Being a fine art photgoraphic printer/service bureau (old school term!) were are often asked, "Should I upload 8 bit or 16 bit?"
Short answer? It doesn't matter.
"It doesn't matter? How can you say that?! Clearly you don't know what you're talking about! Your printing must also be awful!" This was my fear. In fact, my opinion on this topic is why I didn't post a blog sooner, because I know there are so many "experts" out there saying the opposite and, well, bigger is ALWAYS better, right? Well, no, not always (insert Big Gulp joke). You CAN have too much of a good thing in life. In fact, in my opinion, much of life is about learning what to cut out. My decision to finally confront this question was inspired by this old 37signals blog focused on web development, but I feel it's applicable to this discussion:
37 signals "it doesn't matter" blog post here
In our current "pro-sumer" driven photo-technology market, everyone wants your money. Marketers of all sorts of photography software (and yes cameras) have to convince you that you NEED their products to make better images. What makes this really difficult for someone starting out trying to learn digital imaging (aka digital photography) is that the blogs they will find on the subject are often written by people trying to sell a photo product through internet "affiliate marketing." For blogs whose biz model is Ad based, this is less true, but ultimately theses sites have to release fresh content on the latest gear, which goes back to getting you to think you need something new to make good images. Photography forums can be an excellent source for information, but sometimes there is a lot of false info out there. Hopefully those forums can adopt the stack overflow and Reddit practice of "voting up" responses to questions, but I digress.
A professional photographer I work closely with is always showing me the latest Nik software that he's quickly applied to his images, and I'm like, "Well I can do that with a curve, faster and with more control." Don't get me wrong, these products have their place and I do use them for certain things, and if you don't use curves and quickly want to make your images look cool it's great, just like Instagram. But often I find these new tools cumbersome, and that sometimes they are overcomplicating the issue and ultimately slow me down. Not to say speed is everything, but quality from efficiency IS. In fact I consider efficiency necessary for quality, because if something you're doing is terribly inefficient, there is probably a better way to do it. Of course that's not true for everything, I consider myself a SLOW photo retoucher, but I'm slow about the things that DO matter. Knowing what matters and what doesn't is what makes a good practioner of any profession: photographer, designer, retoucher, artist, doctor, lawyer, engineer...you get the point. And the specifics of what matters is gained from research and ultimately, experience. No blog can tackle that, not in one go at least.
16 bit has been around a long time. I don't know exactly when it was released in PS, and "it doesn't matter". The point is, the 16 bit workflow has been available to professionals in the larger graphic arts community for a long time now, and still 8 bit has always been the default color space for image editing. And, I would argue, 8 bit has been sufficient for the most critical of digital photographic applications. In the past, the decision to use 8 bit over 16 bit was made because of storage limitations (in the 90's this was precious), but the point is the most high end photographer and retouching studios all used 8 bit (and generally I would bet still do). I know I still use 8 bit in my day to day retouching projects. I simply have never felt a need for more color data. If I need more highlight detail or more shadow detail, I just go back to the raw, reprocess the file, and mask it in to the main processed file I'm working in. You may say, "Well you could do that all in Lightroom," or whatever latest product on the market, and yes, that is true. And if you ARE NOT retouching, this workflow WOULD be faster and "better" (e.g. wedding photographers who can't spend too much time on an image, and aren't retouching at a high level). But having to spend hours on an image retouching an actress or model I simply find the LR workflow more time consuming and inefficient, because I need to retouch the image at some point and good luck doing that in LR. So what if I do everything in LR, process and retouch in PS? This makes sense, and I have tried it, but generally I still find this cumbersome and find myself going back to my usual, somewhat old school workflow, drum roll please...
Process in Capture One (best raw processing IMO), and work in 8 bit in PS, usually in RGB. If I have banding problems after curving in some contrast or brightness, I NEED to re-process from the raw and mask the problem area in, or in some cases just reprocess and start over. No one hates banding more than me, but it's not a common problem unless your shooting a lot of blue skies and gray backgrounds. You may think, "well if you had been working in 16 bit all along you wouldn't need to reprocess." I thought this too, until I had a banding problem on a gray background. The files I was working with were in 16 bit, as they came to me processed like that. I still had the banding problems, even with 16 bit. The only solution was to re-process the raw file at the correct density, or retouch it out somehow (usually by adding noise and blurring slightly), although I much prefer going back to the raw. Banding is a large topic in itself, but in my experience 16 bit has failed to stop banding from happening, despite my high hopes.
And all printers print in 8 bit!? Actually no, or rather, it depends on the printer and RIP driving it. The Polie for instance CAN print a 16 bit file, and as I understand it, "works in 16 bit." The LightJet actually works in 12 bit. And this is done to prevent banding from being introduced in the printing process, but if there is already banding in the image you'll still get banding. Which is kind of my point again; if you have banding, 16 bit won't get rid of it. Better to expose and process your images properly, avoiding banding altogether (some exceptions aside).