I just realized I was clocked at 3.33GHz when I ran the previous pass. I clocked down to 3.33 a couple of weeks ago to generate some numbers to compare to some 3.33GHz numbers from a previous configuration (not like the difference is noticeable in real use). Here are the 3.42GHz numbers:
4.442 Secs, Addition floating point (0><1) 100.000.000 times
4.476 Secs, Subtract floating point (0><1) 100.000.000 times
4.475 Secs, Multiply floating point (0><1) 100.000.000 times
36.504 Secs, Divide floating point (0><1) 100.000.000 times
4.46 Secs, Addition integer (1><10) 100.000.000 times
4.63 Secs, Subtract integer (1><10) 100.000.000 times
35.899 Secs, Multiply integer (1><10) 100.000.000 times
4.743 Secs, Divide integer (1><10) 100.000.000 times
I think the test is not as clockspeed depend as you think. It appears that way when run on different rigs running the same cpu at different clock rates. Small synthetic benchmarks are always this way--as they fit in the cache leaving only clockrate to differentiate the results.