I disabled dithering on Apple silicon + Introducing Stillcolor macOS M1/M2/M3

DisplaysShouldNotBeTVs Interesting. Comparing now. So apparently you have AppleSummitLCD, whatever that is. It's not there on my M1 Max. The only reference I found to it was here What happens when an M1 Mac starts up? – The Eclectic Light Company, but I didn't google much. Some H7DisplayPipe as well. And some references to debug device.

com.apple.driver.AppleMobileDispH14G-DCP vs com.apple.driver.AppleMobileDispT600X-DCP

It would be really interesting to do the same on an M2 Touch Bar with a bad screen.

jordan

So I am testing a refurbished 13 inch M2 MBP with Touch Bar as mentioned by DisplaysShouldNotBeTVs. I disabled the oled touch bar which indeed flickers under my camera. Edit: While I did not detect any PWM on the backlight at 240fps or at 1/24000 shutter speed photos the opple shows PWM flicker at lower brightness settings (see below and PS). More testing on that later with oscilloscope waveforms.

Under the microscope, I'm surprised to find little pixel flicker. It's still there but if this is in fact pixel inversion or some kind of FRC, it is minor compared to the 7i at 60hz. Edit: I am always running Stillcolor with dithering disabled and uniformity2d disabled. It has comparable low levels of pixel level micro flicker to my T480s panel also running at 60hz. It is still too early for me to rank which is most comfortable (7i, T480s, or this particular MBP).

Update! FWIW, Opple waveform charts at 100%, 50%, and 25% panel brightness. At lower panel brightness their is more fluctuations unfortunately but still very good results at brighter settings. I would personally avoid 50% and lower brightness settings.

https://ibb.co/Z1CSt3k (100% full brightness)
https://ibb.co/pxqJ3GR (50%)
https://ibb.co/Btd5RW3 (25%)

Opple reports the PWM frequency to range from about 41khz to 24khz and reports "no risk" at the three tested brightness settings. The charts:

https://ibb.co/w6D7ZsF (100% full brightness)
https://ibb.co/jZPvCFz (50%)
https://ibb.co/865cMtP (25%)

Tested on white background for showing backlight PWM not temporal dithering.

Thank you DisplaysShouldNotBeTVs for discovering this "hidden" model and sharing!

PS: I should have noted to not take these absolute PWM frequency numbers too seriously from a measurement device like the Opple. See: https://ledstrain.org/d/2686-i-disabled-dithering-on-apple-silicon-introducing-stillcolor-macos-m1m2m3/741

madmozg

I already did (see updated post) and the opple seems to show significant "modulation depth" and flicker. So for myself, I would use this close to or at full brightness in a brightly lit ambient environment.

async I'm 90% sure that I could do 6-bit with AW EDID Editor some time ago, however when trying recently the OS seemed to ignore it, like it does with certain things. I use this for 10 to 8-bit successfully. Figured it was either due to Sonoma upgrade, or that it simply didn't renegotiate or something.

See my post here for my own recent experiment involving an Intel Mac Pro, macOS High Sierra, an NVIDIA GPU, and a 6-bit+FRC display. While subjective, I definitely felt a difference (worse flickering, unstable image) when overriding the display EDID from 8-bit color depth to 6-bit color depth. Furthermore, other than while on the lock screen (and only on one display… weird), the effective depth did not appear to be reduced, so I figured all this did was traded the display's FRC dithering for temporal dithering by the operating system or GPU.

EDIT: Sorry, I just saw that you replied to that post, and therefore already saw it. Interesting that our experiences differ, though.

MBP mini led is still a shit show. Tinkered a ton with it today. Especially when increasing the brightness above stock and changing some flags you can both see the individual backlight leds turn on and off in a pattern, as well as color artifacts to the side of them. There is also this banded blob that changes size around the bright cursor based on how close it is to the leds, so while moving the cursor it keeps scaling up and down, and that part is significantly delayed compared to the leds.

Also it's almost impossible to make sharp contrasted areas like black and white feel solid. After trying a bit I actually think I got a wavy pattern that matched the placements of the backlight leds in my peripheral vision.

Most scrolling gives significant flickering as well. So far only Edge seems to handle it ok. Most measurements done are without any movement, but it is just as important how it behaves when you actually use the machine.

Obviously this is less of an issue when not using extreme adjustments, but it still proves that there are a lot of things going on that manipulates the image far away from the actually moving pixels.

Tried to capture some videos of this, but I have to find somewhere to upload it.

One of the issues is similar to this: Samsung Odyssey G9 Backlight bleed (youtube.com)

simplex

Hmm... it is a fixed 60hz refresh rate screen correct for this model (no ProMotion, no miniLEDs)? Normally, you would think that refresh rate has nothing to do with the flicker of the backlight LEDs. I also film at 240fps (not with microscope) and don't see any backlight flickering. I agree on the noise as factor in general with any measurement. I also agree to not pay too much to the frequency numbers from the Opple and simply just get a rough sense of the waveform and how that might correlate with subjective comfort. Stay tuned for results with the oscilloscope and photodetector.

async It would be really interesting to do the same on an M2 Touch Bar with a bad screen.

There's not really a M2 Touch Bar with a truly "bad screen" / unusable screen IMO, since even my first one with the worse panel was still more usable than every other Apple Silicon Mac. It was more of just visible defects, yellow tint, and screen uniformity issues which made me return that one.

(The "worse" panel had long empty sections of 00000000 in the panel ID and serial number dump, the "amazing" panel has a fully filled out serial number without any empty sections.)

However, since I still have the M1 Air for a little bit longer, I will dump ioreg on there — since I consider the M1 Air very uncomfortable (relative to any M2 Touch Bar Pro).

The M1 Air's panel is a lot different, flickers on camera (in the same way the M2 Air does) and is way worse.

Will be interesting to see if the ioreg has any significant difference.

photon78s Edit: While I did not detect any PWM on the backlight at 240fps or at 1/24000 shutter speed photos the opple shows PWM flicker at lower brightness settings (see below). More testing on that later with oscilloscope waveforms.

Yep this lines up with the notebookcheck review (which is good to know, since sometimes notebookcheck can be really inaccurate)

When I say the M2 TB Pro screen is "flicker-free" I mean it's possible to make it flicker-free, i.e. at higher brightness levels.

Also, I actually did use the computer for a while at lower brightness levels yesterday, and whatever high frequency PWM is there still does not affect me nearly as much as the really annoying low frequency PWM-like flicker on M1 Air and M2 Air.

However, PWM can still be avoided at high brightness!

(Lots of people think that is true with other Macs too, but two 2015 Intel Macs I tested actually also show PWM-like flicker on dark grays in 240hz slow-motion even at max brightness!

M2 Touch Bar Pro is the only MacBook I've tested that just doesn't have this issue at all. This is what makes it so unique!)

And with Stillcolor, it's possible to dim the screen through color overlays or color table adjustments (e.g. with BetterDisplay) without introducing temporal dithering.

photon78s Can you upload a video of the "pixel flicker" you captured vs. the other laptops you recorded that have much worse pixel flicker?