ditherig.exe on Windows 11 24H2

Earlier, I conducted research related to detecting spatial dithering in Windows 10 22H2. The application ditherig was used to control dithering. The results showed that the ditherig application works properly when used with the ARC A770 in Windows 22H2.

Today, I tested the functionality of ditherig in Windows 11 24H2. The test system was an i5-12450H (Intel UHD 48EUs iGPU).

Screenshots were taken using a Blackmagic UltraStudio Recorder 3G:

1. without using the ditherig application;
2. with ditherig installed (dithering disabled) (uploaded to another hosting service because postimg.cc replaces this file with the file from the previous numbered item due to it having the exact same checksum, as both files are pixel-by-pixel identical);
3. with ditherig installed (spatial dithering enabled).

Now let's compare the screenshots:

1. "without using the ditherig application" vs "with ditherig installed (dithering disabled)": the images are identical, pixel by pixel;
2. "with ditherig installed (dithering disabled)" vs "with ditherig installed (spatial dithering enabled)": a characteristic checkerboard pixel pattern can be seen in certain areas, similar to what I described in my post about spatial dithering in Windows 10 22H2. The red dots represent pixels that were altered by the spatial dithering algorithm when compared to the image where dithering was disabled.
3. By default, dithering is not enabled on my iGPU, so temporal dithering is also absent. However, temporal dithering can be enabled using ditherig. Analyzing the recorded video (a photo of the screen, static without movement) shows that the option to enable temporal dithering works as well. And disabling it, accordingly, works too. The video is very large (uncompressed 10-bit stream), so I didn't upload it. But if needed, I can compress and upload it. To determine temporal dithering, the method proposed by aiaf in their post was used.

It demonstrates that the ditherig application functions properly in Windows 11 24H2.

In my case, hardware dithering was completely disabled on the iGPU by default, so there’s no need for me to use the ditherig application. However, your situation might differ; it’s also possible that on some configurations, ditherig may not work correctly. But at least on both of my PCs (ARC A770 and Intel UHD 48EUs), it performed as expected.

simplex

Have you tried same with arc750?

Here I tested the ASRock ARC A770 on Windows 10 22H2. The behavior is identical to that of Intel UHD 48EUs.

Btw, if arc750 and your Intel UHD doesnt have activated dithering by default, what could be the issue with arc750…?

Unfortunately, I don't have a definitive answer to this question.

To me, the default image from the ARC A770 appears sharper and more contrasty compared to the Intel UHD 48EUs on any operating system. My eyes tend to prefer a slightly softer image.

When comparing Windows 11 on the PC (ARC A770) to Windows 11 on MBP M1, the first thing I notice with the Apple Silicon is a much flatter image - less contrasty and less sharp. This provides my eyes with maximum comfort. It seems that certain aspects of image processing in Windows 11 do not function properly when the OS is running on Parallels on an Apple Silicon Mac. Either the Apple Silicon M1 graphics card processes the image in a way that sharpness and contrast are not boosted.

As a result, I think this difference might be due to variations in the processing algorithms applied to the image (including gamma/color/etc correction). These could stem from the GPU hardware pipeline itself or differences in the software.

In my case, dithering is disabled by default, but with a graphics card from another vendor, the behavior might be different. Theoretically, there could also be hidden hardware defects resulting in jitter or high-frequency flickering.

We still know very little about how it all works internally and what the mechanisms of its effect on our eyes are.

DisplaysShouldNotBeTVs

I did this on Windows 10 22H2 with Intel UHD 48EUs and the Intel ARC A770.

The difference in the number of pixels with varying shades is pretty noticeable, even when using the same driver version. Since both the Intel UHD 48EUs and Intel ARC A770 are built on the same Xe architecture, they can use the same Intel UHD driver. However, I didn’t dive into comparing the exact pixel values themselves.

I’m currently looking for an application that can display two images side by side or overlaid on top of each other. Ideally, it would let me, when hovering the mouse cursor over a pixel in one image, automatically receive information about that pixel in the second image:

1. Check the pixel values for both images.

2. See how significant the differences are.

3. Show the actual pixel shades somewhere on the info panel.

It’d also be great if the app allowed zooming in on the images and switching between them with a key press (in case they are overlaid on top of each other) to make evaluating the differences easier.

If I can’t find an app like this, I’ll probably just write one myself. Not sure exactly when I’ll get around to it. Since my experiments with the iMac 27 2015 were unsuccessful (it went out of stock in my region), I have some free time, and I might write this application this month.

jordan

I ran Windows 11 in Parallels (a virtual machine).

do you have to do anything to disable the dithering on win11 with apple silicon?

I didn't do anything. Temporal dithering is 100% absent (I checked). It also seems that GPU emulation in Parallels does not involve spatial dithering. My display is 6-bit+FRC, and spatial dithering enhances the FRC module. If it were present, working on this setup would not be comfortable.

jordan

I thought it would still have dithering since its mac os with the win in vm.

Dithering, which strains the eyes, always comes from the graphics card. In this case, it is enough to use StillColor to disable dithering on the Apple Silicon.

would you say windows on that is the most comfortable?

In my case, it is no different from macOS. Working in both operating systems is comfortable. In this case, the differences include a different graphics card (i.e., not Intel, Nvidia, or AMD) and possibly something else that we do not yet know about.

is windows able to be installed as the main os?

Windows 11 can currently only be run in Parallels; full installation is not supported. I use Windows 11 to run a few utilities that I need for work. Otherwise, its use is debatable, as the virtual machine does not provide the required level of performance, and Windows 11 has an ARM architecture, which also imposes certain limitations. That is, it is not a replacement for a PC.

im assuming this is only for external screen not internal mba screen right?

I use an external monitor because I can't work with MacBook Air/Pro screens.

WhisperingWind

(Do you think this is because running the windows 11 VM is activating the Metal direct display mode and bypassing processing since it's a fullscreen window?)

IMO your experience here is very strange to me. As I've never been able to get M1 macOS (or even Intel macOS for that matter!) to work with any external monitor that's fine on a known good setup

(an example of a known good setup for me is my current solution: Raspberry Pi 2011 Model B, 1st generation with acceleration turned off + the older dispmanx display driver + showing the same macOS or Windows desktop through VNC)

I have an M1 Air which should have the same GPU as your M1, but when displaying the same screenshot between both the Mac connected to a monitor, and the Raspberry Pi Gen 1 ---

the Mac creates extremely strange artifacts on all monitors I've connected it to, everything (in the same screenshot that looks fine when output from the Pi Gen 1) looks either extremely oversharpened or too blurry when output from the Mac. I notice additional color fringing and blue glows around larger objects. All photos seem to have some sort of contrast enhancement or antialiasing effect where smaller objects / AKA "objects in the background of the photo" look more blurry or smudged than they should.

This effect is definitely caused by the M1's GPU, or HDMI controller, or possibly the macOS window server (but at a stage after the plain, unprocessed framebuffer is rendered, meaning that it's not captured by a generic screenshot)

Because: The same photos, displayed on the same macOS desktop, but captured as a screenshot instead (or shown through VNC) and displayed on the Pi Gen 1 connected to the monitor (instead of from the Mac's HDMI output) look fine.

This remains true even with Stillcolor and the Force RGB EDID workaround.