• OS
  • who can use Linux but not windows? Share your Linux setup

DisplaysShouldNotBeTVs

I ran into some color management issues that I need to figure out.

The video recorded by the device uses Limited Range 16-235, which makes white look gray. Switching between Limited Range and Full Range in BetterDisplay for the recorder doesn't fix the issue.

The recorder can also take screenshots (a separate feature from video recording). These screenshots are in TGA format, and in those, the white background in the browser looks white with a value of 255. But there are differences in shades between the screenshots taken directly on the M1 Mac and on the recorder because the profile generated by macOS is not accurate (?). I also tried standard profiles, and it's the same story everywhere; the shades differ.

TGA should be lossless (https://en.wikipedia.org/wiki/Truevision_TGA), but the presence of something similar to aberrations around some colored objects seems strange to me. I need to figure this out as well.

Below are links to screenshots taken directly on my M1 Mac (Screenshot_X.png) and screenshots taken through the HDMI recorder (Recorder_X.tga): https://file.io/7N0aPzOsfdPg

If you need frames directly from the video, let me know, and I'll send those too, but they are in Limited Range 16-235.

    I came across something interesting. In Linux, if I enable 6-bit color, it seems that the FRC module on a 6bit+FRC monitor continues to strain my eyes a bit when reading. I replaced the Linux profile with the one from macOS. The eye strain when reading text disappeared, and the screen became a bit calmer. I've been testing this for two days now, and there's no mistake - the effect is almost immediate in terms of eye strain when reading text or lack thereof.

    It's possible that an incorrect color profile set by default in Linux could have affected color conversion, leading to incorrect interpretation of the 6-bit signal. This might have triggered FRC activation for compensation, even if the output was specifically set to 6-bit. But this is just a hypothesis for now.

      moonpie jeez pricey. I have two 3995wx thread ripper PCs which are much cheaper but 64core

      WhisperingWind

      I'm sure that applies to many different monitors too. I wonder what would be the best way to grab the color profile other than grabbing it off a Mac. What PC is this? The a770 one? I was wondering if the a770 is able to be made fully comfortable or to not bother with it

        WhisperingWind the presence of something similar to aberrations around some colored objects seems strange to me

        if you see this, these are pretty much what's going on

        1 (2011, Samsung):

        https://www.researchgate.net/publication/241222779_Depth_Perception_Enhancement_based_on_Chromostereopsis

        2 (2017):

        https://www.rose-hulman.edu/class/cs/csse451/Research%20paper/2019-papers/ChromaBlur.pdf

        3 (2018, continuation of 2, funded by Intel Labs):

        https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6126933/


        prior work that eventually lead up to this goes back to the mid-2000s, possibly earlier

        https://jov.arvojournals.org/article.aspx?articleid=2121865

        jordan

        The monitor's color profiles have different triangle coordinates (color gamut), different white points, and different gamma. Here's an example comparing two color profiles of my monitor from macOS and Linux. As you can see, there's a shift towards the yellow spectrum in Linux, which theoretically could be interpreted by the monitor as coordinates beyond the 6-bit range (but I'm not an expert in this): https://ibb.co/mRd9YtN

        I checked the color profiles for my TV: the triangle coordinates (color gamut) are the same, but the white point and gamma differ, which should not cause dithering at 8+ bits (TV has 8+FRC). This means that not all devices may have the issue; it definitely won't be a problem if the monitor doesn't support FRC.

        What PC is this? The a770 one?

        I don't know if Windows generates color profiles, but it can be checked.

        jordan

        Try Ubuntu 18, which the autobot installed. It seems that it doesn't support the Xe architecture, on which the A770 is based. It’s necessary to check that the i915 and xe modules are not loaded using the lsmod terminal command.

        If Ubuntu 18 is comfortable, so the monitor should be fine and the graphics card in basic mode is also okay. If that's not the case, then there's probably not much that can be done here without changing hardware. But it's hard to say for sure, as everyone has different sensitivity in their eyes, but for me, reading in this case is quite comfortable.

        If everything is okay, then you can switch to newer versions of Linux where video card support is available and try different desktop environments with various compositors. Try turning off text anti-aliasing in Ubuntu Tweaks (apt install gnome-tweaks).

        P.S. I agree 100% with JTL: you need to understand what specifically strains your eyes (vivid colors, contrast, high brightness of white areas, etc). This way, you'll have a set of small tasks to address, with the ability to check the results. This will be more manageable.

          moonpie What if we disable the GPU, aim at a reasonably expensive fast CPU(for ex. $500-600$), with the goal of running all steam games before a given date(for ex. before 2018) ?

          Would this work reasonably well , for those who are willing to settle on the gaming experience?

          • JTL replied to this.

            autobot Won't work.

            A better project would be to obtain a "good" setup with an older GPU and study what potentially causes strain with newer setups.

            Installed Zorin os lite 17.2, based

            on ubuntu 22.10.

            Driver - i915.

            Disabled xfce compositor(colors became somewhat muted), disabled font anti-aliasing, disabled firefox "hw acceleration" , disabled firefox "pages choose their own font".

            Based on short testing, feels good. But needs more testing.

            I came across an interesting comment in the kernel (module i915):

            'For Display < 13, Bits 5-7 of PIPE MISC represent DITHER BPC with valid values of: 6, 8, 10 BPC.

            ADLP+, the bits 5-7 represent PORT OUTPUT BPC with valid values of: 6, 8, 10, 12 BPC.'

            I think '13' refers to the 13th generation of Intel; ADLP stands for Alder Lake P.

            I wonder if this confirms that ADLP and later generations are free from dithering.

              WhisperingWind That's a great find.

              I think we need to look at the programming manual for those registers.

              I'll try to find the manual.

              • JTL replied to this.

                JTL thank you, that is very helpful.

                I looked there, found this registers manual that includes info about dither registers:

                https://cdrdv2-public.intel.com/703047/intel-gfx-prm-osrc-tgl-vol-02-c-command-reference-registers-part-2.pdf

                • It's iris xe gen 11th but alder lake n is 12th generation, but the registers are Probably the same
                • And yes, registers 7:5 from PIPE_MISC are responsible for dither_bpc, like the source code

                WhisperingWind looking at the manual for "dither" , it says it can be free from dithering, given the right register values.

                My i5-12450H uses PIPE_MISC_A, and the register value in Windows 11 is 00000100000000000000000000000000. Dithering is not enabled by default.

                  WhisperingWind

                  A. There are a few registers interacting, not only from PIPE_MISC, Probably, so we need to understand them and dump them all to test.

                  B. Another way to test whether its hardware or software or driver creating the issue is to try Zorin Os with gpu and with the settings I mentioned previously.

                    dev