16bit anyone?

Yep 12bit for the k10d and for the k20d
Maybe we might see 14bit for the k7 though it seem to become standard with high-midend cameras

The K10D has a 22bit A/D converter, the K20D has a 14bit one. I presume that 22bit A/D conversion with the higher pixel count would slow things down too much.

I work in 16-bit when I need to do serious PP. When "pushing" an 8-bit file too far you can get unwanted posterisation.
Still, I always work with layers, so even if I work on an 8-bit and find it's going pear shaped, then I can reprocess the RAW file into 16-bit and pull all the layers onto the new 16-bit file.

It's quicker to work in 8-bit, and usually it's just fine
Matt

To be honest, I don't see a real difference in time in photoshop, maybe that the 64bit version has something to do with that... I don't know really, maybe I never push it far enough to notice a difference

Yes I believe that 8bit is good enough, I mixed up somethings so 8-bit looks like it was really cutting it short with things.

it's 12 bits per channel so it has 4096 shades for each colour.
The K20D allso stores RAW data as 12 bits/channel.
8 bits per channel is what you call 24 bit colour in display talk, 16,7 million.
12 bits per channel is 36 bits colour in display talk 68.719.476.736 colours
16 bits per channel is 48 bits colour in display talk so: 281.474.976.710.656 colours.

12 bits colour is allready way more then the human eye can show and miles from what your graphics card can display.

But again, it is not about the number of 2^24 (16 something milions).
When working on an 24 bit pic (8 bit per channel), operations like levels, curves, etc operate on 8 bits data (channel), what gives max of 256 possible values, so as we can imagine i.e. streching levels will lead to significant gaps in the histogram - that is the way to posterization.