I have bought a capture card

macsforme I plan to post my findings soon

The results of my investigation into temporal dithering on Intel-based macOS are as follows. This testing pertained to external outputs only (not built-in laptop displays) given that a capture card was used. The machines tested included a 2009 Mac Pro, 2015 15-inch Retina MacBook Pro (iGPU only and iGPU/AMD dGPU variants), 2012 13-inch MacBook Pro (non-Retina), and 2009 15-inch MacBook Pro. The GPUs tested in the Mac Pro included the original NVIDIA GT 120, an NVIDIA GTX 640, an NVIDIA GTX TITAN (Kepler), an NVIDIA GTX 1080 FE, and an AMD Radeon Pro WX 5100.

Notably, the macOS version and firmware version of each machine did not appear to make any difference in the outcome, in the cases where several were tested. Most of the HDMI output testing was done with a standard HDMI cable, but I repeated several of the tests with a high-speed HDMI cable, and it seemed to make no difference on my hardware.

2009 Mac Pro - All NVIDIA cards do have temporal dithering on DisplayPort outputs, and do not have temporal dithering on HDMI outputs. The AMD WX 5100 has temporal dithering on its DisplayPort outputs (which is the only type of output), but only when plugged in at boot time (not after hot-plugging the monitor). Having multiple monitors plugged in and hot-plugging them seemed to cause inconsistent results as far as dithering.

2009 MacBook Pro & 2012 MacBook Pro - The former has an NVIDIA dGPU while the latter has an Intel iGPU; on mini-DisplayPort outputs, no temporal dithering was detected.

2015 15-inch Retina MacBook Pro - The iGPU-only model (MacBookPro11,4) does not exhibit temporal dithering on either mini-DisplayPort nor HDMI outputs. The model with the AMD dGPU (MacBookPro11,5) does exhibit temporal dithering on both mini-DisplayPort and HDMI outputs.

macsforme The model with the AMD dGPU (MacBookPro11,5) does exhibit temporal dithering on both mini-DisplayPort and HDMI outputs.

Does this change if you try the Psychtoolbox kext + Octave commands dither disable method I posted? I've already confirmed this method affects flicker systemwide as I can see noticeable small changes to colors and reduced flicker on an old TN monitor I tried it on. (Unfortunately that monitor also has its own FRC on 8bit signals and macOS seems to not be capable of forcing a 6bit connection)

So I'm curious if the PTB kext method actually stops the "GPU-level" temporal dithering entirely or if it only "reduces" it, before I go and try to search for a true 8bit monitor

DisplaysShouldNotBeTVs Does this change if you try the Psychtoolbox kext + Octave commands dither disable method I posted?

I have not yet had a chance to experiment with that method, but I am curious about that as well. I hope to look into it at some point so we'll have another empirical data point.

jordan Btw in reply to a comment above. if you want that dvi2pcie cheap no need for the duo. Just get this one, it's missing the bracket but I'll also link one that works. Me and another user here have ordered from both of these links.

Thanks for the suggestion. I picked up one of these capture cards and set it up over the weekend.

Is anyone with one of these cards willing to share their method for detecting dithering with it? Is the "videodiff" program used for this? I tried with the ffmpeg time blend method, but so far I could not detect dithering after trying several configurations (including the 2015 15-inch Retina MacBook Pro with AMD dGPU, which did exhibit dithering on the BlackMagic capture card). The ffmpeg time blend command appears set up for 10-bit video, whereas the Epiphan DVI2PCIe card seems to capture 8-bit video, so that could be the difference or simply that DVI output does not dither on these cards.

I have discovered that UEFI has its own video driver configuration - called UEFI GOP. If I understand the idea correctly, UEFI GOP works based on the EDID data of the monitor, imagine if monitor is 8 bit + FRC, you disable 10 bit in windows / linux to get pure 8 bit signal, and UEFI continues to scale it to 10 bit at its own level.

What if UEFI GOP behavior of motherboards, someone call here "non-disabled dithering" of video cards?

DisplaysShouldNotBeTVs I would suggest dumping a realistic EDID from an actual monitor (using BetterDisplay's save EDID function etc), then using one of the "force EDID" methods in order to tell macOS to substitute in the capture card's EDID with the monitor's one instead. then see if the output changes.

Actually. I wouldn't trust forcing the EDID on the target device. Forcing EDID can have different functionality compared to native EDID detection.

The DVI2PCIe is one of the few capture cards that can be programmed to output an arbitrary EDID.

Good points. I understand the goal of replicating the exact configuration of a real display for accurate testing/comparison. My concern with programming an arbitrary EDID would be that the EDID advertises the capabilities of the display (or capture card), so I could end up with a video format that the capture card is not actually compatible with.

Epiphan’s web site has EDID files you can download to force certain resolutions if you don’t otherwise have a way to do so. From what I gather, this seems to be the main purpose of the EDID upload feature.

macsforme Good points. I understand the goal of replicating the exact configuration of a real display for accurate testing/comparison. My concern with programming an arbitrary EDID would be that the EDID advertises the capabilities of the display (or capture card), so I could end up with a video format that the capture card is not actually compatible with.

