910      154.54   0.95     1.98    0.99 
      2       1x1x2    52.75      52.20       199602      155.08   0.95     1.99    0.99 
      4       2x2x1    56.09      55.48       375548      291.78   1.79     3.74    0.94 
      4       2x1x2    55.77      55.18       377632      293.40   1.80     3.76    0.94 
      4       1x2x2    55.05      54.44       382604      297.26   1.82     3.81    0.95 
      8       2x2x2    71.28      70.46       590805      459.02   2.82     5.89    0.74 
     16       4x2x2   122.72     121.23       686962      533.73   3.27     6.85    0.43 
     16       2x4x2   123.22     120.91       683098      530.73   3.26     6.81    0.43 
     16       2x2x4   121.73     119.81       691848      537.53   3.30     6.90    0.43 
     32       4x4x2   244.15     238.92       696029      540.77   3.32     6.94    0.22 
     32       4x2x4   243.19     238.27       698565      542.74   3.33     6.96    0.22 
     32       2x4x4   241.94     237.44       701449      544.98   3.34     6.99    0.22 
 
ZEUS-3D (10 steps) (same layout)

                                                                           Speedup/
 Processors  Wall Clock  tused(s) Zone-Cycles/sec  MFLOPS   C90s   Speedup Processor 
      1         92.00      91.47        57318       46.41   0.13     1.00    1.00 
      2         96.00      95.43       109884       88.98   0.25     1.92    0.96 
      4        111.00     108.14       193937      157.04   0.44     3.38    0.85 
      8        185.00     177.51       236287      191.33   0.53     4.12    0.52 
     16        421.00     399.96       209734      169.83   0.47     3.66    0.23 


Back to Scaling Comparison Main


Powered by Plone CMS, the Open Source Content Management System

This site conforms to the following standards: