• Hardware

  • Intel confirmed no dithering on Intel arc a770

a month later

whystrainwhy That's great news. Can you share the message from them ?

Also is it also true for that the whole ARC line from intel doesn't do dithering ?

    I am also interested in trying out the Intel Arc. However, I find the term 'by default' a bit confusing. In my experience, we often have more issues with drivers than with the hardware itself. Even if dithering is enabled/disabled on hardware by default, it might be possible to change this setting through the driver. This raises a question about Intel's response: are they considering the role of the driver in their answer, or are they only referring to the default settings of uninitialized hardware?

      autobot

      This is indeed the case for my pc: i5-12450H , Intel UHD Graphics Xe (driver v. 31.0.101.5537), Windows 11 23H2, 8-bit monitor.

      Ditherig.exe does nothing for my graphics card. The writing to the register is not performed because there is no dithering. It seems that dithering is disabled by default, and the driver doesn't enable it in my case.

      Out of curiosity, I added logging to ditherig app (https://ledstrain.org/d/1001-ditherig-is-open-source-on-github).

      Here is an explanation of how to read the log:

      RegisterValue - the value read from the register.

      RegisterDataFromDB - bits that need to be set to 0 in the register to disable dithering. The writing is done using a mask to avoid overwriting bits that are not related to dithering.

      NewRegisterValue - the new calculated value to be written to the register in order to disable dithering; If it is equal to the value that was read from the register, then no writing occurs.

      The contents of the log file:

      Address=603c000000 RegisterAddress=70030 Address+RegisterAddress=603c070030 RegisterValue=4000000 RegisterDataFromDB=0 RegisterMaskFromDB=1c NewRegisterValue=4000000

      There is no need to update the register with the value 'NewRegisterValue' because it is already equal to 'RegisterValue'

      Address=603c000000 RegisterAddress=71030 Address+RegisterAddress=603c071030 RegisterValue=0 RegisterDataFromDB=0 RegisterMaskFromDB=1c NewRegisterValue=0

      There is no need to update the register with the value 'NewRegisterValue' because it is already equal to 'RegisterValue'

      Address=603c000000 RegisterAddress=72030 Address+RegisterAddress=603c072030 RegisterValue=0 RegisterDataFromDB=0 RegisterMaskFromDB=1c NewRegisterValue=0

      There is no need to update the register with the value 'NewRegisterValue' because it is already equal to 'RegisterValue'

      Address=603c000000 RegisterAddress=73030 Address+RegisterAddress=603c073030 RegisterValue=0 RegisterDataFromDB=0 RegisterMaskFromDB=1c NewRegisterValue=0

      There is no need to update the register with the value 'NewRegisterValue' because it is already equal to 'RegisterValue'

          When I enabled Auto Color Management (by default, ACM was disabled), I got the following entry in the log:

          Address = 603c000000 RegisterAddress = 70030 Address + RegisterAddress = 603c070030 RegisterValue = 4000110 RegisterDataFromDB = 0 RegisterMaskFromDB = 1c NewRegisterValue = 4000100
          We should update the register with the value 'NewRegisterValue' because it is not equal to 'RegisterValue'. The updating was successful.

          The register value 0x4000110 in binary representation is 0100000000000000000100010000. The fifth bit is set to 1. According to the ditherig database, this is Spatial dithering. Its value in the database is 0x00000010 (00000000000000000000000000010000), i.e., the very same fifth bit.

          14 days later

          reaganry

          It might depend on the vendor, but my Asrock ARC A770 doesn't dither by default in Windows 11 (if ACM is turned off).

            reaganry No good user reports about intel arc yet?

            my 128EU Arc iGPU was bad. Not sure, GPU or screen (or both) issue

                simplex when you turn of the GPU(i.e. choose basic vga driver in windows) does it solve/improve that?

                    autobot does it solve/improve that

                    nope

                    Even safe win10 1809 build didn't solve that. tbh laptop's screen could be reason but I sold laptop half-year ago, only now I understood issue could be in:

                    1) screen (noticable pixel-inversion especially in all new screens I tested)

                    2) motherboard/CPU/GPU (some users noted swithing motherboard solve eye-strain, some users told only CPU (amd 5600g to 5700g) switch give extra eye-strain, and I myself got strain using rtx20 and newer series over gtx10 line keeping exact same PC components)

                        I dont think an OS will do dither at software level, I only know some monitors will do dither at hardware level which hurt eyes.

                        simplex one possible way to do the testing is use a a safe live usb distro, for example lubuntu 18.04, and try at friends with pcs etc until finding some safe hardware.

                        It doesnt use the gpu driver though, but it's much better solving it starting from safe hardware/monitor.

                            autobot starting from safe hardware/monitor

                            I still haven't found a safe business laptop (16-18" with 300nit screen and 72% NTSC, not old amd/intel with iGPU, 16/32gb memory, 60+ watt battery)

                              I have a hypothesis regarding Ubuntu 18: the display might feel easier on the eyes because some new graphics cards aren't supported, resulting in an image that's less straining. If that's the case, we could blacklist the necessary kernel modules to achieve the same effect in newer versions of the OS.

                                autobot

                                I don't think it's specifically about the distribution itself. It's more about a set of useful hacks and configurations that might work for certain hardware.

                                  dev