AMD unveils Tetrahedral Cages technology to optimize ray tracing

AMD has unveiled a new ray tracing technology called ‘Tetrahedral Cages’ designed to significantly reduce the resource demands of large-scale dynamic scenes. The technology aims to lower the overhead associated with Bounding Volume Hierarchy (BVH) updates and memory usage when rendering complex environments like dense vegetation or large crowds.

In a demonstration involving a scene with approximately 25,000 independently animated plants and 500 million dynamic triangles, AMD reported that traditional methods required up to 80GB of GPU memory and over 300ms for BVH updates per frame. By implementing Tetrahedral Cages, memory requirements dropped to approximately 1.7GB—about 1/47th of the original amount—and update times were reduced to roughly 3.3ms.

The method works by wrapping high-density geometry in a smaller, deformable tetrahedral cage. Instead of animating billions of individual triangles, the system only updates the simpler cage, while the complex geometry remains static and reusable. Rays are then transformed back to the static reference space for intersection calculations. While this reduces per-vertex animation control, it makes the computational cost scale with the complexity of the cage rather than the number of triangles. The technology is compatible with Microsoft DirectX Raytracing (DXR) and is currently in the research and development stage.