I have noticed no effective hit for high poly models. This is different then what i'm used to with Lightwave... but the real difference lies in the file size. (800x600 vs. 2000x3000)
The number of polys in an emitter DOES seem to have a huge impact on the scene, but I don't agree with the noise statement. With my work I haven't been able to draw that conclusion, that more polys in an emitter equals more noise. No, the only effector for noise I've seen is the amount of light in the scene, and the materials themselves. Metal creates wierd 'only in some spots' noise. SSS creates noise like crazy.
As for render time, more polys in the emitters does affect render time. As it does in many other packages. With shade, a multipoly emitter increases accuracy but also render time (exponentially) Having said that, there seems to be a limit. I have tested with extremely complex emitters (5000 polys +) and yes, it's slower, but not much slower then 25,000 polys. So, though a 4 poly emitter is much faster, it seems that it's not a mathmatical progression, otherwise at some point the complexity of the emitter would equal unusable rendertimes... 100+hours for level 10, and with my own experience at least that doesn't seem to be the case.
I would be really interested in some solid info from NL regarding this issue. It would be nice to understand what effect subdividing a model has on an emitter and how it changes the number of rays, etc.
cheers
