Afaik the power consumption (and heat generation) is not only dependant of the frequency but as well of the number of transistors that are currently active. Since there are millions of transistors inside a CPU this makes the difference.

CPU power dissipation is increased by voltage to square and linear to frequency. And it decreases with structure size.
That are the basic rules.
But there are leak powers and more that affect this not so easy calculation.

A major development has been with the Pentium 3 mobile chipsets, where they have started with speed step.
During the Pentium 4 fail the Pentium 3 mobile has been developed to the Pentium M and the result were the early Core processors which introduced speed step on desktop processors.
Yes, I am aware that Windows uses the halt instruction if the computer is idle. But speed step is another thing 😀

It's interesting cause the same watt difference is seen during a memtest run (72 watts at the wall plug). So the videocard, the hard disk, cd readers etc.. while powered can't be the reason. I suppose internally the Pentium III already slowdown itself in idle even the early core on these old chipsets. Can't remember in which way.