Three-dimensional computer graphics architecture

Three-dimensional (3D) computer graphics hardware has
emerged as an integral part of mainstream desktop PC
systems. The aim of this paper is to describe the 3D graph-
ics architecture at a level accessible to the general compu-
tational science community. We start with the generic 3D
graphics rendering algorithm, the computational require-
ments of each of its steps, and the basic architectural fea-
tures of 3D graphics processors. Then we survey the
architectural features that have been implemented in orproposed for state-of-the-art graphics processors at theprocessor and system levels to enable faster and higher-quality 3D graphics rendering. Finally, we describe a tax-onomy of parallel 3D rendering algorithms that accelerate the performance of 3D graphics using parallel processing.

1.   Introduction UNTIL recently, real-time three-dimensional (3D) computer,graphics was available only in very high-end machines from Silicon Graphics Inc. In the last few years however, the PC industry has seen an unprecedented growth of cost- effective 3D graphics accelerators. Because a significant amount of industrial research effort has been invested in powerful 3D graphics cards, it is predicted that the perform- ance of these accelerators will surpass the performance of SGI machines by the year 2001 (ref. 1). 3D graphics applica- tions place a stringent demand on the processing power  and on the data transfer bandwidth of the memory subsys- tem and interconnecting buses. The growing importance of 3D graphics applications has motivated CPU vendors to add new instructions to the existing instruction set architec- ture, and to develop higher-bandwidth memory and system buses. In fact, the data-intensive nature of 3D applications has been one of the primary motivations behind the introduction of advanced Dynamic RandomAccess Memory (DRAM) architectures for host memory, and the local memory on graphics cards.  In this article, we start with the basic steps required to render a polygon-based 3D graphics model and their asso- ciated and bandwidth requirements. Then we examine the major design issues in generating photo-realistic images on  esktop machines in real time, and the architectural innova tions that attempt to address these problems. Finally, we present a taxonomy of parallel rendering algo-rithms, which uses parallel processing hardware to render extremely complicated 3D graphics models.