9

Most monitors that I see these days support 32-bit color depth. I can tell the difference between 16-bit, 24-bit, and 32-bit color depth but I am wondering if the human eye could tell the difference between 40-bit, 48-bit, etc. color depth?

4
  • 1
    Given wide enough gamut yes you can. PS 32 is not divisible by 3
    – joojaa
    Commented Feb 11, 2015 at 18:11
  • 1
    32 probably means 8 bits each of red, green, blue, and alpha. Most monitors these days support 24-bit RGB (8 bits per color sample). For gruesome details see en.wikipedia.org/wiki/Color_depth. The human eye can easily distinguish 8-bit color samples, but it's harder to see a difference if the colors are logarithmically (gamma) or sRGB encoded. Commented Feb 11, 2015 at 19:03
  • Relevant. Commented Feb 14, 2015 at 10:42
  • I can personally see the difference between one-off colors in 8 BPC (24/32 BPP) color depth. I can easily see the difference between 222b14 and 232c15 if they're right next to each other. But I don't know if my eyesight is near the average. Commented Oct 26, 2018 at 1:16

4 Answers 4

20

You have to be a little careful with the definitions.

24 bit per pixel and 32 bit per pixel

“24 bit” usually means 24 bits total per pixel, with 8 bits per channel for red, green and blue, or 16,777,216 total colours. This is sometimes referred to as 24 bit RGB.

“32 bit” also usually means 32 bits total per pixel, and 8 bits per channel, with an additional 8 bit alpha channel that’s used for transparency. 16,777,216 colours again. This is sometimes referred to as 32 bit RGBA.

24 bit and 32 bit can mean the same thing, in terms of possible colours. It’s also worth noting that transparency doesn’t need to be sent to your display, because displays are opaque (you can’t see through your display to what’s behind it, unless you’re Tony Stark).

32 bit per channel

Also, 32 bit occasionally means 32 bits per channel (128 bits total per pixel). And, a lot of the time 32 bit per channel uses floating point numbers, rather than integers. (I’m happy to add more detail on floating point vs integer, if you’d like.)

The OpenEXR format supports 32 bit float channels. That might sound excessive, but it’s often used for VFX and rendered material, where heavy processing or colour correction may be involved, and large files are less of a concern.

16 bit per pixel or 16 bit per channel?

“16 bit” can typically mean two different things: 16 bits per pixel or 16 bits per channel. 16 bits per pixel works out to be 65,536 possible colours, and it definitely looks worse than 24 bits per pixel. 16 bits per channel means 281,474,976,710,656 total colours — well beyond human perception, but handy for processing.

RGB or CMYK?

All the above information assumes you’re working with RGB or RGBA images. If an image is CMYK, it could be 8 bit per channel and 32 bit per pixel, with 8 bits for cyan, magenta, yellow and black channels.

I am wondering if the human eye could tell the difference between 40-bit, 48-bit, etc. color depth?

I think 8 bit per channel (24 bit per pixel) is on the fringe of what the human eye can easily distinguish, but that’s only part of the story. Processing can cause rounding and clipping, so additional colour depth can push errors beyond the point where humans and display technology can see them.

That is one of the reasons why it’s common for RAW camera formats to be 10, 12 or even 14 bits per channel, which works out to be 30, 36 or 42 bits per pixel. It’s also common for those working on photos to import RAW to a 16 bit per channel document for further manipulation. Pro video recording can be 10 bits per channel or higher, too.

And, in cases where you might not think there’s any processing going on, there might be — colour management alone can introduce additional processing.

8 bits per channel means there’s only 256 levels of intensity, which really isn’t much. Common causes for rounding errors to be visible when using 8 bits per channel:

  • Using gradients stacked on top of each other, where layers aren’t 100% opacity.
  • Gradients drawn without decent dithering.
  • Shadows with large blurs.
  • Blurred objects.
  • Blending modes and other compositing of two or more layers.

bpp vs bpc

Now might be a good time to mention the shorthand that’s often used for bits per pixel, bpp and bits per channel, bpc. It’s common to write 32bpp etc when talking about these things, to remove the ambiguity of saying 32 bit.

