AMD released a paper on its new Dense Geometry Format (DGF) 3D graphics technology, and the hardware stack needed for it. The technology aims to significantly increase geometric detail in 3D graphics beyond even current raster methods, such as tessellation. It goes about doing this by minimizing the resource footprint of geometry. It packs small pieces of mesh, running up to 64 vertices and 64 triangles, into 128-byte blocks that form compact containers of geometry, positions of vertices, and light topology.

AMD's technology is similar in function although different in form to Mega Geometry introduced by NVIDIA with its "Blackwell" graphics architecture. Team Green defines Mega Geometry as squarely focussed on increasing the geometry of ray traced objects through a hierarchical model of storing geometry data. DGF has a broader application spanning raster and ray traced objects. The paper also deals with the hardware stack of DGF. Apparently, none of AMD's current architectures provide hardware acceleration for it, and the technology is being developed for a future architecture, such as "RDNA 5," or UDNA. On GPUs without hardware support, the DGF stack can be executed on shaders, however this comes with a small performance cost, which is worth the effort compared to rendering all that added geometry without DGF using conventional raster 3D methods.