Face Count vs Polygon Count

We all know that the basic geometry we’re concerned with in real-time rending is usually mentioned in conversation and text as the Polygon. And we also know that a polygon is typically made of 2 faces, especially when dealing with quads. So when we are working with our modeling team, do we need to be concerned with the face count or the poly count? In most cases, the poly count will be half the number of triangles and give a false perception that the geometry load on the render is less than it is. Or is that true? In the case of a quad, we’re really only adding another vertex because the two at the joining edge between two triangles are shared.

My question then is which is the true measure of performance? Do you look at the face count or the poly count?

Thanks for your input.

GPUs only render triangles, so of the two measures you mentioned, the one that matters is triangle count. Vertex count is a separate issue and may also be measured differently between modeling app and GPU. Generally, each vertex that has a sharp edge or uv seam must be duplicated, which means that having many sharp edges or uv-seams will increase your vertex count by a significant amount.

Thanks for the answer.