"I think that, somewhat ironically, we might be seeing engines going back to “software” renderers. CPU core numbers are creeping up, and raytracing for instance can be rather easily parallelized (well…)"
I think that this line is actually blurring. It's a statement which made sense at the time, where GPU's were dedicated to specific tasks and the CPU handled everything else. Now that you can freely code with tools such as Cuda (and more yet to come) you can do the very same things which in the past would have been done on the CPU with the GPU. So you can either resort to standard solutions or code your own, depending on the level of R&D that the developer wants to commit to the project.
At the moment, Intel speculation about the superiority of their future architecture against nVidia looks like a panicked response after Cuda showed to be able to take over the supercomputer market (which, at a fraction of the cost, covers the processing power provided by Intel). Simply put, from Intel's speakings it looks like they want to completely cut ATI and nVidia out of the market, as it seems like they have took this issue as a "we or them" matter. Still, if Intel manages to handle with ease certain GPU specific tasks, this line would be even more blurred, and I can't see ATI and nVidia engineers simply disappear from the scene either way.
looks very good, i didnt think voxels could look that good,
Outcast was very complex and mixed various technologies. Terrain was voxel based, buildings and characters were 3D models, characters had bone animation, water was a vertex matrix which interacted with the player, and areas were really big. The AI was fairly good too, as it was a stealth game among other things.
Sadly, as much as I wanted to like it, I always though that the storyline was dull and the frequent dialogues futile (which isn't uncommon on todays games too, but it detracted too much from the otherwise excellent atmosphere).