An air cooling scheme is proposed for 3-D packaging. It consists of converging, impinging, and diverging flows among stacked circuit packs. Its cooling concept is demonstrated by experimental studies. Using a small Muffin fan (Flight-80 series), the scheme can cool a very compact system, 8 cm × 8 cm × 12 cm, with 56 Watts power dissipation. The maximum temperature difference between the circuit packs and the air is under 50°C, without the use of heat sinks. The flow and heat transfer behaviors of the system are then characterized and optimized by numerical simulation. The sizes and locations of the plates’ openings for the flow passage are identified as the most effective parameters. The optimization adjusts these parameters or power dissipation levels to achieve a uniform distribution of plate temperatures. The uniform plate temperature distribution results in a higher system cooling capability, e.g., 71 Watts, and lower levels of temperature-gradient related problems. The improved cooling capability can be estimated by a simple formula derived by a linear relationship between the plate’s temperature rise and its power dissipation level. The scheme can be applied to higher-power systems by adding more cooling surfaces and a more powerful fan.