Marc, you must have been checking for those results daily to notice them so quickly.
Just in case anyone is confused about the noise differences betwee the 5D2 and 5D3, as shown on the DXOMark graphs, I'll try and summarize them.
(1)ISO sensitivities are about the same for both cameras at lower ISOs, the 5D3 being very slightly more sensitive to an insignificant degree. However, such differences seem to increase as ISO increases, so by ISO 25,600 the 5D3 is about 1/4 of a stop more sensitive.
What this means in practice is that when using the same shutter speed at ISO 25,600 with both cameras, the 5D2 sensor will receive 1/4 of a stop less exposure (in terms of its full-well capacity) and will therefore have 1/4 of a stop more noise, and 0.25EV less DR, irrespective
of any other differences
However, there are
other differences. At the manufacturer's nominated ISO of 25,600, the 5D3 has very close to 1 stop lower noise at 18% grey (the midtones), and very close to 1 stop better DR. Add the 1/4th of a stop difference in ISO sentivities and we get slightly more than 1 stop better DR and slightly more than 1 stop lower midtone noise for the 5D3, at the same shutter speed, with the same lens used at the same aperture.
However, if different lenses are used, then all bets are off. Using different lenses with both cameras, even at the same F/stop and same focal length, using the two cameras at ISO 25,600 and at the same shutter speed, the sensors could either have close to the same performance, or the gap could be widened to as much as a 2 stop difference, depending on the T-stop values of the two lenses.
(2) At the more commonly used ISOs, from base up to ISO 1600, there's an insignificant difference between the two cameras with respect to SNR, DR, Tonal Range and Color Sensitivity. At extreme pixel-peeping levels, one might be able to discern very slightly less midtone noise in the 5D3 image, but I doubt it.
Up to ISO 1600, DR for both cameras is virtually identical, with a very slight edge going to the 5D2, which is surprising. However, such difference are probably within the fluctuation range due to QC differences between different units of the same camera model, and any imprecision in the testing procedures. Nothing's perfect.
(3) Bottom line: any improvement in the 5D3 sensor performance (regarding noise and DR compared with the 5D2), seems to be confined to very high ISOs above ISO 1600. The 12mp D3s still has the edge at high ISO, having on average about 1/2 a stop better DR at the same print size, at high ISO.
Hope this is all crystal clear.