Light Probes vs 1 Realtime light

Hi. I just recently realized that when light probes are used more draw calls are needed because dynamic objects that use different probes are in different draw calls. I know it's been documented that we should use light probes instead of realtime light. I wonder, if I just use 1 directional light to light the dynamic objects instead of hundreds of probes, which one would in theory be better for mobile. I know it would sacrifice the aesthetics a little as all objects will now have same lighting, but would the performance be better due to less draw calls in theory?

 

Dynamic objects need different draw calls anyway though. (Unless they are dynamically batched, which has very restricting rules).

 

Hi yes that what I mean. With probes, they end up not being batching, but with realtime lighting they are batched. So I wonder what's worse, many draw calls but no dynamic light calculation, vs 1 draw call from 1 realtime light but with realtime light calculations.

 

Are they? I'm not sure that's true?

 

I used frame debugger to check. The only ones that didn't get dynamic batched are if they are over 300 verts. But the lighting looks horrible because it doesn't match the scene. But yeah, I wonder if drawcalls are worse or dynamic light. I guess I'll have to profile it with a test scene as it doesn't seem like something people talk about.

 

I wish I could use it. But gles3 performance quite bad compared to gles2 for lower end devices. So with instancing it actually performance worse due to gles3.