Data Parallel Path Tracing in Object Space

TL;DR

This paper presents a novel data-distributed path tracing method that enables collaborative rendering across multiple nodes or GPUs, supporting various partitioning strategies and reducing network communication for efficient production-style rendering.

Contribution

Introduces a new approach to distributed path tracing that improves over traditional spatial partitioning, supporting object-space and spatial partitioning, and reducing network communication.

Findings

Supports different model partitioning strategies

Enables rendering of complex models on moderate hardware

Reduces network traffic during distributed rendering

Abstract

We investigate the concept of rendering production-style content with full path tracing in a data-distributed fashion -- that is, with multiple collaborating nodes and/or GPUs that each store only part of the model. In particular, we propose a new approach to tracing rays across different nodes/GPUs that improves over traditional spatial partitioning, can support both object-space and spatial partitioning (or any combination thereof), and that enables multiple techniques for reducing the number of rays sent across the network. We show that this approach can handle different kinds of model partitioning strategies, and can ultimately render non-trivial models with full path tracing even on quite moderate hardware resources with rather low-end interconnect.