can packing 3 grayscale texture into a single texture and read it once actually become significant perfomance ?? i dont see any difference at all and i removed my graphics card for this purpose, what is the real perfomance win to pack textures then?? the result are still the same but im not sure if im measuring correct but i would know if anyone know the number or the % in increase based on your experience ? also asked chatgpt but said terrified fake lies
Hola Gabriel!
Yep, the aim is precisely to make data loading more efficient and improve performance by reducing memory accesses and draw calls. Many PBR textures – for example – really only need one channel (8 bits), so if you load three channels in a single greyscale texture (24 bits), you’re actually wasting channels that could be used to store other data.
On average, you should expect VRAM reductions of 30% to 40% and FPS increases of 10% to 20% (don’t ask me for references, this is the numbers that I remember). Also, I don’t know how you are measuring performance, but if the scene is too simple (or if the CPU is overloaded), the optimization will not be reflected in the frames per second (FPS). If you really want to test this out, you should use a complex scene (not a benchmark but complex enough), and measure the milliseconds per frame (ms).