Actually I can disagree with that. When rendering large amounts of text on the screen - you'll soon notice performance drops (mostly from individual texture binds for each letter). The way to solve such a problem is to batch all the geometry into a single mesh and render it all in one draw call. Then the problem arises - how do you bind all images for each individual letter - because there is a limitation on how much images can be bound per 1 draw call. And the answer is a texture atlas. Because I've already implemented "batch buffers" for text, I decided to use them for chess pieces rendering - because why not. I know I will not save a lot of performance - but anyway I thought it would be nice to reuse part of the engine that was specifically created for such scenarios.
@@dzima-create hm it also depends on the implementation for sure, if you were to bind each individual letter that for sure can't be good, usually letters are held into atlases. But for smaller things like a few textures I don't think you will see any problem