In 3D visualization, real-time rendering of high-quality meshes in complex 3D environments is still one of the major challenges in computer graphics. New data acquisition techniques like 3D modeling and scanning have drastically increased the requirement for more complex models and the demand for higher display resolutions in recent years. Most of the existing acceleration techniques using a single GPU for rendering suffer from the limited GPU memory budget, the time-consuming sequential executions, and the finite display resolution. Recently, people have started building commodity workstations with multiple GPUs and multiple displays. As a result, more GPU memory is available across a distributed cluster of GPUs, more computational power is provided throughout the combination of multiple GPUs, and a higher display resolution can be achieved by connecting each GPU to a display monitor (resulting in a tiled large display configuration). However, using a multi-GPU workstation may not always give the desired rendering performance due to the imbalanced rendering workloads among GPUs and overheads caused by inter-GPU communication.

In this dissertation, I contribute a multi-GPU multi-display parallel rendering approach for complex 3D environments. The approach has the capability to support a high-performance and high-quality rendering of static and dynamic 3D environments. A novel parallel load balancing algorithm is developed based on a screen partitioning strategy to dynamically balance the number of vertices and triangles rendered by each GPU. The overhead of inter-GPU communication is minimized by transferring only a small amount of image pixels rather than chunks of 3D primitives with a novel frame exchanging algorithm. The state-of-the-art parallel mesh simplification and GPU out-of-core techniques are integrated into the multi-GPU multi-display system to accelerate the rendering process.

Library of Congress Subject Headings

Graphics processing units; Three-dimensional modeling; Rendering (Computer graphics)

Publication Date


Document Type


Student Type


Degree Name

Computing and Information Sciences (Ph.D.)

Department, Program, or Center

Computer Science (GCCIS)


Chao Peng

Advisor/Committee Member

Joe Geigel

Advisor/Committee Member

Christopher Egert


RIT – Main Campus

Plan Codes