A question of optimization.

I am going to make a fence model. I need to know which is more efficient.

1 alphatest, no cull texture for all, effectivly doubleing the polycount (only a quarter of the polygons need to be transparent)

Or 2 textures, 1 with alphatest and no cull. Doubling the drawcalls.

Which is most efficient.

I expect the polycount to be 80-120

same number of polygons but more likely for shit to go wrong with two plains in the same spot. plus it’s more complicated. use one that’s double sided.

EDIT: i think i just misread what you said; it doesn’t really make sense to me at all now. are you talking about textures or the model?

Well nocull will make it so the texture is shown on both sides. Effectivly doubling the rendered polygons.

My model mixes transparent material (wire) with non transparent (poles)

Is it more efficient to make the entire thing use 1 material, so there is only 1 draw call. (but then the poles will have nocull on them)

Or have the model with 2 materials. 1 with no cull for the wire, 1 withnothing fancy for the poles? Doubling the draw calls.

If I remember correctly, the HL2 fences just use a single material. So I suppose that’s more optimized. It’s not like either way has that big of an impact on performance. Actually, depends if by “material” you mean actual texture sheets, or .vmts. Also, alternatively you could just duplicate the chainlink plane and flip it, and then not have $nocull in the .vmt so that the opposite plane isn’t drawn.

The hl2 fences use 2 materials iirc. That flipping plane thing might work! I’ll try it out. Thanks for your help