Originally posted by taks Isn't an image taken at ISO800 exactly the same as the image taken at ISO100 and then increased 3 stops in post in a camera without an accelerator chip?
Depends on the hardware. One would hope that the camera could do a slightly better job of it. There's lots of talk of ISO Invariance at this day (where increasing the ISO in your camera is pretty much no different to using post processing). Seems to be very few articles actually quoting actual lab results for any particular camera (lots of subjective observations about though). Roger Clark has some decent articles around this stuff.
Seems that all of the super crazy high ISO settings on modern cameras are likely using 'simulated' high ISO, meaning it is simply multiplying the digital values just like your post processing software would. This of course would be completely ISO invariant - indeed if the post software is clever enough, it may even do a better job of things.
At low ISO's though, my understanding is that your camera is likely using analogue gain to increase the charge presented to the digital converter. It makes sense to use this to get the signal (after gain) well above the noise that may be introduced later in the chain. This might only be effective up to something like 1600 or 3200 ISO.
For all intents and purposes, the K-3 III sensor might be close enough to invariant that nobody can tell anyway. So I guess pick whatever pleases you. I prefer to still set my ISO's on my K-1ii up to about 3200 max as required...I assume anything else is obsolete and can be done in post process. I think it's always better, if possible, just to get more light in somehow anyway.