Abstract-The quality of atmospheric corrections provided by a dense reference network for centimeter-accurate carrierphase differential GNSS (CDGNSS) positioning is investigated. A dense reference network (less than 20 km inter-station distance) offers significant benefits for mass-market users, enabling lowcost (including single-frequency) CDGNSS positioning with rapid integer ambiguity resolution. Precise positioning on a massmarket platform would significantly influence the world economy, ushering in a host of consumer-focused applications such as globally-registered augmented and virtual reality and improved all-weather safety and efficiency for intelligent transportation systems, applications which have so far been hampered by the several-meter-level errors in standard GNSS positioning. This contribution examines CDGNSS integer ambiguity resolution performance in terms of network correction uncertainty, and network correction uncertainty, in turn, in terms of network density. It considers the total error in network corrections: a sum of ionospheric, tropospheric, and reference station multipath components. The paper's primary goal is to identify the network density beyond which mass-market users would see no further significant improvement in ambiguity resolution performance. It finishes by describing development and deployment of a low-cost dense reference network in Austin, Texas.