AMD is previewing a smarter way for GPUs to handle modern animation and ray tracing, and it could have a big impact on future graphics performance. The company’s Dense Geometry Format (DGF) is a new, GPU-friendly approach to storing and processing geometry that promises higher efficiency, faster ray tracing, and smoother animation—potentially with hardware-level acceleration in upcoming RDNA GPUs.
At its core, DGF is a compressed container for geometry designed specifically for GPU workflows. Instead of feeding massive triangle meshes straight to the pipeline, DGF breaks them into small “meshlet” blocks and stores each block in a dense local format. When those blocks need to be animated, the GPU doesn’t have to unpack and rebuild the whole thing. A per-frame compute shader can re-quantize and update only what’s needed, keeping the data compressed and the processing lean.
This approach is especially interesting for ray tracing. One of RT’s costliest steps is rebuilding BVHs (Bounding Volume Hierarchies) as scenes move and deform. Because the GPU can understand and operate directly on DGF blocks, it can build and update RT acceleration structures more efficiently, cutting memory bandwidth demands and reducing overhead in the ray tracing pipeline. The result is more geometry fitting inside GPU caches, lower latency, and higher, more stable performance.
Today, DGF runs on compute shader units. Looking ahead, AMD hints that future RDNA GPUs could push parts of this workflow into fixed-function hardware, accelerating animation and RT updates even further. Offloading DGF operations from general compute to dedicated units would free up resources and improve per-frame throughput in complex scenes.
What this means in practice:
– More efficient animation updates without unpacking entire meshes
– Faster, leaner BVH builds for ray-traced scenes
– Lower memory bandwidth usage and better GPU cache utilization
– Smoother frame times in geometry-heavy or RT-heavy games and 3D applications
DGF isn’t a magic bullet on its own, but it’s a meaningful piece of the next-gen rendering puzzle. By compressing geometry in a way the GPU can directly leverage, it enables denser worlds, richer animations, and more scalable ray tracing without demanding a proportional increase in raw compute or memory bandwidth.
For developers, DGF offers a clear path to packing more detail into real-time scenes without blowing up performance budgets. For players, it points to future GPUs delivering better ray tracing and animation at the same or lower power and bandwidth costs. Keep an eye on upcoming RDNA generations as hardware-level DGF support matures and moves from compute-driven techniques to dedicated acceleration.
