Show HN: DirectXSplat – D3D12 Gaussian splat renderer, up to 2.4× gsplat FPS

DirectXSplat, a new C++/Direct3D 12 library for rendering 3D Gaussian Splats, achieves up to 2.4× the FPS of gsplat in matched-quality benchmarks on an NVIDIA GeForce RTX 4070 SUPER, according to its developer. The open-source (MIT) library supports multiple scene formats, GPU-accelerated rasterization, and features such as GPU-side culling and compaction, with no CUDA dependency.

Hi all, Over the past few months, I've been working on and off on DirectXSplat, a C++/Direct3D 12 library for rendering 3D Gaussian Splats. It can load PLY, SPZ, .splat, SOG, and lod-meta.json scenes and render them from a host D3D12 application. Features: - GPU-accelerated D3D12 rasterization - Up to 2.4x the FPS of gsplat in matched-quality, resident-scene benchmarks on an NVIDIA GeForce RTX 4070 SUPER - Native embeddable D3D12 renderer with a host-owned device, queue, command list, fences, and render targets - Compact packed GPU scene buffers, persistent uploaded scenes, and reusable renderer resources - No CUDA dependency for DirectXSplat rendering - Support for trained 3DGS scenes in PLY, SPZ, .splat, and SOG formats, plus lod-meta.json scene manifests - Whole-scene updates and uploaded scene/chunk mutation - GPU resource interop for external work that references renderer-owned resources - Optional approximate splat-depth output - Convenience APIs including an interactive viewer through Show ... and offscreen image capture through Draw ... Some of the performance work includes an adapted GPU OneSweep radix sort, GPU-side culling and compaction, and indirect dispatch and draw arguments based on the surviving splats. Scene data is stored in compact packed GPU buffers. Uploaded scenes and renderer-owned resources persist and are reused between frames. Large scenes are internally partitioned into chunks and support hierarchy-based visibility, screen-space LOD selection, configurable splat and residency budgets, and residency caching. The project is MIT licensed and completely open source. Comments URL: https://news.ycombinator.com/item?id=49532043 https://news.ycombinator.com/item?id=49532043 Points: 1 Comments: 0