Dynamic range and gamma

Dynamic range should also be factored in. It’s typical for displays to target sRGB (gamma of 2.2). Wider dynamic range means the number of possible values are stretched further, so more colour resolution is needed.

Is 8 bits per channel enough for final asset delivery?

Yes, most of the time, depending on the use.

Is 8 bits per channel enough for creation?

Sometimes. But, often it is not.


Please note: I’ve added some more information relating to digijim and Warren Young’s comments below.

10
  • 1
    Maybe I missed it, but "32 bit" can also refer to CMYK images having 8 bits per channel.
    – digijim
    Commented Feb 12, 2015 at 21:05
  • @digijim Good point. I'll add in a section about CMYK (just to add more confusion). Commented Feb 13, 2015 at 0:33
  • Epic answer, +1. A couple of additional points, though. First, 14 bpc raw is not uncommon these days at the high end of the DSLR world. For the truly well-heeled, you have things like Phase One medium format digital camera backs that go to 16 bpc. In the pro video world, 30 bpp is quite common. This usually gets stuffed into a 32-bit word per pixel for ease of computer processing, so there's another "32 bit" confusion for ya. The two remaining bits are typically set to 0, with competing companies fighting about which end the 0 bits go on. :) Commented Feb 14, 2015 at 10:46
  • @WarrenYoung I’m not up to speed with recent RAW formats, so thanks for the additional info. I’ve added a bunch of info based on your comments. Also added a credit to the bottom. Thanks again. Commented Feb 15, 2015 at 7:02
  • 1
    @Startec 32-bit float is less accurate than 32-bit integer (if you consider the normalised, 0-1 range). But, 32-bit float allows over and undershooting, which stops clipping. Values like 1.5 are acceptable with 32-bit float. I think that’s usually why float is used when the depth is 32-bit — may as well avoid clipping if there’s enough resolution to avoid rounding. Commented Apr 23, 2016 at 23:40
3

The answer is yes. It is also no at the same time. See a monitor is not able to show all the colors humans can sense. Since humans certainly can sometimes see differences between the colors of the 8bit per color palette it means more could be useful.

But if you extend the color range and dynamics then you also need to boost the resolution of the color system. Thing is if you do then you have a lot of space for resolution improvements.

Allready last gen graphics cards were able to adress 10bits per channel, wether your os or entire pipe could or not. Most dslr cameras capture 12, 14,16 bits etc.

Displays today do not show alpha channels, so i doubt your hardware will be 32 bits etc. So the numers your saying dont really make sense for monitors.

Anyway for editting the images it certainly makes sense to have more data. As otherwise your edit might make the limitation visible.

2
  • In a strange sort of way, "32-bit support" makes sense on a spec sheet. While the monitor may only see 24-bit information (ignoring its own menus, etc.), the user's computer may be set to 32-bit colour (which basically means "my graphics card/processor can handle alpha overlays; don't waste CPU time calculating this") and putting "24-bit MAX" on the spec sheet/user guide can lead the uninformed to put their machine into a state of poorer performance for no good reason. Yes, it's a bit of a lie, but it's a white lie. Commented Feb 13, 2015 at 2:36
  • @StanRogers sure but the graphics hardware of most graphics cards today can do programmable processing so they support 64bit processing or half float processing or full float processing etc why not ay so.... bigger numbers are better no. besides like i said they are mostly capable of more than 24 bit... just clarifying.
    – joojaa
    Commented Feb 13, 2015 at 4:28
0

If height mapping is part of your job, then the answer is no. By the time higher color depth becomes standard height maps will probably be a thing of the past though.

0

Info on 8-bit (24bpp) Vs 10 (32bpp) bit monitor and gpu. Also professional graphics vs consumer.

Professional graphics:

Honestly this matters more for photographers and videographers for professional graphics work than others like design or illustration. It helps with color grading and gradient banning. A good monitor calibrator and 24bpp can go along way for graphic design.

For example if a videogeapher zoomed in over 65% on a metal detail and looked at its gradient they can then color grade more accurately than normal an can tell if there is a shift off with the white color balance to a much higher degree than with a non 10-bit monitor.

