Technonut's results are extremely helpful, but I can't hide my dissappointment. His P4M/1.7 requires 1.70vcore to reach a stable 2.7ghz, and 1.45vcore to reach 2.34ghz. (This seems consistent with the other results.)
This suggests that P4M/1.7ghz is little better for overclocking (as a CPU) than a desktop P4/2.26B or 2.4B. These latter chips already achieve comparable MHz at comparable voltages.
There may be 2 potential advantages to P4M over "regular" P4:
1) The P4M has a much lower multiplier (12x) on desktop boards, allowing insanely high FSB. However, high FSB barely helps because your memory is still single-channel and limited to 400-450mhz. Unless you can run P4M on a dual-DDR GBay board at very high FSB, you don't get much benefit from having 900mhz (QDR) FSB.
2) At the same MHz and voltage, P4M seems to consume roughly 15% less power than its desktop equivalent. (By this I mean, a P4/1.8A versus P4M/1.8, overclocked to any particular speed/Vcore; or P4/2.4B versus P4M/2.4). This could make cooling easier, or allow using a quieter fan.
I don't mean to spoil the party; all the testing has been worthwhile, and P4M does have advantages. However, as of right now, it is not quite the wonder chip we hoped it would be. (For that matter, neither is GBay.. but if it becomes possible to combine the two effectively, the results could be impressive.)