Releases: NVIDIAGameWorks/NRDSample
Releases · NVIDIAGameWorks/NRDSample
v4.11.3
NRD updated to v4.11.3:
- PT: hooked up
NRD_MaterialFactors - PT: improved glass
- PT: SHARC usage for specular made lobe-footprint-based instead of "smc"-based
- UI: added "Glass" checkbox, transforming some cubes into glass!
- added
NRD_RESTORE_INITIAL_STATEswitch - added CPU-side performance markers
- better stated geometry flags
- NRI updated to v1.158
- updated deps
- updated tests
v4.11.1
v4.11.0
NRD updated to v4.11.0:
- PT: added 1-bit mask for glass and hair
- PT: fixed dark tails, left by moving emissive objects
- PT: ray offsetting tweaks
- UI: adjustments
- NRI updated to v1.154
- eliminated awkwardness around
materialID - dropped 3D MV
- improved TAA, also fixed potential ghosting and pixelation, while keeping IQ on glass and hair
- added
USE_CAMERA_ATTACHED_REFLECTION_TEST
v4.10.0
NRD updated to v4.10.0:
- PT: fixed curvature (scaling from matrices was applied incorrectly)
- PT: curvature computed on the fly, using minimal per-primitive data, coming from the CPU side
- PT: simplified curvature usage for hit distance estimation for NRD (removed some hacks)
- UI: added "boost" option, which is available if SHARC is ON (it's a good practice to shorten history by ~30% if SHARC is available)
- hooked new NRD integration features (FP16 promotion and FP32 demotion)
- NRI updated to v1.152
- code improvements
- updated tests
v4.9.3fixed
Same as v4.9.3 but includes a few simple fixes for some major regressions.
v4.9.3
NRD updated to v4.9.3:
- SHARC: lighting model used for cache population switched to "full" model with "true" specular
- SHARC: improved update to keep cache behind the camera alive
- SHARC: fixed energy conservation for Burley diffuse
- PT: fixed energy conservation for Burley diffuse
- PT: fixed glass behavior on top of a PSR-ed surface (fix after fix)
- PT: lighting at PSR decoupled into direct (shadows go through SIGMA) and indirect (goes through NRD)
- PT: indirect lighting at PSR is split into diffuse and specular according to diffuse probability in case of "1 path perf diffuse, 1 path per specular" mode (had wrong energy)
- PT: eliminated visible seams between L1 (prev frame) and L2 (SHARC) radiance caches
- PT: fixed darkening if prev frame reprojection with confidence < 1 ends the path
- PT: metal and non-metal material IDs get used in PSR like in normal frame
- PT: optimizations
- PT: fixed broken mips for indirect rays (were too conservative)
- fixed potential NANs in curvature
- NRI updated to v1.147
v4.9.1
NRD updated to v4.9.1:
- OCCLUSION: better settings
- OCCLUSION: added blue noise usage to compensate the lack of temporal stabilization
- UI: various improvements
- reworked post processing pipeline
- quality of life improvements
- fixed several minor bugs
- updated deps
v4.9.0
NRD updated to v4.9.0:
- PT: ambient estimation replaced with SHARC radiance cache
- PT: fixed NANs in hair lighting model
- PT: added translucency for foliage
- PT: hair material ID should not be used for PSRs
- PT: various hair rendering improvements (demodulation, animation, BLAS update)
- PT: significantly improved ray offsetting
- PT: better "normal guide" for hair
- PT: hooked up latest NRD features
- UI: various improvements
- DLSS: added dithering after DLSS
- improved TAA
- added ability to choose SIGMA variant via a macro switch
- fixed hair color space mismatch
- history reset related improvements
- lots of improvements and bug fixes (nano NRD usage opts, UI, accumulation factors...)
- reduced code entropy
- updated deps
- updated tests
- refactoring
v4.4.1
v4.4.0
NRD updated to v4.4.0:
- DLSS INTEGRATION: refactoring and improvements
- PT: VNDF sampling upgraded to v3
- TAA improvements
- NIS applied after DLSS, not only TAA
- returned accidentally lost USE_INF_STRESS_TEST
- improved sun animation
- updated tests
- updated dependencies