Nikon is claiming image quality that is a full stop better at every ISO, e.g. ISO 12800 is as good as ISO 6400 on the D810. And without losing quality at the base ISO.
John W writes:
I have now looked at some of the full-res sample Nikon D850 images online, and they look very good - except that they have not allayed my concerns about dynamic range. In fact, a few of them have increased my concern. But, I think what I was seeing were straight out-of-camera JPEG, so I guess we really don't know much about the dynamic range until we get a RAW editor.
DIGLLOYD: Evaluations of noise must verify proper exposure, which is why I advise getting and using RawDigger as well as understanding ETTR exposure (Expose To The Right). That’s why most “lab” or “quick” tests are of little value—they do not verify that the sensor dynamic range is actually used, which is insanity unless one wants to evaluate noise based on auto-exposure (which changes with metering mode and subject matter)—untenable as a defensible approach. In short, I find that few noise evaluations have any merit, at least not for any serious about exposing properly.
JPEGs are meaningless for noise comparisons unless one is specifically evaluating JPEG performance. The camera configuration has so many variables that influence how noise the JPEG is, so unless all settings are spelled out, JPEG noise tells us little. Moreover, one camera could use settings that make it look better than another in JPEG. RAW is the only proper way to evaluate noise.
Just because two cameras are ISO 100 does not mean they meter and behave the same at ISO 100. So fixed lighting and ISO is not a defensible metric either. Most recently, see Increase Image Quality by Optimal Exposure (ETTR, Dolls) in my review of the Hasselblad X1D which has commentary on the very different exposure behavior of the Nikon D810 vs the Hasselblad X1D.
Many cameras auto-expose in a manner that gives up a whopping 1.5 to 2.5 stops of dynamic range and that is why I generally include the RawDigger histogram when evaluating noise—see Real World Dynamic Range in my review of the Nikon D810 as just one example as well as Shootout vs Nikon D810: 4-stop Underexposure + Push (Flowers) when comparing the Nikon D810 to the Hasselblad X1D.