IT lexicon › Hardware › GPU particles

GPU particles

Hardware På svenska → Updated: 2026-10-26

Particle simulation that runs entirely on the graphics card, so millions of particles can update without burdening the CPU or crossing the bus every frame.

A traditional particle system updates each particle on the CPU and uploads the vertex buffer every frame — fine for thousands, a bottleneck at millions. Here position and velocity live in GPU buffers (or textures) and are advanced by a compute shader, so the data never leaves video memory. It scales to millions of sparks, smoke or crowd points.

The catch: reading results back to the CPU is slow, so collisions and gameplay-affecting logic are awkward; GPU particles are usually pure eye candy that never touch the simulation. The technique appeared with programmable shaders in the mid-2000s and became standard with compute shaders (DirectX 11, 2009). Unreal's Niagara and Unity's VFX Graph are built around them.

← Back to the lexicon