... the red trace is the voltage, and the purple trace is the current into a switching power supply (as the power supply's load changes, the current spike tallness changes as well). Note that the current is zero most of the time (2/3 of the cycle or more), and large spikes occur at the time of the peaks of the voltage waveform. This shows only one phase. Now think of the next phase, and the timing of the spikes on that phase happens during the zero current period of time on the 1st phase. Likewise on the 3rd phase. None of the current spikes on the 3 phases overlap in terms of time, and thus the neutral gets hammered with current spikes 6 times per 60Hz cycle, while one hot phase wire only sees 2 current spikes per 60Hz cycle. Thus the need for a heavy neutral wire.
Not only can the neutral get hammered by those narrow pulses (at up to 300% of the current of one circuit), but both the neutral and phase lines get hit further with increased I2R losses due to the high current peaks. The narrower the pulse, the higher the current, and you square that for the losses before averaging it back out over time. So the end result is when a pulse is N times narrower, you have N times more I2R loss (N^2 loss during the pulse times 1/N time ratio). These losses will also affect the OCPD thermal element. So what should be 15 amps of load can trip a 20 amp breaker when it has peaks 3 times normal (9 times the dissipated watts for 1/3 time). It basically becomes a derating issue when it gets bad.
Fortunately switch-mode power supplies for computers are getting better.
