Ok - results.

I used a 4 point sine wave, and heavily down sampled. (like Gabriel
did). Intuitively, I personally would have agreed with Gabriel - 
Gen08 *does* look "smoother". The peaks are definitely rounder
with Gen08. However, as surprising as it may seem, upon analysing
the output with a spectrometer, oscil3 is actually doing 
a better job. The signal to noise ratio is reported to be
13.1 dBA for oscil3, and 10.3dBA for gen08. Total Harmonic
Distortion + Noise (THD+N) for oscil3 is 22.1%, and for gen08 it
is 30.7%. Again, this is in oscil3's favour.

So I don't think we can say gen08 is better. It is certainly different
though.
And oscil3 may even be superior to gen08 if my measurements are valid.

I have again attached my orc & sco files.

The spectral analysis software I am using is SpectraPRO
from http://www.eskimo.com/~cetacean/software.html
(unfortunately I only have the demo copy and mine expires
in about 11 more days. Does anyone know of a cheaper program
than this one that can do the above noise calculations?)  
Note that I had to paste together multiple cycles in order to
get a long enough analysis duration for SpectraPRO to work.
I think it has a bug with very short waveforms. Yes, I *was*
very careful in my cutting and pasting. :)

Greg.

ORC
----
sr = 44100
kr = 100
ksmps = 441
nchnls = 1

instr 1
aout oscil3 30000,p4,1
out aout
endin

instr 2
aout oscil 30000,p4,2
out aout
endin

SCO
----
f1 0 4 10 1
f2 0 2048 8 0 128 1 128 0 128 -1 128 0 128 1 128 0 128 -1 128 0 128 1 128 0
128 -1 128 0 128 1 128 0 128 -1 128
i1 0 3 86.1328125
i2 3.5 3 21.533203125 
e