Octane Standalone Documentation

Adaptive Sampling is a method of sampling that determines if areas of a rendering require more sampling than other areas instead of sampling the entire rendering equally. In Octane, Adaptive Sampling disables sampling for pixels that have reached a specified noise level. With adaptive sampling, Octane is able to stop rendering on areas which no longer need to be rendered thus, freeing more GPU power to render pixels that still need to be rendered. This allows for bumping up the maximum samples quite high (even more than 30,000) and then rely on the adaptive sampling to figure out which pixels actually need that many samples and which do not.

Figure 3: Accessing Adaptive Sampling for the Direct Lighting kernel

Adaptive Sampling Parameters

Adaptive Sampling - Enables adaptive sampling.

Noise Threshold - Specifies the smallest relative noise level. When the noise estimate of a pixel is less than this value, sampling switches off for this pixel. Good values are in the range of 0.01 - 0.03. The default is 0.02, which produces a fairly clean render.

Min. Adaptive Samples - Specifies the minimum number of samples to calculate before adaptive sampling kicks in. A pixel's noise estimate has a large initial error. The higher you set the noise threshold, the higher you should also set this parameter to avoid artifacts.

Pixel Grouping - Specifies the number of pixels handled together. When all of the pixels in a group reach the noise level, sampling stops for all of these pixels.

Expected Exposure - This value should be close to the same value as the image's exposure, or 0 (the default value) to ignore these settings. Adaptive sampling uses this parameter to determine what pixels are bright and dark, which depends on the Octane Imager's exposure setting. If the value is not 0, adaptive sampling adjusts the noise estimate of the image's very dark areas. It also increases the Min. Adaptive Samples limit for very dark areas, because very dark areas tend to find irregular paths to light sources, resulting in over-optimistic noise estimates.

To visualize the progress, a noise pass AOV can be enabled. The pass is calculated when Adaptive Sampling is enabled. The green pixels in that pass mark those pixels that have reached the specified noise limit. This mask is re-calculated every time a new result is blended into the film buffer.

Figure 4: Noise AOV visualization