Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GRTX: Efficient Ray Tracing for 3D Gaussian-Based Rendering

Junseo Lee, Sangyun Jeon, Jungi Lee, Junyong Park, Jaewoong Sim

TL;DR

Gaussian ray tracing of 3D Gaussian Splatting suffers from bloated BVHs and redundant multi-round traversals. GRTX addresses this with a two-level TLAS/BLAS structure where all Gaussians reference a single shared BLAS of unit spheres, combined with hardware traversal checkpointing to resume from checkpointed nodes across rounds. The paper introduces GRTX-SW (software-only) and GRTX-HW (hardware-augmented) and demonstrates end-to-end speedups up to 4.36× over a baseline, with modest hardware overhead (~1.05 KB of per-RT-core storage) and strong cache locality benefits. These contributions enable efficient Gaussian ray tracing across vendors and support dynamic scenes via multi-level instancing, bringing high-quality Gaussian-based rendering closer to real-time performance.

Abstract

3D Gaussian Splatting has gained widespread adoption across diverse applications due to its exceptional rendering performance and visual quality. While most existing methods rely on rasterization to render Gaussians, recent research has started investigating ray tracing approaches to overcome the fundamental limitations inherent in rasterization. However, current Gaussian ray tracing methods suffer from inefficiencies such as bloated acceleration structures and redundant node traversals, which greatly degrade ray tracing performance. In this work, we present GRTX, a set of software and hardware optimizations that enable efficient ray tracing for 3D Gaussian-based rendering. First, we introduce a novel approach for constructing streamlined acceleration structures for Gaussian primitives. Our key insight is that anisotropic Gaussians can be treated as unit spheres through ray space transformations, which substantially reduces BVH size and traversal overhead. Second, we propose dedicated hardware support for traversal checkpointing within ray tracing units. This eliminates redundant node visits during multi-round tracing by resuming traversal from checkpointed nodes rather than restarting from the root node in each subsequent round. Our evaluation shows that GRTX significantly improves ray tracing performance compared to the baseline ray tracing method with a negligible hardware cost.

GRTX: Efficient Ray Tracing for 3D Gaussian-Based Rendering

TL;DR

Gaussian ray tracing of 3D Gaussian Splatting suffers from bloated BVHs and redundant multi-round traversals. GRTX addresses this with a two-level TLAS/BLAS structure where all Gaussians reference a single shared BLAS of unit spheres, combined with hardware traversal checkpointing to resume from checkpointed nodes across rounds. The paper introduces GRTX-SW (software-only) and GRTX-HW (hardware-augmented) and demonstrates end-to-end speedups up to 4.36× over a baseline, with modest hardware overhead (~1.05 KB of per-RT-core storage) and strong cache locality benefits. These contributions enable efficient Gaussian ray tracing across vendors and support dynamic scenes via multi-level instancing, bringing high-quality Gaussian-based rendering closer to real-time performance.

Abstract

3D Gaussian Splatting has gained widespread adoption across diverse applications due to its exceptional rendering performance and visual quality. While most existing methods rely on rasterization to render Gaussians, recent research has started investigating ray tracing approaches to overcome the fundamental limitations inherent in rasterization. However, current Gaussian ray tracing methods suffer from inefficiencies such as bloated acceleration structures and redundant node traversals, which greatly degrade ray tracing performance. In this work, we present GRTX, a set of software and hardware optimizations that enable efficient ray tracing for 3D Gaussian-based rendering. First, we introduce a novel approach for constructing streamlined acceleration structures for Gaussian primitives. Our key insight is that anisotropic Gaussians can be treated as unit spheres through ray space transformations, which substantially reduces BVH size and traversal overhead. Second, we propose dedicated hardware support for traversal checkpointing within ray tracing units. This eliminates redundant node visits during multi-round tracing by resuming traversal from checkpointed nodes rather than restarting from the root node in each subsequent round. Our evaluation shows that GRTX significantly improves ray tracing performance compared to the baseline ray tracing method with a negligible hardware cost.
Paper Structure (20 sections, 2 equations, 24 figures, 3 tables)

This paper contains 20 sections, 2 equations, 24 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Ray Tracing
  4. 3D Gaussian-Based Rendering
  5. Motivation
  6. 3D Gaussian Ray Tracing
  7. Performance Analysis
  8. Observations and Opportunities
  9. GRTX: Gaussian Ray Tracing Acceleration
  10. GRTX-SW: Leveraging Two-Level Acceleration Structure for Gaussian Primitives
  11. GRTX-HW: HW Acceleration for Gaussian Ray Tracing
  12. Evaluation
  13. Methodology
  14. Performance
  15. Source of Performance Gain
  16. ...and 5 more sections

Figures (24)

  • Figure 1: 2D visualization of Bounding Volume Hierarchy (BVH) and ray traversal.
  • Figure 2: Ray tracing pipeline.
  • Figure 3: Overview of 3D Gaussian Ray Tracing moe:mir24.
  • Figure 4: (a) Rendering performance of rasterization (3DGS) and ray tracing (3DGRT). (b) Execution time for a single round of tracing while isolating each operation in 3DGRT.
  • Figure 5: (a) Rendering time and (b) BVH size when using triangles or custom primitives.
  • ...and 19 more figures