Apparently there are still issues with Photoshop zoomed in over 65% reverting to 8-bit color as of 2017, but photographers can still get a benifit from 10-bit screens if their trying to shoot over standard color gamunts or want to gain just a tiny bit more adobeRGB gamut for print proofing - though they still need to proof the prints them selves regardless anyway -.

Consumer graphics:

People who own a consumer grade or better 10-bit monitor actually have already had support from AMD Radeon gpu's for years and as of 2017 Most Gtx gpu's support it as well. Windows 10 and macOS both just implemented native 30-bit - aka 32bpp - support, with linux having it a few years already. All three OS still have their own issues relating to 32bpp support.

For games, video, and other applications 32bpp support really depends on the developer and the color standards individuals publish content at, many of the latest games in theory support 32bpp color already.

Monitors:

You should know that exteneding the 32bpp color range of a monitor will not mean that that extended color range will be in the desiered color gamut. While most of them will fully cover sRGB there will often be a compromise between Adobe RGB, DCI-P3, or a even lower standard video gamut.

Even if it is a wide gamut monitor it may not cover Adobe RGB or DCI-P3 gamuts respectively over 95% for color grading. I recommend reviewing the monitor your looking at on tomshardware.com, they will also have a list of the best professional monitors based on their testing each year.

It is generally recommended that you get a 27in or larger monitor if your going to run a single monitor setup, and as high a resolution as you can afford. You'll want one that is a 60hz or better refresh rate with a response time of under 10ms for daily workflow.

GPU:

The GTX Gpu family only supports 32bpp for DirectX applications, most professional software use Opencl - or possibly in the future 'Vulkan' based off Opencl -. Professional applications won't work with GTX gpu as of 2017 or earlier for 32bpp color.

Quadro Gpu will support 32bpp, mainly starting from 2015 or later please review each spec sheet for said Quadro as I do know their general support for 32bpp color started later than with Radeons or FirePros. Quadro are workstation class gpu but have less hardware power per dollar then the GTX line up, so for graphic design areas other than 32bpp they may not be the best solution.

FirePro gpu - or the radeon WX series - have long had 32bpp support even though computers for the must part until now haven been able to benefit from it. FirePro are workstation workstation class gpu but have less hardware power per dollar then the regular radeon line up, so for areas other than 32bpp they may not be the best solution for graphic designers.

Radeon gpu use opencl or vulkan as standard in their drivers and some can be modified to use 32bpp with professional applications, but some can not. In theory 32bpp in genereal has been supported in the Radeon gpu line up for years now.

Either adobe or AMD will not now allow Radeon drivers to work with the adobe suite of applications even though in theory it is natively supported by the gpu in the first place. However if you can infact find one that works for 32bpp in adobe and has been recently released, it is probably the best value per dollar.

Note the Quardo and FirePro are better for higher precision for 3d printing and other high precision areas like medical imaging.

Other general PC hardware:

For Cpus more cores are better for multitasking or anything to do with video. For non-video adobe programs they only use a few cores mainly so the faster frequency matters more, but only to a small degree.

For RAM you'll want a minimum of 16GB for professional use, 32GB or more can help for video editing but not as big of a difference as you'd see compared to anything under 16GB. When buying RAM over 16GB wait to get the best price on it you possibly can as the price range fluctuates easily, example would be 16GB of ram jumping from $45 to $265 for the same exact ram sticks in under a month only to fall back down to $45 a half a year later. Just plan it based off your motherboard.

For Memory you'll want to use a SSD for your main OS - like windows - and programs for speed, M2 SSDs actually show even more improved program speed in some tasks over an SSD so consider them for your daily drive too.

For memory storage You'll probably store your finished images or videos on a HHD for a larger area to store stuff like working files, photos, or videos since they have a better price per GB. Note that HHDs while cheaper have a higher rate of failure than SSDs though.

For Cases and Cooling Most professionals prefer to buy cases and cooling based off how quiet they are as well as their cooling abilities before anything else. Noctua are by far the quietest cpu coolers and fans for the money, the NH-D14 is one of the best cheap cpu coolers on the market. The fractal Design R5 is claimed to be the quietest case as of November 1st 2017.

Hope that helps!

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.