Developing a massively multiplayer online game which utilizes physically based simulation to provide realistic behaviors requires numerical integration functions with inherently high computational costs. This simulation, performed on the individual clients of a peer-to-peer networked game or for a client/server online game, presents challenges due to many factors, including limited computing resources at the client level and network latency in the propagation of a client's state to other clients. Computationally intensive simulation may adversely affect performance and result in a situation in which little processing capacity is left for other aspects of the game. In this paper, we explore how a game developer who is aware of these issues might create a game for IBM's recently announced Cell Broadband Engine[TM] processor; we also present an example of the development of a game in which multiple human and robotic characters interact with static and dynamic objects in a virtual environment. Although our experience suggests that porting code to the Cell Broadband Engine core with minimal use of its synergistic processing elements (SPEs) should not be expected to produce significant performance gains at this time, the potential of the Cell SPEs to improve performance is considerable. We discuss performance and design and implementation decisions, with programmability issues being especially noted.

The video game industry had an annual revenue of approximately $25.4B in 2004, and this is projected to increase to $54.6B by 2009, according to Price Waterhouse Coopers. (1) This figure represents a 16.5 percent compound annual growth rate. Game platforms (both PCs and consoles) host non-networked and multiplayer networked games. Although revenue from offline games has been dominant over the years, analysts' predictions suggest that multiplayer online game sales will eventually dwarf those of traditional console PC games, with revenues approaching $5.2B by 2006. (2)

Currently, online games are predominantly turn-based role-playing games with limited behavioral dynamics incorporated into game play. Several problems, such as low network bandwidth and high latency as well as low server-side computing density, are limiting the state of the art in massively multiplayer online games (MMOGs). As the number of households with broadband access increases in the United States and around the world, the network limitations should be significantly reduced. The next generation of MMOGs will likely employ a much higher level of physically based modeling, artificial intelligence, dynamic story lines, user-created content, and gesture-based input devices, and a higher level of visual quality, motivated by a new family of high-performance game consoles equipped with integrated Ethernet controllers. Powerful servers based on the same microprocessor technology used in these consoles can enable the simulation of virtual worlds on a massive scale.

Game developers using current technology are forced to use a design model that binds the scope, quality, quantity, and complexity of interactivity between players. This limits both the type of game that developers can create and the market penetration of the game. The design model is limited due to many problems, including communications latency, bandwidth constraints, algorithmic and computational complexity, development and implementation costs, implementation time, and the inherent complexities of combining visual, musical, and literary arts with technology. MMOGs have traditionaily addressed these challenges by imposing design constraints on the level of interactivity or realism provided in the game. For example, games such as EverQuest ** or Ultima Online ** do not attempt to provide a level of realism that depends on physical simulation, whereas Guild Wars ** provides a higher level of visual realism and attempts to include better simulation than traditional role-playing games.

Physical simulation allows computer-generated objects to behave in a physically realistic manner. This is accomplished by formulating and evaluating a mathematical model consistent with the laws of physics governing the behaviors being simulated. Physical simulation is very computationally intensive and consequently very slow. It can also result in unrealistic behaviors due to instabilities at extremes in the simulation, such as very heavy or very light objects, extreme forces exerted at joints, or very large velocities. These challenges often lead simulation models to use physics for some aspects, but rely on nonphysical simulations for others. For example, the simulation of a multibody object using articulated joints may constrain one body to move itself after each time step to maintain a consistent position or orientation to an adjacent body. Higher-performing and more stable "pseudo-physics" is typically used in video games, but may be inadequate for other simulations such as those required for simulating animal behavior or mechanical systems.