Display port or HDMI ?

AshX

macsforme I am not too familiar with analyzing for dithering with a microscope and camera, although I see others here claim that this works. I would be concerned that regularly occurring camera sensor noise, image processing, and other physical factors such as pixel inversion, can introduce barriers. On the other hand, I have seen dithering so obvious that I could see it with the naked eye, and I did use a camera to record it in those cases.

From what I understand 240 fps is insufficient to detect dithering occurring past a certain frequency. It should be sufficient for 60Hz refresh rate screens, in theory, because they are limited to the amount of dither frames that can be introduced compared to say, 100Hz refresh rate monitors. 960 fps would be ideal - or even higher - but those cameras are tens of thousands of dollars.

I think we can discern camera noise from flicker at the lower frame rates. I’ve seen videos like this one taken at 240 fps where monitors go up to 100Hz+ and it becomes very difficult to discern, as you said. https://youtu.be/LZMWduGGJvk?si=JAqpWpTwOSQmKrBc

If we are going by the epileptic risk range then 30Hz and below can absolutely trigger seizures in sensitive people. Even though most of us do not have epilepsy, the same part of the brain is being affected so if we can identify the specific frequency dither is occurring at, then we should be able to reasonably map symptoms. The medical literature has mapped this out already, and AI Chatbots can collate that data from those studies and should be able to aid in mapping out what is occurring. It should also allow us to determine what cables limit what frequencies of dither; if any.

simplex

macsforme

Can you share gtx1080 results?

Regarding EDID, your capture card is able to edit EDID, or can you also share it there? Not sure but it could be visible in CRU -> monitor list. Capture card can set 4:2:2 and that is, you cannot capture 4:4:4 from 4:2:2 and get proper difference results

macsforme

simplex I apologize, it appears that I only saved the time blend renderings from the GTX 1080 captured by the Blackmagic device, and not the original captures. I will try to make a mental note to preserve some of them the next time I conduct experiments (the equipment is split between home and work). Here are the time blends (please download only what you need, as this connection is metered):

http://joshb.us/~josh/noshow/Time_Diff_Mac_Pro_1080_HDMI.mov.zip
http://joshb.us/~josh/noshow/Time_Diff_Mac_Pro_1080_DP_to_HDMI.mov.zip

I do have a few of the Epiphan DVI2PCIe raw captures (8-bit uncompressed RGB) from the GTX 1080:

http://joshb.us/~josh/noshow/GTX_1080_DVI_Windows.avi.zip
http://joshb.us/~josh/noshow/GTX_1080_DP_to_DVI_Adapter_Windows.avi.zip

Here is another Epiphan DVD2PCIe capture I sent to JTL for testing a while back, from a 2015 MacBook Pro with AMD dGPU. The Blackmagic capture (10-bit RGB) from this same laptop showed dithering via the ffmpeg time blend method, but I was not able to come up with a test method for this 8-bit RGB capture from the Epiphan device. I would be very curious to have this analyzed:

http://joshb.us/~josh/noshow/Epiphan_Capture_MBP_2015_15_dGPU.zip

Here is one of my earlier posts with the EDID info from the Blackmagic capture device. It appears to advertise both YCbCr 4:4:4 and 4:2:2. The capture software purports to allow you to capture RGB, so I'm not sure how to explain that discrepancy when RGB does not appear to be offered in the EDID. I believe I also tried creating an EDID override in macOS so the Blackmagic device would have the same effective EDID as one of my monitors, but I do not recall observing any difference in output.

simplex

macsforme

Why GTX_1080_DVI_Windows.avi is 29fps…? Or is it frames sequence that has no speed at all..?

Which tool I need to extract frames?

simplex

macsforme

GTX_1080_DVI_Windows.avi is 29fps, you know why? Could they be blended?

Images I extracted were equal in terms of color. I also found text has fringing ( green tint on the left and red tint on the right ) which is not captured in windows screenshots, it mean your record post-process very close to source which monitor get

simplex

macsforme

nvidia grapfic pipeline have multiple stages, and nvenc module is near the starting point of signal transformation ( otherwise all video rendered by NVIDIA cards would eye-strain users - in grapfic editors output, or YT/twitch streams )

Also windows use DXGI which is also near the start of transformation

Could this be issue for getin bad ( no symptoms ) capture?

JTL

simplex Which tool I need to extract frames?

If you have ffmpeg, you can use it to extract frames without quality loss.

ffmpeg -i $file -r 1/1 $output/%03d.tiff

$file should be the input file
$output should be the output directory

macsforme

simplex To clarify, I would not say that the GTX 1080 causes me symptoms, at least not severely like with certain Macs. I have it connected via DisplayPort to a 120Hz monitor, so while there is likely dithering (DP connection) the high refresh rate likely helps. I use it mostly for occasional gaming, where dithering probably is less of a factor.

I believe the framerate was the maximum supported by the Epiphan DVI2PCIe capture card for 8-bit RGB uncompressed. Beyond that, I don’t know. The Blackmagic capture device supported a higher framerate I believe.

You mentioned no color changes between frames for the DVI output capture. In my experience, DVI and HDMI do not dither on NVIDIA, while DisplayPort does. Did you (or can you) evaluate the DP output capture to see if you make a different finding?

simplex

macsforme

Okay

Its interesting that 1080 doesnt bother you. Last month I tried 1080ti and sympthoms were similar to 2070s ( not calm as 1060 )

macsforme

simplex I am curious whether you were able to analyze the DisplayPort capture I posted at all? In my experience, DP output is what dithers on NVIDIA (on the cards I tested).

simplex

macsforme

I analyzed DP of gtx1080 and found nothing criminal, "clear" as HDMI

Do you remember, OS grapfic settings also were 30hz for DVI2PCI card, or it drop each 2nd frame?

macsforme

simplex Do you remember, OS grapfic settings also were 30hz for DVI2PCI card, or it drop each 2nd frame?

That is a good question, but I do not recall. If only every other frame was being captured, I suppose that would probably not be a great way to test for dithering. If I repeat the test, I will try to be mindful of that.

Thanks for looking into it.

Aquila

Any update on 2025

« Previous Page