RE: Model with different materials: best practices?
Well, for our project we ideally would like to be able to change armour and the like to give the player the possibility to upgrade... but: as we want to do a strategy game, the impact of a single character on the performance is significant. Because there will maybe be hundreds of characters on screen. So I think, othere than aggressive LoDs, grouping Stuff together might be a good idea, even at the cost of having 3 times the same character in the game files, once with each of the three sets of armours.
there is also the question, if it makes sense to do the same with weapons / equipment, and this is where this really becomes impractical. Thats why I'd like to have some insight, if an extra drawcall makes a big impact with EE... because it did with other engines (tested with tousands of individual 3d objects, but if 2000 objects bring the FPS to a grind, a few hundred might also have a big impact... so every drawcall might count). I made a test once in EE with the standart Warrior Character, without any equipment, the FPS dropped below 30 after some hundreds on screen, 2000 brought it down to 3-5 FPS. Now I'm fully aware the EE Characters are RPG Optimized characters not really suitable for this kind of game (at around 7000 tris, without LOD they would be way overboard), still, this means if you want to do an RTS, every performance bottleneck counts.
Also I noticed that having not all of them visible made no difference for the Engine, but I think this is something already on the roadmap.
Most probably I need to test this.... is there any function that returns me the amount of drawcalls sent to the graphics card by EE? Would help with testing the impact. Also, is there any way to get the amount of Tris rendered / VRAM Usage?
|