Just a couple of viewpoints:
1) The print size assumed in DxO tests is a pretty small one, around A4. They have chosen this so they could compare cameras having 8 MPixels and upwards. Not necessarily adequate comparison if you print A2s.
2) I don't care that much about small differences regarding noise. Although i wouldn't argue with DxO's way of measuring noise, it works the way that what you measure is not always what you see.
Digging to all information on DxO I would say that the 5DII seems to have about one stop advantage over the A900 in most regards at high ISO.
Regarding this article on LL:
Michael showed the "Screen" comparison for SNR 18%, whereas the more appropriate comparison would have been the "Print" comparison. It makes little difference given that the megapixel count is close, but the "Print" results show less than a 3dB difference between the two cameras at any measured ISO. According to Michael, a difference less than 3db ought to be visibly insignificant at any comparable ISO. Looking at the attached figure, A900 manufacturer ISO 6400 is only 1/2 stop noisier than 5D II manufacturer ISO 6400. Even this minor difference is partially explained by the difference in measured ISO (4706 for the A900 versus 3990 for the 5D II).
Michael's states that "At ISO 800 and higher the Canon is visibly superior, especially in the noise department." I don't discount his observation, but it doesn't jibe with the DxO data presented in the same article.