Sparse voxel octree ray-guided GPU streaming @Alexander_Sannikov
Sparse voxel octree ray-guided GPU streaming  @Alexander_Sannikov
Uploaded September 2021 | Updated September 2026, 1 week ago
This is a demo of my sparse voxel octree GPU streaming/rendering engine. I used a fractal for this demonstration, but the engine handles arbitrary voxel data. Octree nodes are streamed in and out of GPU memory when the rasterizer asks for them, so the algorithm always tries to keep as few of them in GPU memory as possible.

Here's some stats from this demo:

Effective resolution of voxels : 4k^3, so 64,000,000,000 voxels max
Total number of octree nodes in the static storage: 137M
Total storage space used by the static storage: 5.4GB
Maximum number of octree nodes resident on the GPU: 600K
GPU memory usage: 23MB
Target octree node rasterized size: 1px (unless changed by the slider)
Sparse voxel octree ray-guided GPU streamingWalker 2 fall recoverySparse voxel signed distance octree ray-guided GPU streamingExploring 3d holographic radiance cascadesunarmedRadiance Cascades Rendered DirectlyTome4 Fire Surge shader v2Glittergram demo #1: loading models (sort of)Crawling pattern generationElastic SPH collision testSound wave simulationShadow maps technique
Alexander Sannikov |

Sparse voxel octree ray-guided GPU streaming

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER