Well, I did mention that the Mac version of Vue D'Esprit *is* AltiVec enhanced - in other words, it makes use of the Velocity Engine too, so the comparison with the Athlon is probably a fair one in that respect. I'm also pretty sure the PC version makes use of whatever AMD's current SSE equivalent is.
However, I don't see the performance disparity as necessarily meaning anything much in the "processor X is faster than processor Y" sense. The scenes I was testing were big, showpiece scenes used by Vue D'Esprit's publisher to show off the application. So the polygon counts were high, which suggests the scenes were also pretty memory hungry. In those circumstances, DDR has an advantage over PC133, esp if the data access patterns are cache unfriendly. I've no idea what the data locality of the dataset was like, either - a complex raytraciing scene with volumetrics, depth of field, motion blur, lots of transparency and reflections, blurred transparency, etc could potentially have lousy data locality, which would take away any benefits from that big L3 cache.
Then there's the dual processor aspect. I'm not too clear about how Vue splits up a scene for rendering, but if there's rays bouncing everywhere, there's potential for data contention, cache invalidation, all sorts of fun stuff like that.
So, basically, while comparing a notionally 2GHz-ish Athlon with two 1GHz G4s might *seem* a reasonable thing to compare, I think comparing a 1GHz Athlon with a single processor 1GHz G4 (esp if it's a new one with DDR 🙂) might produce more interesting results, although there's enough differences in processor architecture that even those would be dodgy.
Sorry, spent too much time studying processor architecture and performance in my youth, which has resulted in a strong belief that system X being faster than system Y on application Z means only that system X is faster than system Y on application Z 😀
