A question regarding polycounts and in game performance

[alan2525]

Just a couple of questions regarding texturing and 3D modelling for low poly in game models.

As far as in game performance is concerned what is more important, poly counts or texture resolution?

I’ll give an example of this below. The two models are a very basic demonstration of a protruding nut and bolt from a vehicle in a game.



The model on the left has 40 polys and the model on the right has only 32 polys. This is because the model on the right is composed of two separate cylinders, whereas the model on the left is a continuous mesh.

There is actually an area on the top of the lower cylinder on the model on the right which would not be seen in game as it is covered by the top cylinder.

As this example is so low poly there is only a small reduction in polys, however if this feature is repeated many times as a detail on a tank for example the reduction in polys could be significant.

There is also the significance of the part of the texture which is concealed by the model. If this was a large area then there would be a large number of pixels which wouldn’t be seen in game however these would obviously still use memory and could cause a performance issue?

Which model would be preferable in game -one uses less polys but the other would use a larger texture and part of that wouldn't be seen in game?