Advances in computer graphics and related technologies over the last several years have resulted in dramatic performance increases, numerous feature enhancements, and significant cost reductions for graphics computers. These factors, coupled with the heightened interest in virtual reality, have spurred the demand for low-cost visual simulation subsystems. The availability of relatively inexpensive graphics accelerators and workstations has helped fuel this demand with the promise of capabilities approaching those of special-purpose image generators. As the prices of traditional image generators fall and the capabilities of graphics workstations improve, the markets for these two product technologies will continue to converge. Caveat emptor applies for those not in tune with the rapidly blurring distinctions between general-purpose graphics workstations and special-purpose image generators.

Different kinds of simulation and training systems often require different capabilities and levels of visual simulation performance. This paper identifies typical selection criteria, including some that are common across many applications as well as those that are more specific to the simulator's intended use. Overall fidelity, realism, feature density, and other issues pertinent to the real-time simulation community are identified, and their trade-offs are discussed. These criteria are then used to compare the applicability of both special-purpose image generators and general-purpose graphics workstations to meeting different visual simulation requirements. This comparison provides a basis for specifying and selecting simulator visual systems in lieu of such esoteric metrics as polygon capacity. Understanding these issues and trade-offs will help those involved in the purchase of visual systems to make selection decisions that are not only cost effective but also meet the functional and performance needs of the end user.