Render farm efficiency: The strategic use of samples in 3D rendering

In 3D rendering, understanding the role of samples is fundamental. Samples are more than just a technical term; they are the backbone of how we translate a 3D model into a visually stunning image, whether it's for a high-end commercial project or a personal endeavor. 

This guide aims to demystify the concept of samples, examining their importance in both local rendering environments and on render farms. By dissecting what samples are, why they matter, and how to optimize their use, we'll provide insights that are essential for both beginners and seasoned professionals in the field of 3D graphics, especially when managing costs when using a render farm.

What are samples

In 3D rendering, "samples" refer to individual calculations of how light interacts with objects in a scene to produce the final image. Imagine a scene composed of various elements – objects, light sources, textures, and so on. When rendering this scene, the software needs to determine how light behaves as it hits these elements, which includes reflecting off surfaces, scattering, diffusing, and more. Each of these interactions contributes to the overall lighting and shading of the image, and a "sample" is essentially one calculation of these interactions for a tiny portion of the scene.

To break it down further:

• Ray Tracing and Path Tracing: Samples are crucial in rendering methods like ray tracing and path tracing. Ray tracing involves simulating the paths of light rays as they bounce off surfaces. Path tracing is a more advanced version, accounting for complex light paths, including indirect lighting and reflections. In both methods, each sample represents a path that a light ray might take from a light source, bouncing off objects, until it reaches the camera (or viewpoint).
  

• Building the Image: When rendering an image, thousands or even millions of these samples are calculated. Each sample contributes a small piece of information about the light and color at a specific point in the scene. The renderer accumulates these samples to build up a complete image. The more samples you have, the more accurate and detailed your image becomes.
• Noise and Clarity: A lower number of samples can result in an image with noticeable noise or graininess, as there are fewer calculations for light paths, leading to less accurate lighting and shading. Increasing the number of samples generally improves the image's clarity and realism but at the cost of longer rendering times.
  

• Balance in Usage: In a practical sense, the number of samples used in a render is a balancing act. More samples mean a clearer, more detailed image, but also longer render times and higher computational costs. This is especially crucial when using render farms, where costs can be directly tied to the amount of computational work required.
  

• Optimization Techniques: Many rendering software programs offer tools to optimize sample usage. For instance, adaptive sampling adjusts the number of samples in different areas of an image based on complexity or noise level, concentrating effort where it's needed most. Denoising techniques can also be used to reduce the appearance of noise in renders with fewer samples, making it possible to achieve high-quality results without excessively high sample counts.

Why care about samples when rendering on a render farm?

The leap from local rendering to a render farm is significant. Render farms, with their vast computational resources, can handle an enormous number of samples, drastically reducing render times. But why should this change your approach to samples?

First, consider cost-efficiency. Render farms often charge based on the computing resources used. More samples mean more processing time and, consequently, higher costs. Balancing the number of samples becomes crucial to manage expenses without compromising on quality.

• Render Samples: This is the number of paths traced for each pixel in the final render. As more samples are taken, the solution becomes less noisy and more accurate.

• Adaptive Sampling: With adaptive sampling, the render engine automatically reduces the number of samples in areas that have little noise, for faster rendering and more even noise distribution.

• Bounce Samples: In Cycles render for example, when enabling branched path tracing, one can specify the number of diffuse/glossy/transmission etcetera, as well as bounce samples for each Anti Aliasing sample. This allows you to give sampling priority to materials that take longer to converge.

• Denoising: Denoising removes noise while previewing scenes in Rendered mode in the 3D Viewport or for final renders1. In Blender for instance, there are different denoisers available, such as Intel’s Open Image Denoise, an AI denoiser that runs on the CPU, and OptiX, an artificial intelligence algorithm to remove noise from renders.

These settings allow you to control the balance between render quality and render time. By adjusting these settings, you can optimize your render times while still achieving the level of detail and realism you need in your final image.

Second, time constraints are pivotal. In professional settings, deadlines are non-negotiable. A render farm can churn through millions of samples rapidly, but misjudging the required sample count can lead to unnecessary delays.

Lastly, quality is key. Render farms enable the use of higher sample counts, enhancing the final image quality. However, overdoing it can lead to diminishing returns, where increases in sample count have minimal impact on image improvement.

How to know how many samples your project actually needs?

Determining the right number of samples is part art, part science. Start by considering the complexity of your scene. Scenes with intricate lighting, reflections, and textures typically require more samples to accurately capture these details.

Preview renders are your best friend here. Run test renders at lower sample rates and incrementally increase the count. Watch for when the improvements in image quality start to plateau - that's your sweet spot.

Software tools also come in handy. Many 3D rendering software like Blender, 3ds Max, Maya, and Cinema 4D have built-in sample calculators or noise reduction tools that can guide your decision.

When to go ham on samples and when to manage them?

Knowing when to push samples to their limits and when to rein them in is crucial. For still images where ultra-high quality is paramount, such as in architectural visualization or product design, maxing out samples can be justified.

Conversely, in animations or when working under tight deadlines, managing samples becomes essential. Opt for a balanced approach, focusing on optimizing other aspects like textures and lighting to compensate.

In both scenarios, always remember: more samples don't always mean better quality. Beyond a certain point, the improvement becomes imperceptible, yet the cost and time implications continue to grow. Always perform test renders locally when possible, and determine at what point increases in sampling do not make any noticeable improvement to the quality of your image.

Also keep in mind that certain cases may require more samples than others, such as SSS, Glossiness, Refraction or Volumetrics. If you have a scene that relies on these properties at certain points, be sure to test those accordingly.

Conclusion

Samples in 3D rendering are a powerful yet intricate tool. They are the building blocks of your digital imagery, dictating the quality, cost, and time efficiency of your projects. Whether you're rendering on a personal setup or using a render farm, a deep understanding and strategic use of samples can help you more effectively balance quality and speed, and get your projects rendered in time for your personal or professional deadlines.