STREAM Results for PowerMac G4/867

From: Stefan U. Hänßgen (stefan.haenssgen@ginit-technology.com)
Date: Thu Aug 16 2001 - 01:33:03 CDT

  • Next message: tls@reefedge.com: "STREAM results for DDR Athlon and interleaved SDRAM P3 machines"

    Hi,

    just came across your STREAM benchmark and had to try it on my new G4 "Quicksilver" - I'm not sure how much you're still actively interested in current results, since the ones posted on your web site don't show more recent machines, but here they are anyway :)

    I'm quoting the results in the alphabetical order the different Mac versions of the benchmark suite are located in the folder. Do you have a theory why the 601 optimized version of the benchmark performs so much better (>70%!) than the rest? The G4/867 has 256kB of 2nd level cache and 2MB of 3rd level cache, but that shouldn't make any difference since the test data doesn't fit completely in either. Interesting...

    yours
            Stefan Hänßgen

    -------------------------------------------------------------
    This system uses 8 bytes per DOUBLE PRECISION word.
    -------------------------------------------------------------
    Array size = 400000, Offset = 0
    Total memory required = 9.2 MB.
    Each test is run 10 times, but only
    the *best* time for each is used.
    -------------------------------------------------------------
    Your clock granularity/precision appears to be 1 microseconds.
    Each test below will take on the order of 16257 microseconds.
       (= 16257 clock ticks)
    Increase the size of the arrays if this shows that
    you are not getting at least 20 clock ticks per test.
    -------------------------------------------------------------
    WARNING -- The above is only a rough guideline.
    For best results, please be sure you know the
    precision of your system timer.
    -------------------------------------------------------------
    Function Rate (MB/s) RMS time Min time Max time
    Copy: 376.5149 0.0171 0.0170 0.0174
    Scale: 352.6559 0.0182 0.0181 0.0184
    Add: 419.2872 0.0230 0.0229 0.0232
    Triad: 418.9944 0.0231 0.0229 0.0232

    -------------------------------------------------------------
    This system uses 8 bytes per DOUBLE PRECISION word.
    -------------------------------------------------------------
    Array size = 400000, Offset = 0
    Total memory required = 9.2 MB.
    Each test is run 10 times, but only
    the *best* time for each is used.
    -------------------------------------------------------------
    Your clock granularity/precision appears to be 1 microseconds.
    Each test below will take on the order of 16165 microseconds.
       (= 16165 clock ticks)
    Increase the size of the arrays if this shows that
    you are not getting at least 20 clock ticks per test.
    -------------------------------------------------------------
    WARNING -- The above is only a rough guideline.
    For best results, please be sure you know the
    precision of your system timer.
    -------------------------------------------------------------
    Function Rate (MB/s) RMS time Min time Max time
    Copy: 355.5753 0.0181 0.0180 0.0186
    Scale: 349.0592 0.0184 0.0183 0.0185
    Add: 402.3133 0.0240 0.0239 0.0242
    Triad: 409.2246 0.0238 0.0235 0.0259

    -------------------------------------------------------------
    This system uses 8 bytes per DOUBLE PRECISION word.
    -------------------------------------------------------------
    Array size = 400000, Offset = 0
    Total memory required = 9.2 MB.
    Each test is run 10 times, but only
    the *best* time for each is used.
    -------------------------------------------------------------
    Your clock granularity/precision appears to be 1 microseconds.
    Each test below will take on the order of 16411 microseconds.
       (= 16411 clock ticks)
    Increase the size of the arrays if this shows that
    you are not getting at least 20 clock ticks per test.
    -------------------------------------------------------------
    WARNING -- The above is only a rough guideline.
    For best results, please be sure you know the
    precision of your system timer.
    -------------------------------------------------------------
    Function Rate (MB/s) RMS time Min time Max time
    Copy: 629.7973 0.0104 0.0102 0.0112
    Scale: 615.9176 0.0105 0.0104 0.0105
    Add: 609.3690 0.0159 0.0158 0.0163
    Triad: 680.0793 0.0143 0.0141 0.0144

    -------------------------------------------------------------
    This system uses 8 bytes per DOUBLE PRECISION word.
    -------------------------------------------------------------
    Array size = 400000, Offset = 0
    Total memory required = 9.2 MB.
    Each test is run 10 times, but only
    the *best* time for each is used.
    -------------------------------------------------------------
    Your clock granularity/precision appears to be 1 microseconds.
    Each test below will take on the order of 16231 microseconds.
       (= 16231 clock ticks)
    Increase the size of the arrays if this shows that
    you are not getting at least 20 clock ticks per test.
    -------------------------------------------------------------
    WARNING -- The above is only a rough guideline.
    For best results, please be sure you know the
    precision of your system timer.
    -------------------------------------------------------------
    Function Rate (MB/s) RMS time Min time Max time
    Copy: 402.5410 0.0160 0.0159 0.0161
    Scale: 392.9756 0.0164 0.0163 0.0165
    Add: 615.2663 0.0158 0.0156 0.0159
    Triad: 471.3044 0.0205 0.0204 0.0206

    -- 
    GINIT Technology GmbH      stefan.haenssgen@ginit-technology.de
    Dr. Stefan U. Haenssgen         phone:          +49-721-96681-0
    Technologiepark                mobile:         +49-172-7193-544
    Emmy-Noether-Str. 11              fax:        +49-721-96681-111
    D-76131 Karlsruhe                http://www.ginit-technology.de
    



    This archive was generated by hypermail 2b29 : Wed Oct 31 2001 - 11:26:46 CST