Microsoft has finally released a second DirectX Ray Tracing (DXR) functional specification file that outlines what its ray tracing pipeline is expected to look like, the goals the company is pursuing, and what the technology does behind the scenes. In the original file, Microsoft described the ray tracing pipeline from ray shader generation, scheduling, and acceleration structure, all the way to the shading of the actual game. This time, the company has shared insights into areas such as clustered geometry, partitioned top-level acceleration structures (TLAS), and indirect acceleration structure operations.

Firstly, Microsoft introduces the concept of clustered geometry. Readers need to understand that the core graphics elements are triangles, which are the building blocks of the 3D worlds we have today. However, DXR clustered geometry treats groups of nearby triangles as common building blocks or multiple building blocks, allowing the GPU to build, move, and instantiate geometry in bulk. Instead of handling this separately with multiple calls for triangles, the GPU's task is now much more simplified. DXR even defines compact vertex encodings and predefined template formats to ensure that the GPU memory and bandwidth required to execute bulk geometry building and moving are sufficient. As a result, the GPU doesn't have to update or duplicate existing geometry, and DXR will help render foliage, crowds, and in-game props once, allowing them to be moved around easily. This reduces the GPU load and improves the performance of ray tracing in games.