Skip to content

RTXGI v2.0 Update including Neural Radiance Cache and Spatial Hash Radiance Cache

License

Notifications You must be signed in to change notification settings

NVIDIAGameWorks/RTXGI

Repository files navigation

RTXGI

banner


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.

Project structure

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

Getting up and running

Prerequisites

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

Further steps

Contact

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].

Citation

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},
}

License

See LICENSE.md

RTXGI v1.x

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".