Advances in path tracing techniques have allowed for the capture of lighting data from the environment, enabling the use of indirect illumination in real-time with both improved accuracy and speed. RTXGI SDK implements two such techniques, replacing traditional probe-based irradiance caching with a world-space radiance cache which can be used to sample outgoing radiance each time scene geometry is hit during path tracing.
These techniques may be combined with a regular path tracing pipeline for the primary rays, sampling cached data only for indirect bounce evaluation. By replacing the whole path trace with a single ray hit evaluation and cache lookup, the cost is reduced with little to no compromise in signal quality, while remaining responsive to change and supporting large-scale dynamic scenes with complex lighting setups.
RTXGI SDK provides an example integration (DX12 and Vulkan) of two state-of-the-art radiance caching techniques for path tracing - a (currently experimental) AI-based approach known as Neural Radiance Cache (NRC), and Spatially Hashed Radiance Cache (SHaRC). The former requires Tensor Cores while the latter has certain limitations but is currently supported on a wider range of hardware without any vendor-specific requirements. RTXGI SDK also hosts documentation and distribution corresponding to both of these techniques, see Project Structure section for further details.
| Directory | Details |
|---|---|
| /docs | Documentation for showcased tech |
| /donut | Framework used for the samples |
| /external | Helper dependencies for the samples |
| /media | Assets and scene definitions |
| /samples | Samples showcasing usage of NRC, SHaRC |
| /sdk-libraries | Binaries, src, includes for NRC, SHaRC |
Any DXR GPU for SHaRC | NV GPUs ≥ Turing (arch 70) for NRC | CMake v3.24.3 | Git LFS | Vulkan SDK 1.3.268.0 | VS 2022 | Windows SDK ≥ 10.0.20348.0 | Driver ≥ 555.85
- Quick start guide for building and running the pathtracer example.
- NRC integration guide and the SHaRC integration guide respectively.
- Change log for release information.
RTXGI SDK is actively being developed. Please report any issues directly through the GitHub issue tracker, and for any information or suggestions contact us at [email protected].
Use the following BibTex entry to cite the usage of RTXGI in published research:
@online{RTXGI,
title = {{{NVIDIA}}\textregistered{} {RTXGI}},
author = {{NVIDIA}},
year = 2024,
url = {https://github.com/NVIDIAGameWorks/RTXGI},
urldate = {2024-03-18},
}See LICENSE.md
Version v1.x of RTXGI which includes the DDGI algorithm is located in a separate repo at the following url: https://github.com/NVIDIAGameWorks/RTXGI-DDGI".