DNA microarrays are becoming a frequently used research tool. Whilst several studies have confirmed the reproducibility of analysing the same RNA samples on duplicate arrays, there is little analysis of the reproducibility of the results of transcript profiling between microarrays carrying different probes to a common set of genes. To address this question, we compared the performance and reproducibility of two microarrays commonly used in plant research, the Affymetrix Arabidopsis AG array containing more than 8000 probe sets and the Affymetrix Arabidopsis ATH1 array containing more than 22,000 redesigned probe sets. A total of 21 different RNA samples were labelled and hybridized in parallel to the two microarray types. Focusing on the overlap of more than 7300 targets detected with both arrays, we found a high degree of reproducibility. Despite the use of different probe sets, both signal and signal log ratio were very similar for most genes. However, genes that were called absent or not changed by Affymetrix' statistical algorithm implemented in MAS5.0 showed considerably less conservation of expression patterns. Moreover, we identified about 300 genes that yielded strongly different measurements with the two microarrays, emphasizing that RNA profiling data need careful interpretation. Overall, this study shows that results obtained with ATH1 and AG arrays are very comparable and hence that the analysis is largely independent of probe sets. However, the result emphasize the need for appropriate filtering schemes such as those based on the present and change calls provided by MAS5.0 rather than reliance solely on signal values.