Slow render of object in play mode

Have a model of building, wich draws fast enough in edit mode, both in Unity and modelling applications. But then running game mode, draw speed falls down catastrophically, no matter, the building have small or big scale, and no matter is visible in viewport or behid the camera. Is someone know, what’s the problem? I am using the last version of Unty 4.0.1f2.

Tried some more complex buildings, with bigger polycount, and they work just fine.

It’s most likely not the mesh itself (in terms of the polycount) which is causing that trouble. How much draw calls do you have once the object is visible? How many and what kind of materials are assigned to the object? Any rigid bodies or other stuff?
And last but not least - are you absolutely sure that it’s that specific object causing trouble?

Draw calls: 4-32, depending is the building fully visible or lower enrance section is off screen.
Material: 1, no matter which - simple vertext lit drawed equally slow, no matter, light is present in the scene, or not.
Triangles: about 17000
Textures not used

Scеne is drowed at 4-5 fps. Also depends of optimization flag, and was mesh intially combined in modelling application or not (combined is slower). Also, as I said, this framerate stay the same, then camera look away from building. Probably, something wrong on software processing of mesh, not videocard.
No colliders for this mesh was generated.

Maybe it’'s because mesh have mаny separate small objects.

Also, scene handled and twisted fast and smooth in edit mode, but not in play mode.

I agree with you, that’s most likely not a rendering problem. Is there possibly a script attached to the object which only runs in play mode and causing a lot of calculations? Like nested loops, a massive Debug.Log() to console or similar?

no, just mesh without any colliders or scripts.

Could you pack and upload an example? Sounds interesting to tinker on!

I think i found the problem.

The mesh consists of many small subobjects, for example, each window (two triangles) is a separate object, each wall segment, etc. Several hundred subobjects total. For test perposes I created a scene from 10000 separate meshes (each is a single quadrangle), a grid 100x100, imported it as mesh, and drop down frame rate to 0.1 fps. So, object list in the scene processed buy CPU, and many small objects is much worse, then several high polycount objects.

The fbx-file is too big to attach (22 MB), anyone can create a grid 100x100 quadrangles in any modellng application a play with it, but I think question is solved.

Also, combinig these submeshes into single mesh don’t solve the problem, because internally the subobject structure is stil exists. Mesh optimization in Unity also do not help. So, the only way to solve the problem - boolean union of submeshes into truly single mesh, after that subobjects can’t be easily separated.

What program have you modeled this in? In the case because of merging all those “base objects” as i call them from your modeling suite, chances would be there are a lot of face selections still in the objects structure.

How did things look when you tried to import the model back to the modeling suite?
(Cinema4D i.e. also exports "Polygon Selection"s from all the seperate parts, at the least when different materials, that were combined which can be interpreted as submeshes if you didn’t make an “object bake”)

I used maya. In maya combine and union meshes is different operations. Combine simply treat several submeshes like single mesh and can be undone by separate. Union is volume union, some new faces and vertex might appear, and some lost. Then back imported (fbx), it looks the same as exported. Last time I combined 10000 quadranles. I didn’t use union, because this operation works for 2 meshes, so I have to select consequently each quadrangle and unite it with overs.

May be some simple way exists, but I am still beginner in maya.