I've been working on a project with React Three Fiber and Three.js where I have multiple 3D houses made in Blender and exported as GLB files. Each house has quite a few objects, different materials, furniture and stuff like that. I'm trying to understand how people usually handle performance when a scene starts getting more complicated, especially on mobile.
I've been reading about reducing draw calls, reusing materials, limiting DPR and using frameloop="demand". But I'm still not entirely sure what makes the biggest difference in practice. For example, if you have several GLB models that users can switch between, is it generally better to keep them loaded in memory or dispose of everything when switching scenes?
I'm also curious how you figure out whether the main bottleneck is React, Three.js or just the models themselves. I know profiling is important, but I'm still learning which metrics to pay attention to.
Would appreciate some advice from people who've worked with larger R3F projects.…