U3223QE, Nvidia, problems with MST and HDR

Question

Hi,

My new U3223QE works perfectly well on Intel iGPUs (both Alder Lake UHD 770 Graphics and Comet Lake UHD Graphics). MST and HDR both work as expected.

But on Nvidia Ampere (RTX 30 series) I have observed the following:

HDR: looks severely broken. Black levels are raised considerably. For example in Netflix one can see an obvious difference between the "black" included in the video and the black background of the app website. Many videos show blocking in dark areas, because they are not shown as dark as they should be.

I own other HDR capable monitors like the AW3821DW, U2520D that do not exhibit this behavior. Their HDR and black levels look identical between Intel and Nvidia graphics.

I have already tried everything I could come up with, including different driver versions, DDUing the graphics-driver, different color-depth / mode options (8-Bit with dithering, YCbCr, Chroma-Subsampling, limited output range), nothing changes it. The picture looks extremely bad, until I turn off HDR on my Nvidia graphics card, while on Intel graphics I can hardly notice the difference between HDR and SDR, like I hoped with such a monitor.

I should also mention that I am using the same desktop PC for this, it has the iGPU and dGPU active at the same time so I can easily compare.

MST: When enabling MST in the monitor, the monitor itself will remain inactive on the Nvidia graphics. Windows reports it as active, MST-Out can be used to attach different displays that will also work. But the U3223QE itself will show "No input" and remain inactive. Even if I detach anything from the MST-Out it will remain inactive, while it will work just as expected on my Intel graphics devices.

Sadly, MST cannot easily be turned off in this state, because the monitor will not allow changing this option back to disabled until it is showing some input from any source. I had to resolve to factory resetting or using a different PC to get it to turn on.

Here I have at least found that Nvidia driver versions prior to 497.29 (including 497.29, 496.13, 472.12, 471.41, 456.71) have this working. But 511.23, 511.65 and 511.72 break in this way. Sometimes a simple downgrade just using an older installer worked, sometimes DDU was necessary to get it working again after having upgraded to a broken driver.

Other MST-capable devices such as the WD19TB or U2520D were not affected by this bug, so it seems to be something specific to the U3223QE (and maybe U2723QE).

Has anybody else experienced something similar?

While the MST thing is annoying, I have some hope that it will get fixed again by Nvidia, as it has not been broken for that long and I can stay on the older drivers for the time being.

But the HDR thing is very infuriating, because HDR is basically unusable in this state. I have not found a sensible workaround (using 2 GPUs simultaneously works, but draws much more power and causes some stuttering).

And I am unsure, whether the U3223QW has just broken HDR, that only works on Intel GPUs or whether this is also a bug with Nvidia, that might get fixed with a future driver update.

Anybody using this monitor on current AMD graphics? If it works on 2 out of 3 manufacturers GPUs, I'd be more confident, that this is not a design error with the monitor.

Ray519 · Answer

When retrying with the newest 511.79 driver, I saw correct HDR black levels once and looked into it further.

I was able to make it work correctly by making the U3223QE the primary monitor and disconnecting all others. I toggled HDR off and on again and it started working and has done so, even after re-adding my other monitors and switching back to my original main monitor.

The MST-Issue remains still broken though. But I was able to confirm that a GTX 1050 was not affected with the same driver version.

My guess is, that it only happens with DSC enabled, which the GTX 1050 cannot do. Nvidia also supports a newer version of DSC (1.2a) instead of DSC 1.1 still in use by Intel, so the problem might have something to do with that.

csandras · Answer

I believe the U3223QE works differently or has a different interface/electronics than U2723QE.

I have both the 27 and 32 version on my desk, 27" (Rev A00, Mfg Nov-2021) arrived and being used sine 2 weeks, 32" just arrived (same Rev A00, Mfg Nov-2021, Fw M3T101).

27" was producing perfect HDR since the first time I connected.

32" is doing me this broken blocky black image, exactly as you described. It looks like the monitor try to increase the black level in a way it converts colours close to black to full black and it does so in big squares altogether for some reason. This makes HDR unusable on the 32" version. And this is not solved for me after a reboot on the Mac and reset of the monitor.

Same computer, same cable, same movie, same player. MacBook Pro 2018 with Intel Iris Plus 655, no dedicated GPU. 4k/60Hz. Monitors are used with exact same setup in the OSD and colour profiles are installed/applied in OS.

The funny think is that this is not happening in non-HDR mode, so there is something to do with the HDR implementation. Also note that 27" has quite neutral colors in HDR - I could use it for generic use - while 32" has slightly exaggerated colors. Both has perfect colors in non-HDR mode though.

Ray519 · Answer

SDR has also been perfect on every system so far. Firmware is the same for me.

In the meantime I had my monitor working correctly in HDR with my Nvidia GPU for a few hours. It started after the U3223QE being the only monitor attached and a reboot, but only lasted for so long until it reverted back to the broken state.

This makes it look like a driver / implementation problem with Nvidias driver. Similar problems are described in Nvidias forums, together with the reboot workaround, that sometimes works for a while.

Also the effect, where dark colors seemingly get rounded to different colors also occurs on my Notebook (Dell XPS 15 9500), when I use the "Dell Premium Color" App to force the display to a more sensible color space for windows (sRGB, AdobeRGB). It looks quite similar and lasts until I reset the GPU color settings. From this I am assuming, that the conversion between different color spaces (SDR to HDR for the U32) seems to be the problem and might not have enough precision.

