Would a user created city have the same FPS impact as one completed by a development team?

[Mr_sukebe]

Here's a question.

A mission creator cannot to my knowledge change underlying ground textures, add airfields roads, or change elevation on maps.

We can however add new buildings and clear existing ones.
If a user were to build a city with the same number of buildings that exist in a professionally created theatre, would that have the same resource requirements on rendering and similar?

The background to my question was a thought about the viability of populating cities on maps that are outside of the detailed areas.

[cfrag]

On 12/19/2024 at 4:35 PM, Mr_sukebe said:

If a user were to build a city with the same number of buildings that exist in a professionally created theatre, would that have the same resource requirements on rendering and similar?

That rather depends on how the rendering engine works and its built-in optimization. If we assume DCS's rendering engine and in your hypothetical case a city is user-built (in ME) by adding one object (building, street, lamp, car, fence, whatever) after the other, then the result will be orders of magnitude worse than if that same conglomeration of objects was created in a 3D app (like 3ds Max or Maya), processed, optimized (!!!) and added as a single object (that's how it's usually done). The reason for this is that the latter requires only a single call to GPU draw: fully scale the draw call over all available GPU cores (with many single smaller objects this doesn't scale well), with all vertices having to be camera-transformed just once (especially those that are shared: if two buildings touch walls there are a number of vertices that can be eliminated during optimization - in cities that happens a lot), all textures loaded once, all LOD texture calculations being done once, and the entire thing is then z-transformed and drawn once. For thousands of little objects, there'd be next to no scale effect on transformations, millions if not billions of pixel double-processing (the same pixel being dawn over and over by different objects), same for texture buffering and LOD, making the result greatly less efficient and a much greater drain on performance.

Note that this is only true for scenes rendered with painter's algorithm (almost all are, ray-tracing as main algorithm as in RTX isn't standard yet because of its highly inefficient method). IIRC, DCS uses the painter's algorithm.

So, I expect that a city built from individually placed objects in ME would have an exponentially worse (by number of objects) performance curve compared to exactly the same set of source objects, pre-optimized and professionally pre-processed into a single object.

 

Edited December 25, 2024 by cfrag