Post
A memory trick that lets games display impossibly large textures by only loading the pixels the camera can actually see.
Virtual texturing (also called sparse virtual textures or partially resident textures) treats the GPU's texture memory like a virtual memory system. Instead of loading entire textures into VRAM, only the specific tiles (small rectangular sections) that are currently visible at the needed resolution are loaded. A massive 32K terrain texture might have millions of tiles, but the player only ever sees a few hundred at any given camera position. The system uses a feedback pass to determine which tiles are needed, then streams them from disk or generates them on-the-fly. id Software pioneered this with Mega Textures in RAGE (2011), and modern implementations in UE5 (Virtual Texture Streaming) are far more sophisticated.
Example
RAGE (2011) used id Tech 5's MegaTexture system to give every surface in the game a unique texture, eliminating the repetitive tiling visible in most open-world games. While the implementation had pop-in issues, it demonstrated the principle. UE5's Nanite and Virtual Texturing work together to stream both geometry and textures, allowing film-quality assets to be used directly in games without manual LOD creation.
Why it matters
Virtual texturing breaks the link between world size and memory requirements. Without it, massive open worlds must reuse textures constantly, creating visible repetition. With it, every surface can be unique. It's a foundational technology for next-gen open worlds and explains how modern games maintain visual density at massive scale.
Related concepts