koobas.hobune.stream
Why has gpu instancing worse perfomance than dynamic batching?
Hi!
In my case I render thousends of quads sharing the same mesh and same material. I tried gpu instancing instead of dynamic batching and the performance got worse. Whereas the drawcalls remained the same, the number of batches were dublicated. Render time for cpu and gpu did not change.
Best, Andre
Edit: In Frame debugger i can see that roughly half of the meshes are instances and other half are without instancing, what might explain the double number of batches. Still no clue why there are still meshes not instanciated even though they share same meshes and materials.
