I should mention first that the lighting thing wasn’t an issue; I was stating that I have chosen to only use shadows for the character meshes.
Here’s what I can think of that you could use to repro our game’s processing setup:
- About 100 draw calls at any time. Textures for most of them are very tiny, at 64x64. Larger ones are around 256x256~512x512. Characters use a single texture atlas at 1024x1024, with normals and opacity.
- Including the player, about 5 complex skeletal meshes at any time.
- ALL materials have color map going in to base color, and also (color map)x(.1) going into emissive. It gives a some pep to the cartoon style.
- There are 2 large meshes that are batched. One is around 30K polys and the other around 15K polys. There are dozens more that sit around 2K that are also batched. Characters are about 7K each.
- One mesh has (color map)x(10) for emissive, since it’s the batched mesh that represents the ceiling lights.
- Bloom is on.
- Single dynamic light for casting dynamic shadows from characters.
- Several static lights that have settings to not cast/bake shadows.
- Shadows are turned off for everything except the characters.
- HDR is on.
- Devices tested on are typically: Galaxy Note 3, Nexus 7, iPhone 4S, iPhone 5S, iPhone 6, Galaxy S5, Galaxy S4.
- Galaxy Note 3 is primary testing device where we’re aiming for min 20~30fps, with iPhone 4S being the furthest back phone that we want the game to work on, but are not too worried about fps on it. It’s an old phone, so if it’s getting ~10fps, we’re fine. Newer iPhones virtually never have performance issues. They consistently run 30+ fps.
Maybe this would be easier if you could test the actual project itself? I’m not sure how that would even work, but it’d probably be a lot easier.
I’m definitely encouraged by you saying you didn’t see an fps change. I’m really hoping I’m just missing some setting somewhere and can move into 4.7 asap.
Really appreciate your time.