PC speaker

I've been reading about outputting PCM audio through the internal PC speaker using "Pulse-width modulation". If I understand correctly, this is acheived by retracting the speaker cone before it has a chance to fully extend. My question is now this: would it be possible, by spending all CPU resources on this task, to keep the speaker cone in its current mid-extension position, by rapidly (as fast as the CPU will allow) toggling the current to it?

The reason I'm asking, as some may have already figured out, is because by freezing the state of the speaker between samples would, according to my theory, output a higher quality signal, because of lowering the amplitude of the carrier. Is this correct, or have I completely misunderstood something?