I would like to check something with you, especially if you have access to both U2732QE and U3223QE: in the EDID-Info of my Display are

 Color Characteristics:
    Red  : 0.6767, 0.3144
    Green: 0.2724, 0.6757
    Blue : 0.1445, 0.0576
    White: 0.3134, 0.3291

These are distinct for all my different displays (all different models). If (I am guessing here) these values factor into the color space conversion done for the monitor, then subtle differences here could explain why some displays might be more problematic than others.

I would like to compare these values to both your displays, to check, whether they are unique to each display or just unique to each model of display (where a difference in panel could explain different values).

I extracted the raw EDID infos and decoded them with the linux tool 'edid-decode', but other tools to do the same should also exist.

All the other details in EDID, like HDR Static Meta Data etc are pretty much identical to my AW3821DW and U2520D (which has much worse HDR), but both work reliably in HDR on Intel and Nvidia graphics.

Under windows one can manipulate the EDID with tools such as CRU, so if we identify a difference I can try, whether this gets my U3223 working more reliably (although one can never be entirely sure, whether a change is just ignored or does not make a difference).

csandras · Answer

Hello Ray, I have already returned the 27" as that one had a bad pixel... I was try to extract the EDID for the 32" but could not complete it and just have no more time now.

What I got is the hex EDID data, you might be able to decode it.

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

Different SDR color modes are working just fine for me, I was try DCI-P3, SRGB and both are nice with no artefacts and great color.

Even though both displays has/had the worse backlight bleed/IPS glow I have ever seen (and I am using IPS only since 20 years or so) I think I will go back to 27" (a new one) as the contrast/black level is so amazing in normal office usage that it worth it... and HDR is just working on that. I also use LG 27UL850 and were using EIZO EV2785 recently and yet to see another monitor with this black level... Nevertheless this monitor with this panel should not wear the UltraSharp title...

Andras

Ray519 · Answer

I could decode your complete EDID data and the color characteristics are identical to mine, so looks at the very least like a per-model config.

If those values actually play a role in HDR conversion, it would fit if all the U32 all have the same values and U27 have different ones.

Backlight bleed has been pretty much the same as my U2520D, which is far from perfect but I only notice it when explicitly looking for it in a dark room... But the AW3821 shows how much better an IPS monitor can be at this and there is probably still room to improve upon it.

Ray519 · Answer

Ok, so with Nvidia driver version 512.15 the Issue with enabling MST has been resolved. The broken HDR black levels on Nvidia still remain.

csandras · Answer

Attached a few photos I took earlier to illustrate the HDR difference between 27" and 32" displays. Both were connected to the same MBP on USB-C, both were switched to Movie HDR, color profile applied and displays were switched to hardware mirroring. Movie playing is Rome series, 1080p, BlueRay. Exact same settings on both displays, including same contrast. Color tone is also very different.

27 U3223QE U3223QE27 U3223QE U3223QE

csandras · Answer

For some reason I cannot upload more photos now, so will upload some more later.

Ray519 · Answer

Ok, that looks precisely like what I saw and still see most of the times on Nvidia drivers with HDR enabled, but never once on Intel GPUs. I am so curious to find out what is wrong here. But given Dells communication with me, especially about software issues, I won't hold my breath.

csandras · Answer

I do not think this is chipset related as I use Intel and not nVidia.

csandras · Answer

273227322732

DmitryP · Answer

The following topic explains some of the HDR problems on U2723QE/U3223QE: https://www.dell.com/community/Monitors/U2723QE-Smart-HDR-is-useless-b-c-of-the-dimming-implementation/m-p/8175482

U3223QE has also a flaw in HDR colors that is not reproducible on U2723QE.

Ray519 · Answer

Since it went away for a short time on Nvidia when being the only attached display and has absolutely reliably worked on Intel, it basically has to be a software not hardware issue.

The question is just, whether the monitor reports wrong things to the GPU and some drivers cope better than others or whether it is purely on the GPU/Driver not supporting standardized things correctly.

If I knew it was the display, I would return it immediately. But the other Nvidia driver issue was fixed by Nvidia, so I still have some hope.

Ray519 · Answer

It does not, please look at the pictures csandras was kind enough to provide, especially the first 2. This is anything but a local dimming issue.

And although I have not counted the dimming zones one by one on my U3223QE, it is on par with my U2520D in that regard. Sure it could be much better, but Dell did not advertise the amount of local dimming zones and DisplayHDR400 is specced to be possible without ANY local dimming. I did not expect much more here.

The remaining issue here is clearly a color mapping bug that affects at least Nvidia and Apples drivers for Intel GPUs, but somehow not Intels drivers for Windows.

If you do not have those issues, please report the GPUs and drivers with which you are having success, maybe this allows us to narrow down who to blame for the HDR issues discussed here. I am particular interested in AMD GPUs, as we still know nothing of their compatibility.

Wasabisushi · Answer

Those blocks in the dark areas look also like compression artefacts. Compression would be aggressive in very dark and usually undistinguishable tones. Especially the smooth area around the text looks blocky and rasterized. I have the same experience with my U3223QE on M1 Mac.

The workaround I found is to turn on HDR400 on the display and leave it disabled on the os.
I was just playing around a little with that stuff. I don't actually use it because you loose brightness settings, it is uncalibratable and "HDR"400 is a lie. Still love the contrast in SDR.

Monitors

U3223QE, Nvidia, problems with MST and HDR

Was this post helpful?