A render farm is a computer cluster to render computer generated imagery (CGI), typically for film and television visual effects. The rendering of images is a highly parallelizable activity, as each frame can be calculated independently of the others, with the main communication between processors being the upload of the initial source material, such as models and textures, and the download of the finished images.
As a rule of thumb, CG images take roughly an hour per frame to render. The amount of time it takes to render an image has remained roughly constant over two decades, in spite of huge improvements in computer processing power. The reason for this is that advances in computer power are absorbed by increased computation in order to meet demands to achieve state-of-the-art image quality, while the hour-per-frame figure arises from the demands of production timescales.
To manage large farms, one must introduce a queue manager that will facilitate automatic distribution of processes to the many processors. The software is typically a client-server package that facilitates communication between the processors and the queue manager, although some queues have no central manager. Some common features of a queue managers are: re-prioritization of the queue, management of software licenses, and algorithms to best optimize throughput based on various types of hardware in the farm.
The use of render farms in the entertainment industry can be viewed as one early application of grid computing.