Performance Settings

The General settings tab includes the render settings that affect the appearance of the rendered image.

UI Path: ||Render settings window|| > Corona > Performance settings tab

Lock sampling pattern – When enabled, the noise is 'locked' in place, both in still frames and animations, making it less visible. Disabling this option produces a completely different noise pattern for each frame, which is required by some third-party de-noising applications (but not by the Corona denoiser). When rendering using Distributed Rendering or a render farm service, the noise might be different from frame to frame regardless of this option.

Conserve memory (slower) – When enabled, Corona consumes less memory at the cost of rendering performance. This option can decrease peak memory usage by up to 30%. Consider turning this on if you are running out of memory during rendering.

Adaptive light solver – Toggles an improved method of rendering lights. This algorithm can speed up your renders by learning which scene lights are important at a particular location. Those lights then receive more rays than the others, resulting in a lower noise level. Enabling this option may significantly reduce render times, especially in scenes:

Automated out–of–core textures –  When enabled, Corona automatically offloads unused parts of textures to HDD to consume less memory. Additional settings are located in the Corona Preferences window.

GI/AA balance – This is the number of GI samples per each anti-aliasing sample. Higher values make global illumination and light noise clean up faster, while lower values make the antialiasing, motion blur, and depth of field noise clean up faster. Setting this value under 2 or above 64 is not recommended.

Light samples multiplier – This is the number of direct lighting samples per each GI sample. Increase this value in scenes with visible noise in direct lighting. The default value of 2.0 works well in most cases.

Max sample intensity – This is the maximum brightness of the secondary GI samples. This value is a tradeoff between rendering performance and physical accuracy.
Lower values suppress noise but produce a darker image with lower intensity of reflections and caustics. Higher values produce brighter reflections and caustics but also more noise in the image.
The default value of 20 works well in most scenes. Value 0 is a special case that enables an unbiased mode (this mode is not production-ready as it will never clear noise in some scenes).

Max ray depth – This is the maximum number of light bounces. Lowering this value slightly improves performance; increasing it slightly improves accuracy. However, the change in Corona is very small compared to other renderers, so keeping the default value is advised.

Screen size (px) – When selected, displacement tessellation is performed adaptively in the screen space. This is usually a lot more efficient. Size sets the number of pixels each tessellated triangle spans in the image. Lower values improve displacement quality at the expense of memory usage and preprocessing time.

World size – When selected, displacement tessellation is performed absolutely in the world space. This method is usually inefficient except for special cases (to prevent popping artifacts in animations). Size sets the maximum length of each tessellated triangle in world units. Lower values improve displacement quality at the expense of memory usage and render preprocessing time. Beware of setting this value too low and using this method on large objects, as this may consume all the available memory and result in a crash.

Max passes – Limits the maximum number of passes rendered in the interactive rendering mode. Note that for interactive render regions, this value is best left set to 0 to ensure that new or changed regions will render (otherwise, they will not render if the max passes have already been reached).

Force path tracing – When on, Path Tracing is always used as a primary and secondary solver in interactive rendering, even when UHD Cache is selected for regular rendering. This keeps the interactive rendering responsive, avoiding the cost of UHD Cache precomputation.

Enable motion blur – When enabled, this option allows the motion blur to be computed in the interactive rendering.

Fast preview denoise during render – When set to an option other than None, the image is iteratively denoised during the render. This replaces a noisy render preview with a noiseless, albeit low-quality image that gets progressively refined as the render converges. It can help to get a quick estimate of the overall lighting in the scene. There are two denoising modes available:

Enable – Generates physically correct caustics at the cost of lower rendering performance. Note that to render refractive caustics, you need to enable them in the Corona Material (either if it's a legacy or physical material).

Caustics only in caustics pass – When enabled, caustics are rendered only in a special Caustics multi-pass. When disabled, caustics also appear in beauty, light select, and all other applicable multi-pass elements.

Generate caustics from environment – When enabled, the caustics solver generates caustics from the scene environment lighting. This option does not influence caustics generation by the sun.

Enable caustics adaptivity – When enabled, this option uses an improved method for caustics calculation.

Include/Exclude – These options allow you to include or exclude specific objects from being affected by caustics.

Available Include/exclude modes:

Apply caustics only on these objects – Using this mode, caustics are only visible in the defined objects from the list.
Don't apply caustics on these objects – Using this mode, caustics are not visible in the defined objects from the list.

Enable – Enables the DOF Highlights Solver that improves the speed of rendering of out-of-focus objects (at the cost of casting more rays).

Budget (%) – Increasing this value improves the rendering of out-of-focus objects, but it leads to slower rendering of all other effects.

Reprojection count – For debugging purposes only. When a highlight blurred by depth of field is found, it is projected back to the image. This number changes how many times the highlight is projected back.