I think that concern is partially unfounded. As long as the color format being output by the target device is supported by the capture card it should work. And if not, just reprogram the "stock" EDID and you're no worse than what you started with.

I tried with the ffmpeg time blend method, but so far I could not detect dithering after trying several configurations (including the 2015 15-inch Retina MacBook Pro with AMD dGPU, which did exhibit dithering on the BlackMagic capture card).

If you send me some files and information in PM. I might be able to help.

macsforme All NVIDIA cards do have temporal dithering on DisplayPort outputs, and do not have temporal dithering on HDMI outputs.

macsforme The iGPU-only model (MacBookPro11,4) does not exhibit temporal dithering on either mini-DisplayPort nor HDMI outputs.

After further reflection, the before-and-after of my vision strain is starting to make sense (if temporal dithering is the cause), given my finding that GPUs I tested tended to dither on DisplayPort outputs only and not on HDMI (and I assume not on DVI).

Years ago I ran dual 8-bit (purportedly) displays on a 2014 Mac Mini, which (if consistent with my iGPU MacBook Pro) would not have had dithering on the Intel iGPU outputs. I later upgraded to a Mac Pro (with various GPUs over time) and a 3-monitor setup, but my standard refresh rate monitors (same purportedly 8-bit ones) were on either HDMI or DVI, while the DP output went to a high-refresh gaming monitor (likely reducing the effects of dithering on that output).

Later, after ditching the horrible 14" M1 Pro MacBook Pro, I ran my Mac Pro with an AMD Radeon Pro WX 5100 (4x DP outputs) with DisplayPort going to all three monitors (including the standard refresh rate ones). I also tried some of my NVIDIA cards using the DisplayPort outputs. After that, I never had 100% of the comfort level I had previously.

For the last few weeks, I've been using one of my original ASUS VS239H(-P, I think) monitors with a 2012 non-retina MacBook Pro with Intel HD 4000 graphics, running a fairly clean install of the latest Ventura. So far this has been giving me no issues (so far as I can tell… there is still some unavoidable strain from other screens at work and LED lighting). I am tempted to try going back to a 2014 Mac Mini with an Intel iGPU, or just keep running one of my MacBooks with Intel iGPUs in clamshell mode.

Interestingly, I just booted into my main data drive running Monterey (which was a migration from my 14" M1 Pro MacBook Pro originally) for the first time in about a month and a half, using roughly the same hardware, and I could instantly tell there was some kind of strain there. So, either I screwed up some macOS graphics settings on the 14" M1 Pro MacBook Pro while trying to stop the strain (and those settings migrated over), or else there was something inherent in the Monterey setup on that Apple silicon machine that came over with the migrated settings. This is consistent with my observation in another thread:

macsforme When I happened to boot another disk with a clean Monterey installation on that same computer, it looked "calmer" than my main Monterey environment, making me wonder if some strain-inducing setting(s) were migrated back with my data (although plausibly, my condition could just be deteriorating).

macsforme just curious did that also apply to HDMI being dither free while dp dithered for the gtx 1080?

jordan I wonder if it's your cable that is triggering it due to lack of bandwidth maybe ?

This is quite possible, although I'm not sure that the math backs up this theory. The BlackMagic capture device itself is HDMI 2.0, and my DisplayPort to HDMI active cable is supposedly DisplayPort 1.4 (but the supported resolutions on the product page suggest that HDMI 2.0 is the bottleneck). For what it's worth, I did repeat most of the tests with an "Ultra High Speed HDMI" (48Gbps) cable, with identical results. The GTX 1080 also has HDMI 2.0, so with both ends supporting HDMI 2.0 as well as the cable, that should not be a bottleneck. I'd like to determine what the actual negotiated bandwidth is in macOS, but in any case this seems like plenty of margin for 1080p 60Hz (even with 10-bit color). The fact that we have to (actively) convert DisplayPort to HDMI or DVI (which appear to be very different signals) introduces a complexity that I wish we could avoid.

jordan I know someone for example did test dp off the quadro rtx 4000 (turing)gpu with the epiphan and he confirmed no dithering(I have the vid).

I am still trying to figure out how to identify dithering in an Epiphan capture. However, DVI and HDMI seem to be similar signals, and I only detected dithering over HDMI (with the Blackmagic device) on one machine to begin with, the MacBook Pro 15-inch Retina with AMD dGPU.

jordan Do you plan to try more gpus?

I am open to it. My primary goal was to find a Mac-compatible GPU with a DisplayPort output that doesn't dither, but so far I have not found one. I do have some interest in finding a safe GPU on the Windows side to use at work, however.

macsforme I am still trying to figure out how to identify dithering in an Epiphan capture.

What aspect do you need help with exactly?

macsforme If you can find some way to send me the files and PM me I can investigate.