Page tree

Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.


  • Substantial improvements made to Denoiser performance on the CPU, speed improvements of 2x to 4x are possible.
  • Default settings for PxrMarschnerHair have changed.
  • Improved sampling results when using mesh lights.
  • Improved default Shadow LPE with improved accuracy and now works with light filters. Previous behavior is preserved by using the 'shadows214;' LPE prefix, i.e. lpe:shadows214;CDL
  • PxrSurface now has more accurate behavior when thin shadows are enabled on glass-like surfaces; the improved behavior more accurately models the energy loss due to multiple reflection/refraction bounces. This change may cause some darkening of shadows compared to earlier versions.
  • The GGX specular option for PxrSurface should see reduced noise as well as a slight change at grazing angles on anisotropic materials.
  • Dramatic improvement in PxrMarschnerHair rendering speed, this might result in a slight brightness change when specularcompensation was active.
  • Updated the light selection algorithm. We expect, especially in situations with very many lights, that the new algorithm will demonstrate faster iterations times and/or improved convergence per iteration.
  • Max continuation length is now at least 256. Unless you have a significant number of overlapping volumes, you should not need to worry about this parameter.
  • Displace shaders now receive Synchronize messages.
  • Minor improvement to OSL texture access.
  • Multiscatter now supports multiple samples, just like single scattering does. Also, equiangularWeight is only controlling single scatter now. Multi-scatter ignores this.
  • We've added velocityMultiplier to PxrVolume, it could be helpful when converting velocity per frame to per second, etc.
  • The max resident memory for PxrVCM has been reduced by around 30%.

  • The time to first pixel for scenes with volumes undergoing deformation motion blur has been greatly reduced, especially when the motion blur is large and many threads are used.

  • Fixed a bug where Denoise would not take advantage of all possible hardware capability on Linux machines in either CPU or GPU mode.