My friend with an Athlon XP 1600+ (1333mhz) tried it and got:
FPU: 181.91M per sec, 7.70161 clock ticks
MMX: 154.27M per sec, 9.0815 clock ticks
Another with vanilla Athlon 1400mhz:
FPU: 184M per sec, 7.61 clock ticks
MMX: 157M per sec, 8.94 clock ticks
It seems that Athlons work better with the regular FPU rather than with MMX. Then again, I haven't done any serious AMD tuning.