The evaluation of the agreement between measured and calculated dose plays an essential role in the quality assurance (QA) procedures for intensity modulated radiation therapy (IMRT). Film dosimetry has been widely adopted for this purpose due to excellent film characteristics in terms of spatial resolution; unfortunately, it is a time-consuming procedure and requires great care if film has to be used as an absolute dosimeter. If this is not the case, then an independent ionimetric measurement is mandatory to assess the absolute dose agreement. Arrays of detectors are now replacing films for routine IMRT QA, since they permit very simple verification procedures. They show excellent characteristics in terms of linearity, repeatability, and independence of the response from the dose rate, but at the same time present a poor spatial resolution, due to the limited number of detectors available. In our institution, a diode matrix (MapCHECK, provided by Sun Nuclear) is adopted for routine QA. The aim of this work is to compare the performances of absolute film dosimetry with this matrix in QA procedures and to investigate the origin of possible discrepancies between the two methods. The results we present show a very good agreement between the two detectors when used to assess the mean dose deviation between calculated and measured doses (in both cases 0.2%). If the y matrix method is adopted, MapCHECK response shows a slightly better agreement with computed dose distribution than film dosimetry (mean percentage of points satisfying the constraint y < or = 1: 96% versus 94%). This difference is shown not to depend on the different field sampling, but on the detectors' capabilities. Moreover, we show that the diode matrix is able to identify eventual delivery errors as well as film. Our conclusion is that the diode matrix may effectively replace both film dosimetry and ionimetric measurements in routine IMRT QA.