AMD unveils new ray tracing technique to reduce VRAM usage

AMD has introduced a new ray tracing research technique designed to significantly reduce the memory and computational overhead required for rendering complex, animated environments. The method, detailed in the research paper “Ray Tracing Massive Amounts of Animated Geometry,” utilizes tetrahedral cages to enclose dense geometry.

By animating a simplified deformable cage rather than the entire high-detail mesh, the technique allows for the real-time ray tracing of hundreds of millions of triangles without the massive memory requirements typically associated with updating Bounding Volume Hierarchy (BVH) structures. In a technical demonstration using a Radeon RX 9070 XT, AMD showcased a scene containing approximately 25,000 independently animated foliage items, totaling roughly 500 million ray-traced triangles per frame.

The results demonstrated a massive reduction in memory usage for acceleration structures, dropping from as much as 80 GB to approximately 1.7 GB. Additionally, the update time for these structures was reduced to 3.3 ms, compared to over 300 ms using conventional methods. The demonstration achieved over 60 FPS at 1080p resolution. AMD is currently developing DirectX Raytracing samples and a C++ library to facilitate wider industry integration.