EdipisReks, I'm not sure how sound your logic is here...you say that as compilers and software become optimized and the G5 ramps up in speed, it will be able to compete with high-end P4's...but your assumption kind of rests on something silly like "the P4 isn't going to get any faster". One word for you: Prescott. When Intel released the Pentium 4, they had a clear roadmap of where they were going with it. Everyone (at least, everyone who stayed informed), knew about the upcoming Northwood core revision, and nice people like Wingznut were there to let us know that it would offer a significant performance boost.
Sure, the P4 wasn't a stellar performer when it debuted. BUT, it had tons of performance potential and a clear way to reach it. Yeah, the 2GHz G5 is a big leap in the Macintosh world...but it's not like the PC world doesn't get those leaps, either. Prescott will be able to scale clockspeeds wildly, whereas it has yet to be shown exactly how much headroom the G5 has to offer. Prescott will have larger caches, improved Hyperthreading, and (eventually) a faster FSB (with memory to match). It also is manufactured with strained silicon. What does the G5 have down the line?
I'm not saying that it has no headroom, but Apple/IBM have yet to demonstrate that it does. Intel (and AMD to some extent) make their roadmaps relatively well known, and they show obvious growth and progress. Apple, however, had stagnated up until recently. What's to say that they won't do the same again? Will the G5 scale a few MHz, then sit a lag behind the rest of the world, just like the G4?
There's no doubt that the G5 is an impressive effort from Apple, but it still pales in comparison to what Intel (and perhaps AMD) will have to offer. A 1GHz bus is great, but only if you've got memory to use it. A new architecture is nice, but how long will it last? Netburst has a lot more headroom than what it's showing right now.
P.S. I'm not sure what kind of 3D animation work you do with 3D Studio Max, but I found that the 4.2.6 upgrade that optimized for Intel's SSE2 instructions really didn't speed anything up except for a few special cases (intensive raytracing). If you're doing something that utilizes 3ds Max's raytracer that much, you should consider upgrading to a real raytracer, like finalRender or VRay. They offer much better speed boosts than Intel's SSE2 instructions did.