It's also worth noting why DXO doesn't like this kind of in-camera RAW NR.
According to DXO, NR is best done out of camera because:
- in-camera RAW NR is not as likely to be as sophisticated as out-of-camera NR software;
- in-camera RAW NR is hardwired into the camera, and can't improve over time, unlike out-of-camera RAW convertors;
- in-camera RAW NR can destroy fine detail.
DxOMark - Noise reduction
and:
DxOMark - Conclusion
The first point is sensible - in-camera RAW processing is not likely to be as powerful as the RAW processor you might run on a modern PC. This is also a logical position to take if you also happen to sell RAW conversion software
Less certain is the notion that in-RAW NR is bad because it is hard-wired into the camera and can never improve, whereas out-of-camera RAW processing can evolve over time as algorithms and the software get better. This argument is weak because if the in-camera RAW NR is implemented in firmware, that code can also change over time too with a firmware update. This is also a position you would expect a RAW software vendor to make
The most crucial point DXO raise is that in-camera RAW NR can destroy fine detail, and once it that detail is destroyed and gone at a low-level within the camera, you can never get it back.
However this argument may be a weakened if it can be demonstrated that a strong level of detail is indeed still emerging out of a sensor even if some in-camera RAW NR is suspected.
In other words, if there is a ton of fine detail still evident in the RAW output at high-ISO from a camera, no-one will care (even DXO) if there is any in-camera RAW NR going on. That may turn out to be the story of the K-5.