NVIDIA Enhances Path Tracing in Indiana Jones™ with Shader Execution Reordering

NVIDIA has unveiled significant advancements in path tracing optimization for the Indiana Jones and the Great Circle™ game, utilizing Shader Execution Reordering (SER) to enhance performance on NVIDIA GeForce RTX 40 Series GPUs. This optimization is part of a broader series of enhancements aimed at improving the game's graphical fidelity and performance, as reported by NVIDIA's developer blog.

Enhancements with Shader Execution Reordering

The use of Shader Execution Reordering (SER) in the path tracing mode of Indiana Jones and the Great Circle™ has resulted in a 24% reduction in GPU time for the main path-tracing pass, TraceMain. This improvement is achieved by reordering threads in RayGen shaders, leading to more efficient execution and better utilization of active threads per warp.

In the profiled scene, captured on a GeForce RTX 5080 GPU, SER was able to increase the metric of active threads per warp from 38% to 70%, significantly enhancing the efficiency of the graphics processing unit. Such optimizations are crucial for handling complex scenes within the game, such as dense jungle environments with high graphical demand.

Optimization Techniques

To optimize SER, NVIDIA relied on the NVIDIA Nsight Graphics GPU Trace Profiler to identify and address inefficiencies. The RayGen shader's large number of ray-tracing live-state bytes were pinpointed as a bottleneck, leading to targeted reductions in variable sizes and the elimination of unnecessary GLSL variables.

Additional optimizations included converting certain floating-point variables to half-precision (FP16), which reduced the size of ray-tracing live-state data without compromising image quality. These changes helped decrease the GPU time from 4.08 ms to 3.08 ms, a notable improvement in rendering efficiency.

Impact and Future Developments

These advancements highlight the potential of SER to substantially accelerate graphics workloads with minimal code changes. By reducing the number of spilled bytes, NVIDIA has made SER an even more potent tool for developers looking to optimize performance in graphically intensive applications.

Looking forward, NVIDIA plans to explore further optimizations in path tracing, including the use of Opacity Micro-Maps and the compaction of dynamic BLASs to enhance performance in scenes with alpha-tested materials. These ongoing efforts underscore NVIDIA's commitment to pushing the boundaries of real-time graphics rendering.

Overall, the integration of Shader Execution Reordering in Indiana Jones and the Great Circle™ represents a significant step forward in the evolution of path tracing, offering players a more immersive and visually stunning gaming experience.