The primary advantage of deferred rendering is that the lighting is only computed for fragments that are visible (not blocked by others). Consider a scene with many lights:
In forward rendering, each fragment requires a loop over all light sources and for each, an evaluation of the reflectance model.
For many fragments, the lighting is computed, only to find that the fragment is not visible due to occlusion.
In this case it’s better to only computing lighting for visible fragments.
Deferred rendering provides better scalability as the number of light sources increases.
Cons:
The primary disadvantage of deferred rendering is the additional storage(memory overhead) for the g-buffer. Everything used in lighting must be stored there.
This becomes a difficulty when different materials are required for different objects in the scene.
</aside>
Forward Rendering v.s. Deferred Rendering
Forward Rendering
Deferred Rendering
Compare:
Pros
Cons
Forward Rendering
* computation expense
Deferred Rendering
* avoid unnecessary fragment shader runs (every pixel’s lighting calculation is ONLY done once)
support larger number of light sources | * extra memory needed for storing G-buffer
doesn’t support blending (since we only have info of the top-most fragment)