If i have a polyphonic FM synth using poly~ and i want to oversample, what do people think is the most efficient strategy?

1 - use one, single voice "master poly~" patch that is upsampled (with antialiasing filters before the outputs). Inside this poly would be another poly with say 8 voices for my synth.

2 - have an 8 voice poly~ and put another (oversampled) poly~ inside each voice for as little of the signal path as possible (including antialiasing filters before downsampling).

I suppose i have to try it and see...

oli