We present a detailed comparison of predictions at NNLO accuracy
computed by four publicly available computer codes for Drell-Yan
processes at the LHC and Tevatron colliders. We point out that while
there is agreement among the predictions at the next-to-leading order
accuracy, the predictions for fiducial cross sections at the
next-to-next-to-leading order (NNLO) differ, whose extent depends on
the observable. The sizes of the differences in general are at least
similar, sometimes larger than the sizes of the NNLO corrections themselves.