I think most of us are pretty solidly confident that Canon has resolved their read noise issues, and are probably getting much closer to that 13.9 stops of maximum DR that Sony Exmor sensors are getting...so the difference is probably less than a stop, (personally I hope and believe it will be in the realm of 0.25 or less stops), of DR difference between any one of the 1D X, 5D III, D800, D4 and D7000.
I had hoped that but earlier today someone on DPR measured 6.02 ADU for the 5D3 at ISO 100 while my 5D2 measures 6.09 ADU.
Who did that measurement? If your talking someone on the forums, thats not what I'm referring to. If it was an official step wedge test based on JPEG output (which are usually what official DPR DR tests are based on), those tests are generally worthless (JPEG compression obliterates DR.) The DXO tests are very accurate, and measure the sensor hardware at a lower level than a JPEG step wedge test does. From a headroom standpoint, i.e. the ability to recover highlights or shadows, DXO's tests are more accurate (as demonstrated by the many videos on the net showing unbelievable shadow recovery with the D7000, where you can see before your eyes the recovery of 4-6 stops of what appears to be solid black in under exposed photos.)
Not a wedge test. Measuring the masked off black area of an ISO 100 RAW, finding the std dev in raw levels, then taking the max raw level - the raw black point and then dividing it by the read noise and then taking the log2 of that to get dynamic range in stops. JS provided the raw read of 6.02 ADU. It's the same thing DxO does and same thing I did in my tests of the 40D,20D,50D,7D,5D2,1D3.
To date, from ISO 100 to ISO 25600 test images, both JPEG and RAW, not one SOUL outside of you has mentioned any amount of banding noise visible in ANY image from the 5D III (most of which are from pre-production models even!) From the noise levels at ISO 6400, even at 25600, the 5D III, with black and shadow levels cranked up as high as they can go in Photoshop, the only thing visible even in the darkest blacks is clean, random noise. I cranked up EV on every ISO 100 image I could get my hands on, and I have seen nothing but a few speckles of random noise in the darkest shadows. You would need to produce an actual sample image that clearly demonstrates banding noise before I'd even consider that you are not purely fabricating the details our producing (i.e. 6.02 ADU.) You would also need to produce a link to whatever forum thread or article on DPR that you are referring to that supposedly measured the 5D III's read noise, because I'm highly skeptical anyone has had a chance to do any accurate measurements with the proper gear and care to produce anything remotely resembling a valid result.
Additionally, if you are trying to replicate DXO tests on your own with Photoshop/ACR, Lightroom, Aperture, or any other mainstream RAW editor, your not working with accurate data. Outside of DCRAW, every RAW processor applies a tone curve (picture style/image style/etc.) to the RAW image data before rendering to screen or file. You aren't looking at the unmodified RAW image data, even if you use a "neutral" tone curve...since even neutral is non-linear. You need to use the right kind of image processing software, as well as specially designed test targets, to test the full DR capable by a digital sensor. DXO not only uses a very specialized system to measure noise and DR (http://www.dxomark.com/index.php/About/In-depth-measurements/DxOMark-testing-protocols/Noise-dynamic-range
), they use the same system to measure every camera, so comparisons are accurate.