Archive notice: This article was originally published on May 22, 2019. Links and embedded videos are preserved as part of the historical record.
Originally planned for April, the developers have finally released the first Redshift 3.0 release for existing users. Progress is said to have been somewhat delayed by the recent acquisition by Maxon. The latest Nvidia drivers are required for the experimental Redshift 3.0.0 build. Here we provide an overview of the current and outstanding 3.0 features:
Current Redshift 3.0 Features
RTX (RT Core) acceleration: There is a new checkbox for enabling Optix RT acceleration for all RT-capable GPUs. All functions of the (old) CUDA-based ray tracer are now supported, apart from trace sets and light/mesh linking. With the new hardware acceleration, rendering times up to 40% faster are possible, depending on the scene (!). You can find a realistic assessment here.
Increased sampling depth: As previously announced, Redshift now supports a sampling depth of up to 64 diffuse, reflection and refraction bounces, as well as 256 transparency levels. This now makes it possible to achieve significantly more precise results. RS 3.0.0 also contains several sampling optimizations to further accelerate render times. (Automatic sampling is not yet enabled).
Improved AOVs: The higher sampling depths fix various artifacts and errors in lighting-related AOVs (e.g. specular or diffuse).
Outstanding Features
The following functions are intended to follow shortly:
- Automatic sampling for a significant performance boost (up to 4x)
- NVLINK will be supported as soon as RTX features are stable.
- DeepEXR handling with large datasets
- Cinema 4D Noise – native support (without baking) in all DCC tools
- You can find further information here in Redshift’s forum post
Availability
Redshift 3.0 versions require NVidia driver version 419.67 or later (Windows), 418.56 or later (Linux) and 387.128 or later (macOS). As a registered user, you can download the experimental build here.
