The performance of a swarm of robots depends on the hardware quality of the robots in the swarm. A swarm of robots with high-quality sensors and actuators is expected to out-perform a swarm of robots with low-quality sensors and actuators. This paper directly investigates the relationship between hardware quality and swarm performance. We take three common components of swarm algorithms (trail following, swarm expansion, and shape formation) and measure how they are affected by two common types of hardware inaccuracy (communication bearing reception error, and movement error) both in simulation and with E-Puck robots. We find that large amounts of both types of hardware error are required before performance appreciably decreases.