Modeling Buildings or Building segments?

Development World Creation
, , ,
1

I’m not a modeler, just a programmer and music developer, but I’ve started learning blender to help out on minor levels that the modeler doesn’t have time for. I chose a small side mission level with very low light so I my inexperience would be hidden, and populated the landscape with the necessary buildings and roads. To prototype several brick buildings, I grabbed the large collection from MegaScans and used blueprints full of building components to create the buildings necessary. These buildings were pieced together by the wall sections found on MegaScans. I very quickly found myself with a small level containing an astounding number of polygons, far more than anything the modeler creates. On closer examination, it was that the MegaScans components were extremely high poly due to very complex geometry, probably very good for renders but not so good for games without some decimation or redoing the wall components. I ended up with a 20 FPS drop with my modular buildings and the level wasn’t close to done.

That’s where I grabbed and started learning blender so I could redo the meshes on the building components. With the help of the modeler, the building segments went from 1000s of polygons to 100s. He pointed out that if I just made each building in blender, the entire building would be lower poly than a few segments of a modularly assembled building. This turned out to be true and additionally provided more flexibility in creating the buildings such that they didn’t look so cookie cutter. Unfortunately, neither of us know if the number of polys mattered as much in UE 5.

So I’m stuck with this question. With Unreal 5, which is better from a poly and performance standpoint, building a building from components in unreal or building the building in a modeling tool and and bringing it in as a single mesh? I would assume that if numerous repetitive meshes such as, for example, a window segment used 1000 times, that window segment would only have one copy to the renderer even though it’s far larger in a poly count. Yet overall the individually modeled buildings have vastly lower poly counts. Are the number of polys as relevant now with UE5?

2

You’ve got the right approach. Those Megascans models will have your PC on its knees before you can blink.

A real UE5 enthusiast would just say ‘but we have Nanite, you don’t need to worry about polys any more’. Nanite certain works wonderfully in certain circumstances, but for my money ( and for most average PCs ), low poly still wins hands down.

You can also use instancing to improve the situation ( where all those windows because copies of the same thing ). Give it a go, and see for your self. I recently change a mesh that I had instanced to blueprints all over the map, and the frame rate went up!

So experiment :wink:

3

Thanks, we had our suspicions. Seems that picking the method that produces lower polygons would be best. I’ll look at the instancing and possibly come up with a hybrid. Highly pertinent buildings need to still be modeled just because they get closer player scrutiny (and we don’t want someone saying “They just ripped that window/door/wall off from…”), but possibly background buildings could be modular creations.

We were guilty of assuming that possibly nanite would make this moot. We haven’t worked with it yet because we are on Macs.

1 Like