MeshNet: GPU-Driven Visibility Culling as a Network Transport Filter for Real-Time Multiplayer

We present MeshNet, a technique that repurposes GPU visibility culling data — already computed for rendering — as
a real-time network transport filter in multiplayer applications. By reading back the results of GPU-side frustum and
occlusion culling to the CPU, the network layer can selectively transmit only the state of entities determined to be
visible by the GPU. Preliminary measurements on a proof-of-concept implementation show that GPU culling reduces
the visible entity set to near zero when the player faces away from the scene, suggesting potential bandwidth
reductions of 8-15x in occluder-dense environments — though end-to-end network filtering has not yet been
validated. The technique eliminates an entire class of information-disclosure cheats (wallhacks) and imposes nearzero additional computational cost since the visibility data is a free byproduct of the rendering pipeline. We describe
the architecture, analyze its properties, discuss server-authoritative variants, present preliminary measurements, and
outline the conditions required for full validation.