IT lexicon › Hardware › Virtual texturing

Virtual texturing

Hardware På svenska → Updated: 2026-10-24

A huge texture surface is split into pages, and only the ones actually visible are streamed into a small physical cache.

The idea is that a texture can be far larger than video memory if you treat it like virtual memory: carve it into tiles ("pages"), keep only the visible ones in a physical cache, and look up where each landed through an indirection table. id Software's id Tech 5 marketed this as MegaTexture in Rage (2011); Sean Barrett's "sparse virtual textures" described the same principle openly.

The payoff is that the texture budget is decoupled from the total amount of texture — unique texels across an entire world become feasible. The cost is streaming complexity and visible pop-in when the cache misses.

← Back to the lexicon