When vector-based GIS-packages are used for traffic noise impact assessments, the buffer-technique is usually employed for the study: 1. For each road segment buffer-zones representing different noise-intervals are generated, 2. The buffers from all road segments are smoothed together, and 3. The number of buildings within the buffers are enumerated. This technique provides an inaccurate assessment of the noise diffusion since it does not correct for buildings barrier and reflection to noise. The paper presents the results from a research project where the traditional noise buffer technique was compared with a new method which includes these corrections. Both methods follow the Common Nordic Noise Calculation Model, although the traditional buffer technique ignores parts of the model. The basis for the work was a digital map of roads and building polygons, combined with a traffic- and road database and the Danish Building- and Residence Register (BBR). In order to test the methods, a total noise impact assessment was made for the municipality of Middelfart (20,000 inhabitants). As the new method considers buildings barrier to noise diffusion, efforts were made to synthetically create a 3D-model from the 2D digital map by utilising the information in the BBR-register. Thus, the method can also estimate the noise on each floor and it takes care of the differences in barriers from tall buildings versus low buildings. The practical testing of the methods in Middelfart showed that the traditional noise buffer technique compared with the new method systematically overestimated noise affected residences by 17% and overestimated strongly affected residences by 282%. Thus, the study showed that the traditional buffer technique provides unreliable results in urban areas compared with the more detailed method.