Trainz is both blessed and cursed by the community-created content business model. On the one hand, we have a huge range of content covering a vast range of prototypes, and it's possible to put together some fantastic routes, running exquisitely detailed models. On the other hand, this community-based approach means that the game engine has to cope with some hugely inefficient models that haven't been optimised like professionally produced game assets. Latest versions have introduced many features to make the game more efficient, such as mesh stitching, libraries, DXT compression etc, and look better but the game will always struggle as long as there's a huge base of inefficient content (some of mine included) in use.
N3V appears to be doing its best to improve standards by more stringent error checking, but the recent fuss about single colour textures and thumbnails demonstrates that there appears to be a bit of community resistance to these steps. The lack of rigid standards in the early days may well be a factor- I contrast this to the original guides for registered creators for Rail Simulator, which were pretty prescriptive. Trainz has always made content creation much more accessible, but at the cost that inefficient content can find its way onto the DLS fairly easily.
In my opinion, a route that is built entirely from assets built to modern standards, with properly configured LOD, mesh libraries, environment and normal-mapped materials, baked shadows etc. has the potential to look and perform as well as many more modern games- albeit without some of the fancy environmental effects. To achieve the highest standards requires rigorous discipline on the part of the route builder, with very careful asset choices, judicious use of repetition (many routes seem to use an enormous palette of materials and assets, when one that is already used elsewhere in the route may be close enough to achieve a good effect at less cost to performance). All of this can easily be ruined by a few poor performers such as a station lamp with many thousands of polys, huge, single colour textures and no LOD. Although there are some pretty good routes out there, I don't think it would be possible to put together one to the standards I've described without some serious effort, probably from a team, to put together a set of efficient, route specific assets.
Most of the time we're happy to sacrifice some of these inefficiencies for a good looking end result, and I think it's a testament to the performance of the game engine that it manages as well as it does in the face of all the rubbish that we sometimes throw at it.
R3
