Could someone please explain the performance cost with reguards to the Bevel / Face weight technique - There is lots of content out there on how to use the technique but not a lot on the performance cost!
My understanding is you got a "Free Bevel" due to the way that split normals works. IE when you add a hard edge it splits the vert into 3 normals and the smoothed, beveled + weighted version has the same number of split points.
The project I'm working on budgets are based around Verts, a prop budget for example is 1k verts if I add the weighted normals that would raise my vert count to 3k ish which is way over my budget but does the technique mean this is not a problem?
If there's strict vert budgets, it should be based off the vert count in engine, the vertex count in a model program is meaningless. It's 24 verts for the cube with hard edges, but the 24 vert beveled and smooth shaded version would get more verts once it has UVs (one on each UV seam).
in that third one you've got more 'split points' than 24 - you've not accounted for the verts on the bevel faces (another 32 at best)
and.. Quibbling over whether extra vertices get sent to the GPU or not is basically a waste of effort from xbone onwards - that stuff is cheap and unless you do something stupid your top couple of lods are almost never a problem in terms of rendering cost.
The primary impact on performance is almost always from having geometry that is too dense for the view distance or from overdraw (related)
Replies
and..
Quibbling over whether extra vertices get sent to the GPU or not is basically a waste of effort from xbone onwards - that stuff is cheap and unless you do something stupid your top couple of lods are almost never a problem in terms of rendering cost.
The primary impact on performance is almost always from having geometry that is too dense for the view distance or from overdraw (related)