I have a confusing problem i hope you guys can help me with.
Im trying to reduce drawcalls. I mean, really reduce it to a bare minimum.
A model has only 1 material (base color, metallic, height, emission, normal map).
If im using the URP and remove the PP volume and everything that requires a drawcall (skybox for instance),
i get 5 batches as a starter base. 8 setpass calls. If i have the directional light selected.
Now for each model (1.5k triss, 1 material) i have in the scene, it adds 2 batches if the material is opaque and doubles the triss to 3k. if i switch it over to transparent it gives me 1 batch and 1.5k for the model.
is it suppose to do that? isnt it 1 material = 1 batch if i dont have any extra hooked up, like PP and stuff like that? and transparent? shouldnt transparent materials be more expensive?
The Standard RP gave a smiliar result.
everything turned off, opaque gives 2 batches, transparent 1 batch.
but the starting batches is only 1, and 1 setpass call with absolutly nothing in the scene.
At least the setpass calls work when i group the models in one empty GO and add static to it. If i increase the model count by duplicating one, it stays how many copies i make.
But the draw calls increase with 2 for each 1 model i add. if im not using transparent, which is not what i want.
how can i make unity only draw 1 batch for 1 material in opaque? in both URP and Standard RP.
ask me anything and ill happily answer gaps in this explanation.
Hi, you can open the Frame Debugger to see how Unity renders a scene.
For example, in the picture below, URP is executing a pass called “DepthNormalPrepass”, which outputs scene normals & depth. (required by the SSAO effect)
On some platforms (usually OpenGL which Unity cannot copy depth), Unity will execute “DepthPrepass” to output scene depth if you/effects require a Depth Texture.
It depends, but for most cases, they are more expensive.
That’s because GPU cannot skip meshes that are behind a (semi-)transparent mesh.
This can lead to severe overdraw issues (GPU bottleneck) when there’re many of them. (you can use URP Depth Priming / “mobile” GPU early-Z to reduce opaque overdraw costs)
The reasons that sometimes a transparent material is performance cheaper can be:
Transparent materials usually don’t write depth to the camera’s depth buffer (not Depth Texture, it’s copied from camera depth at a time), so this makes them a bit less expensive compared to depth write enabled objects.
Transparent materials won’t execute any Prepass, so they can be less expensive and will be ignored by some post-processing effects.
I think you can try these common batching techniques:
GPU Instancing (requires the same mesh & material, not automatically enabled)
Static Batching (idea is to combine meshes into a big one and use 1 material only)
SRP Batcher (only for SRPs, reduces the cost of each draw call)
