106.2

Release Highlights

In this release, geometry streaming is now supported, which optimizes the use of GPU memory by dynamically loading the most relevant geometry relative to the view and available memory.

Auto Generation of LODs (level-of-detail) has also been added as an option, which enables generating, caching, and loading lower resolution representations of the geometry when necessary to help fit within the memory constraints, which can also improve performance.

Additionally, while the RTX renderer is able to render millions of instances, a high instance count can reduce performance. The streaming system allows reducing the maximum instance count while selecting the most relevant subset of loaded instances to render.

Both texture and geometry streaming are balanced against each other to maintain an optimal distribution of memory for high-quality rendering in real-time, even with large datasets.

See the GPU Resources Management section for more details.

Added

  • RTX: Added Geometry Streaming feature. This feature allows balancing GPU memory usage for geometry separately from textures, automatic LOD generation and streaming, and instance streaming.

  • RTX: Added mesh attribute to disable LOD generation per mesh.

  • RTX Real-Time: Added a bool setting to skip reflection for translucency rays when hits are close to the camera, which can improve performance. /rtx/translucency/skipReflectionForCloseToCameraHits

  • RTX Real-Time: Added a Translucency setting ‘Invisible Light Behind Translucency in Reflections’. If enabled, invisible lights behind translucent geometry won’t be visible in the reflection. Depends on the roughness thresholds of the light visibility settings.

Improved

  • RTX: Removed deprecated ‘Max Fisheye Resolution’ and ‘Front Face Resolution Scale’ from ‘Fisheye Lens’ parameters.

  • RTX: Support for the following camera Projection Types are now deprecated: fisheyeEquidistant, fisheyeEquisolid, fisheyeOrthographic.

  • RTX: Optimized instance creation time for geometry that has more than one instance.

  • RTX Real-Time: MDL Material Distilling is no longer optional, it is always enabled, and the setting has been removed from the Preferences.

Fixed

  • RTX: Improved CPU-based Vertex Displacement performance by preventing unnecessary mesh geometry computation during displacement editing.

  • RTX Interactive: Fixed an issue where the alpha would appear aliased if the Bloom post-process was enabled.

Known issues

  • RTX Interactive: UI responsiveness can lag when rendering complex volumes.