A healthy debate is going on on the official forums regarding module size as applied to persistent worlds. Mr. Sawyer has weighed in several times in the thread:
Zarathustra217: I am a bit disappointed in your faith in your average (or even above-average) person. But moreso, I wonder why you even bother to cater for this usergroup? Obviously, E4 Oblivion will still hold the crown of visual beauty.
I'm not so sure about that. Every screenshot that has been released so far for NWN2 has lagged by a few months. Within the last two weeks alone, point light shadows and water have gone into the game.
Art quality is obviously very subjective. Very frequently, art can be popular without all the latest bells and whistles (Guild Wars). On the flip side of the coin, there are games like EQ2 which have caused a great deal of grumbling despite having great graphical features. NWN2 will have just as many graphical features (if not more) than Oblivion. For one thing, our characters self-shadow. It's possible that people will like the look of Oblivion more than the look of NWN2, but it won't be for lack of technology.
Alerias05: I suspect this will also be the case in NWN2. The devs havent took the time to make such optimisations; theyve been cutting corners ever since the launch of the project ...
We'll need to stick to small I'm afraid. 4 GO ram minimum for a decent server rig.
And frequent reboots because of the likely memory leaks too.
Actually, it was specifically their recent attention to memory profiling and optimizations (including fixing memory leaks) that caused some of this information to come forward.
Pentane: I don't think that's what he meant percisely (I honestly can't see each fragment of the OC being only 10 exteriors total) and I think he meant "large exteriors" as in 32* 32.
Exteriors are kind of tricky for a number of reasons. First, any given X by Y "tile" dimension for a NWN exterior is dwarfed by a NWN2 exterior. A 32 by 32 NWN2 exterior is enormous. Populating them with realistic foliage can make their memory size jump up quickly. Though this fine control over object placement does tend to make walk meshes more complicated, Jason Keeney recently implemented some fancitude that helps abstract clumps of static objects into simpler shapes. This also helps on interiors. Hopefully with more improvements and optimizations like this, the currently complex walkmeshes will become significantly smaller.
The thread begins here.
