rendertime

Rendertime is the amount of time required to generate a rendered image or frame from a 3D scene, from the start of the rendering process to the final pixel output. It is commonly measured in milliseconds per frame (ms) or as frames per second (FPS, where FPS = 1000 divided by rendertime).

Rendertime is distinct from latency, which is the time between a user action and its visible effect, and from the overall frame time, which includes compositing and display delays. Rendertime is primarily concerned with the computation needed to produce the image, while latency involves input-to-display delays and frame time encompasses the entire rendering-to-display pipeline.

Rendertime is influenced by scene complexity (geometry, shading, textures), lighting, effects, resolution, anti-aliasing, and post-processing, as well as hardware performance (GPU and CPU, memory bandwidth), driver optimizations, and the rendering pipeline (software renderers versus GPU-based renderers and APIs). When evaluating performance across hardware, rendertime is often estimated with benchmarks that simulate representative workloads.

Measuring methods include timing hooks around the render call, profiling tools, and standardized benchmarks. In real-time rendering, maintaining a target rendertime is essential for sustaining a desired frame rate; if rendertime exceeds the target, frame rate drops. In offline rendering, longer rendertime may be acceptable for higher image quality or more complex scenes.

Optimization strategies focus on reducing unnecessary work and improving data flow. Common approaches include lowering polygon counts or using level of detail (LOD), disabling nonessential post-processing, employing instancing, optimizing shaders, caching lighting, using texture compression, and leveraging hardware acceleration and parallelism. For large offline tasks, render farms or distributed rendering can reduce wall-clock time.