I added information to the my main post regarding the fact that on AMD video cards dithering is always enabled by default in win10 and in linux.

After you set it in 0x0, or 0x11, it feel much better. As someone wrote in web, dithering is blackbox which need very accurate adjust for your current display (which has own pixel-invertion type and FRC). If all 3 types of pixel moving sync bad between themselfs, it produce headache

upd: according to BetterDisplay research, only part of the dithering has been disabled. I think I got same results, only partitial disable - final signal before sending to the screen is now free of dithering. But first part, which still strain, could be… ColorSpaceConversion. AMD 780m let me choose gamut I need: CIE RGB, rec2020, rec709 etc. Safe IRIS XE miniPC - doesnt use such convertion. It can make CSC, but it disabled by default. And IRIS XE is comfortable. When AMD 780m, even after setting all values to 0x0 - not.

All old hardware, use only sRGB colorspace. New hardware can make convertions. All pixel data re-calced with formulas, and need dithering to smooth results. We need to find, how to disable CSC in 780m, is it real, or hardware use it without any user controls

my settings are response time set to super fast , custom coloor because I calibrate with spyder x pro and everything else set default … i only play with brightness levels acording to enviroment ..I wish companys would include this by default ..

moonpie Check cables and bandwidth.

I tried HDMI/DP cables. I also tried UMR with FMT_BIT_DEPTH_CONTROL settings to disable dithering in each port (FMT0…FMT2), in parallel I disabled all FMT_DYNAMIC_EXP_CNTL and FMT_CLAMP_CNTL settings. After researching registers, I set same values in win10 via ditherig, and get same disabled (0x0) output values. I also tried different gamma and gamuts, sharpening, scaling etc. The dithering results was same - 0xC900 as driver_default in linux/win10, 0x0 in each FMT after turning it off. Moreover, old vega6/7 has same dithering values by default 0xC900 but okay for eyes. And next fact - without AMD drivers in win10, dithering not working (0x0 values) in AMD cards.

After that, I suppose 2 variants: another pipeline (not dithering) give eye-strain and controlled via another registers, or grapfic has hardware acceleration which is not controlled

    moonpie amdgpu.dc=0 ??

    no, it a bit complicated 🙂 it is 4kb values from registries address of your pci-e device, you need to read/write

    I had read Mike's research, it is similar as mine. He used old hardware, I use modern. But old/new AMD GPU's controlled via ADL in same way. The dithering 0x0 reduce strain, but not to 0 level. Same as mates from Apple thread there did, disabling dithering in Macs.

      3 months later
      10 days later
      dev