This research examines the accuracy of computer programmes to simulate the illuminance level in atrium buildings compare to the measurement of those in physical models. The case was taken in atrium building with 4 types of roof i.e. pitched roof, barrel vault roof, monitor pitched roof (both monitor pitched roof and monitor barrel vault roof), and north light roof (both with north orientation and south orientation). The results show that both methods have agreement and disagreement. They show the same pattern of daylight distribution. In the other side, in terms of daylight factors, computer simulation tends to underestimate calculation compared to physical model measurement, while for average and minimum illumination, it tends to overestimate the calculation.