AMD has released research describing a tetrahedral cage approach that lowers memory and compute costs for rendering animated scenes with ray tracing. Holger Gruen, a researcher at AMD, explained that the technique encloses dense geometry within deformable cages, allowing real‑time ray tracing of hundreds of millions of animated triangles while dramatically reducing memory overhead. In a demonstration reported by Wccftech, the method was tested on a Radeon RX 9070 XT, where a terrain scene containing 25,000 foliage elements — approximately 500 million ray‑traced triangles per frame after level‑of‑detail filtering — maintained a frame rate above 60 FPS at 1080p. Updating acceleration structures took 3.3 ms, a stark contrast to the 300 ms required by conventional bottom‑level acceleration structures. Memory consumption for those structures fell from as much as 80 GB to roughly 1.7 GB. Traditional ray tracing typically rebuilds or updates BVH structures for each independently animated object, but AMD’s solution keeps the underlying dense mesh and its BVH static, animating only the outer tetrahedral cage and transforming rays back into the original reference frame during intersection tests. The approach has limitations; tetrahedral cages provide less precise vertex control, making them more appropriate for dense foliage, grass, crowds and distant characters than for highly detailed mesh deformation. The paper, titled “Ray Tracing Massive Amounts of Animated Geometry,” earned third place at the High‑Performance Graphics 2026 conference. AMD noted that DirectX Raytracing samples and a C++ library are currently being developed to facilitate broader adoption.