thenextguy wrote in post #17942976
On the 1DX II I believe it occurs at 51,200 ISO. And I'm sure someone else will chime in that knows better, but "regular" ISO is done via hardware and "high" (or "low") is done via software in camera. Basically ISO 50 is a shot at ISO 100 and then pushed to ISO 50 via camera software.
It doesn't necessarily matter all that much. Analog gain only serves two purposes; to shrink noise that occurs after gain, and to avoid posterization when bit-depth is low. Other than that, it doesn't matter whether gain is math or analog gain. If the read noise at ISO 51,200 in RAW numbers is exactly the same as 2x that of 25,600, there is no point in using analog gain for 51,200, because there is nothing left to gain on.
ISO at the end of the day is nothing more than taking a certain amount of exposure of a gray card or object in a frame, and displaying it at medium gray. Anything else implied by ISO is just an artifact of temporary technological limits.
The fact is, using pure gain for very high ISOs is counter-productive; it tosses highlights unnecessarily, and makes compressed RAW files larger than they need to be. The last couple stops of analog-gain-only ISOs offer little or no noise benefit on most cameras, especially now that the post-gain noise has little banding and blotching compared to older cameras.