Well, true, looks like a draw. Having said this, the performed tests are likely to be unreliable, as they have been executed in runlevel 5 with X and shed lods of various applications running. It would be fair to run the tests in init 1 on both systems. As I have lods of work to do for now, i'll try to make some time in the evening to reboot the machines and post more reliable results from 5 runs Andrei On Wed, 2006-03-15 at 12:50 -0500, Lennart Sorensen wrote: > On Wed, Mar 15, 2006 at 05:33:02PM +0000, Andrei Mikhailovsky wrote: > > Here is my info: > > > > AMD64 3200+ > > # nice -n -19 /usr/lib/atlas3-test/xzinvtst > > > > NREPS ORDER UPLO N LDA TIME MFLOP RESID > > ===== ===== ===== ===== ===== ======== ======== ============ > > > > 0 Col GE 100 100 0.005 1594.32 9.536429e-03 > > 0 Col GE 200 200 0.033 1936.05 1.116727e-02 > > 0 Col GE 300 300 0.096 2247.54 1.244014e-02 > > 0 Col GE 400 400 0.213 2401.87 2.111688e-02 > > 0 Col GE 500 500 0.392 2549.50 2.246591e-02 > > 0 Col GE 600 600 0.660 2616.94 1.968228e-02 > > 0 Col GE 700 700 1.017 2697.10 2.345843e-02 > > 0 Col GE 800 800 1.501 2727.98 2.368572e-02 > > 0 Col GE 900 900 2.041 2856.67 2.717980e-02 > > 0 Col GE 1000 1000 2.778 2879.13 2.801465e-02 > > > > 10 cases: 10 passed, 0 skipped, 0 failed > > > > And here is an intel box (as you can see intel slighly faster in this > > test) > > Looks like a draw. The first 5 lines the AMD wins, and the last 5 the > intel wins. I would even say the ones the AMD wins are won by more than > the intel wins the last ones. > > > Intel P4 3.2Ghz with HT > > > > nice -n -19 /usr/lib/atlas3-test/xzinvtst > > > > NREPS ORDER UPLO N LDA TIME MFLOP RESID > > ===== ===== ===== ===== ===== ======== ======== ============ > > > > 0 Col GE 100 100 0.007 1138.73 1.006733e-02 > > 0 Col GE 200 200 0.043 1485.82 1.071981e-02 > > 0 Col GE 300 300 0.106 2035.50 1.416881e-02 > > 0 Col GE 400 400 0.234 2186.32 1.663189e-02 > > 0 Col GE 500 500 0.431 2318.80 2.447828e-02 > > 0 Col GE 600 600 0.646 2673.66 2.339260e-02 > > 0 Col GE 700 700 1.014 2705.08 2.376800e-02 > > 0 Col GE 800 800 1.445 2833.70 2.600841e-02 > > 0 Col GE 900 900 2.010 2900.72 2.869086e-02 > > 0 Col GE 1000 1000 2.707 2954.64 3.397687e-02 > > > > 10 cases: 10 passed, 0 skipped, 0 failed > > Len Sorensen > >
Attachment:
signature.asc
Description: This is a digitally signed message part