Field Robots are often equipped with a Real Time Kinematic (RTK) GPS to obtain precise positioning. In many precision agriculture applications, however, the robot operates in semi-structured environments like orchards and row crops, where local sensors such as computer vision and laser range scanners can produce accurate positioning relative to the crops. GPS is then primarily needed for robust inter-row navigation. This work evaluates a new low-cost GPS. Static tests were used to test the absolute accuracy. To test the GPS in a precision agriculture environment it was installed on a robot driving in a simulated row crop field. The GPS supports raw data output as well and similar experiments were performed to evaluate the GPS when used in a RTK setup. In field tests more than 95% of the position errors were estimated to be within 2.6 m. In RTK field tests more than 95% of the position errors were estimated to be within 0.2 m. It was concluded that the GPS can be applied to selected applications in row crops and orchards if augmented by local sensors and mapping techniques. Using the GPS in a RTK setup applies to general applications where position errors of 0.2 m are acceptable.