Post
A multi-pass rendering approach that decouples geometry from lighting, allowing hundreds of dynamic lights without crippling performance.
Deferred rendering splits the rendering process into two main phases. First, all geometry is rendered into a set of screen-space buffers called the G-buffer, storing information like position, normals, color, and material properties at each pixel. Then, lighting calculations are performed as a separate pass using only the G-buffer data, meaning each light only affects the pixels it actually reaches. This decoupling is revolutionary for scenes with many lights because the lighting cost scales with the number of pixels affected, not the number of objects in the scene. The tradeoff is higher memory usage for the G-buffer and difficulty handling transparent objects and anti-aliasing.
Example
Grand Theft Auto V uses deferred rendering to handle its dense urban environments with thousands of streetlights, neon signs, headlights, and building lights simultaneously. Battlefield games rely on deferred rendering to handle the chaotic lighting of explosions, muzzle flashes, and vehicle lights across large maps. The S.T.A.L.K.E.R. series was an early adopter of deferred rendering, using it to create atmospheric lighting in the Zone that would have been impossible with forward rendering.
Why it matters
Deferred rendering made modern open-world games possible. Before it, every additional light source was extremely expensive because lighting had to be calculated for every object against every light. Deferred rendering broke that bottleneck and enabled the dense, light-rich environments we now take for granted in games like Cyberpunk 2077 and Spider-Man.
Related concepts