Sponsored Feature: Experiments In Cloud-Based Ray Tracing

Author: by Staff

Rendering in a server cloud is fast becoming part of the video game industry's business model, but what implications could cloud technology have for development? In Gamasutra's Intel-sponsored feature today, researcher and ray-tracing specialist Daniel Pohl explores how leveraging cloud-based ray tracing can bring consumers better graphics. "Ray tracing is a rendering technique that uses the laws of physics to create more photorealistic graphics by accurately calculating effects like reflections, refractions, and shadows," Pohl explains. "However, the computational requirements of ray tracing limit its use for interactive consumer applications like games." "The cloud gaming model provides one method to give more people access to the high-end hardware needed to perform the ray tracing in real-time," suggests Pohl. In the in-depth feature, Pohl demonstrates the combination of a cloud-based gaming approach and utilizing ray tracing for rendering using Intel MIC Architecture, using a small laptop that represents an average gamer's hardware and the German version of 2009's Wolfenstein. "For this initial experiment we chose to start with the best-case networking scenario, a high-bandwidth Gigabit Ethernet connection, to relax the compression requirements and achieve the best visual results," he explains. "That allowed us to use a different form of compression that is based on each image separately (compared to a video stream where a later image is often a modification of an earlier image), which results in very high image quality that remains stable independent of the amount of motion." When it comes to the graphical rendering itself, Pohl explains why ray tracing offers more detailed geometry: It uses acceleration structures that organize scenes to more efficiently calculate the way rays interact with objects, which means that when it comes to static content, the amount of calculations required increases only in a logarithm with the number of triangles -- hence higher detail with lower performance impact. For example, says Pohl: "The complete Wolfenstein game level used for the demonstration consisted about around 300,000 triangles in the original form. To showcase the ability to have more geometry available, two objects were changed to use extreme high polygonal details," he says. The first was the roughly million-triangle chandelier model, the second a 300,000-triangle detailed car model. "The performance impact of changing these models and rendering them with the triangle debug shader was only 15-20 percent," Pohl notes. "Looking at the content generation pipeline, this means that such high-resolution models can just be taken out of the 3D modeling software and can be used in the game directly without the need for artists to painfully reduce every possible triangle, and without the need to use tricks like normal maps to fake the detailed geometry," he adds. The full Intel-sponsored feature explains specifics of the setup, plus its implications for shading, particle systems and much more.