MadVR in use with LG OLED Thread

[SamuriHL]

I calibrate mine in RGB using Colourspace. I don't think I played with YCbCr 444 limited when I did PC mode but I can take a look at that. You'd think I'd have little to no banding with RGB given that's how it was calibrated. Maybe I'll do a fresh calibration in PC mode when I get a free minute and see if that improves it. I have no way to bit accurately calibrate in YCbCr.

[VBB]

(Part of) the reason YCbCr 4:4:4 has less banding is because Nvidia dithers it by default. RGB is not dithered by default, but if you dither it with, for example, Vince's tool, the result is pretty much the same as YCbCr (you can also remove the default dithering from YCbCr to get it to look like RGB). And the fact that a 10-bit input has more banding in PC mode seemingly has to do with PC mode switching the TV into 8-bit mode. Why LG decided to do it this way and to this day is anyone's guess. Most likely to save on processing.

SamuriHL, did you change to black level LOW and madVR set to 8-bit?

[Asmodian]

I have no way to bit accurately calibrate in YCbCr.

Does ColourSpace (HTL?) use a full range pattern for the 1D LUTs with the internal generator or does it not use the internal generator? I could not avoid failing calibration when trying to do a full range calibration with the internal pattern generator with Calman.

I got pretty good results for PC mode calibration with Calman but less or no banding/posterization with full range RGB input would be wonderful. :(

I am seriously considering switching to using my 48" CX in HDMI mode with 120 Hz 10 bit RGB full input. The banding in PC mode is frustrating me more than the lost resolution does, as long as it is the TV that does the subsampling. Maybe some sour grapes too, since the CX cannot do 120 Hz with full resolution chroma even when in PC mode. :o

And the fact that a 10-bit input has more banding in PC mode seemingly has to do with PC mode switching the TV into 8-bit mode.

Wow! that makes a lot of sense. Well, it actually doesn't make any sense but it does match what it looks like.

(Part of) the reason YCbCr 4:4:4 has less banding is because Nvidia dithers it by default. RGB is not dithered by default, but if you dither it with, for example, Vince's tool, the result is pretty much the same as YCbCr (you can also remove the default dithering from YCbCr to get it to look like RGB).

Yeah, but PC mode really seems to need this dithering and limited range so YCbCr 444 is a good compromise. I haven't done much testing between limited YCbCr and full range RGB in HDMI mode but with what I have done limited YCbCr still seems better. It might still be the dithering in that case too though.

[SamuriHL]

SamuriHL, did you change to black level LOW and madVR set to 8-bit?

Yes, the LG is set to low for all modes.

[SamuriHL]

Does ColourSpace (HTL?) use a full range pattern for the 1D LUTs with the internal generator or does it not use the internal generator? I could not avoid failing calibration when trying to do a full range calibration with the internal pattern generator with Calman.


I have a C8 so your question is invalid. :p :D Also, I don't do a 1D lut. I keep the factory 1D lut because any time you touch the 1D lut on these foolish LG OLED's it introduces all kinds of problems. Literally just resetting it will cause problems. So I skip that and only generate a 3D lut after doing a service menu white balance adjustment. And yes, HTL.


I got pretty good results for PC mode calibration with Calman but less or no banding/posterization with full range RGB input would be wonderful. :(


I gave up on Calman. I own it. I'm not renewing it. I don't want to get into that discussion.


I am seriously considering switching to using my 48" CX in HDMI mode with 120 Hz 10 bit RGB full input. The banding in PC mode is frustrating me more than the lost resolution does, as long as it is the TV that does the subsampling. Maybe some sour grapes too, since the CX cannot do 120 Hz with full resolution chroma even when in PC mode. :o



Wow! that makes a lot of sense. Well, it actually doesn't make any sense but it does match what it looks like.


Seriously, it's complete and total BS. Thanks, LG. LOL


Yeah, but PC mode really seems to need this dithering and limited range so YCbCr 444 is a good compromise. I haven't done much testing between limited YCbCr and full range RGB in HDMI mode but with what I have done limited YCbCr still seems better. It might still be the dithering in that case too though.

I may screw with that and see what it looks like. YCbCr 422 limited 10 bit worked well for me. This sounds even more promising. Yea, the panel is calibrated for RGB but it is what it is.

[chros]

What do you think I'm seeing? LOL Banding and posterization. These TV's just DO NOT like PC mode. :(
Yes, there is posterization.

(Part of) the reason YCbCr 4:4:4 has less banding is because Nvidia dithers it by default. RGB is not dithered by default
Yes, but madvr does! So what's happening with that? :)

but if you dither it with, for example, Vince's tool
This one (https://www.avsforum.com/threads/2020-lg-cx%E2%80%93gx-dedicated-gaming-thread-consoles-and-pc.3138274/post-59699820)? Is this the latest version?

Which command line did you use for different scenarios? I used:
- NvColorControl 12 RGB 1 2 4
- NvColorControl 12 RGB 0 2 0
- NvColorControl 8 RGB 1 1 4
- NvColorControl 8 RGB 0 1 0

And the fact that a 10-bit input has more banding in PC mode seemingly has to do with PC mode switching the TV into 8-bit mode. Why LG decided to do it this way and to this day is anyone's guess. Most likely to save on processing.
I know that Ted is pushing the PC mode is crap with 8bit agenda, but It's not 8 bit, try it out with Mehanik's HDR10 test pattern:
"03. Grayscale\04. 10bit test\01. test_10-bit_23.976.mp4"

And I saw the difference in real scenes as well (back in the day when I checked it) in shadow details.
That's why I refuse to put out 8bit with hdr.

Lastly, as I said, check (in a dark room) II. from here (https://www.avsforum.com/threads/2020-lg-cx%E2%80%93gx-dedicated-gaming-thread-consoles-and-pc.3138274/post-59700054) for posterization, shitshow starts from 3 min :)

And about Full vs Limited mode of the TV: if you calibrated your chain with Full and you switch to Limited then you have to calibrate it again :)

[SamuriHL]

There's a couple things you're glossing over. Yes madvr dithers. However it still has to get sent through the gpu after that before making it to the tv. And that tends to be a black box as to what it ends up doing to the video. It depends on how it's configured... See below.

As for full vs limited, people get that confused too. I'm not ever running my tv in full. Even if I set the gpu to full! Gasp whaaaaattttt?! Lol this is because in order to send a full signal madvr and the gpu both have to be set to full which I don't do. If I have the gpu set to full I have madvr set to tv levels. If I'm playing with one of the limited output in the gpu then madvr will be set to pc levels.

Obviously the best case is to run everything at full including the tv with black levels set to high. This ensures the gpu and madvr play nice together and no conversions are done. But we have lg oled. So while that's a nice theory the reality is the oled is doing stupid things to the full signal which ends up causing artifacts. These tvs really expect a limited signal. So second prize is the way I have it set up where madvr embeds the limited signal inside a full gpu signal. Gpu leaves it alone and the tv sees it as a limited signal. This gives the least amount of conversions and least amount of artifacts.

I've tried calibrating to full and it doesn't fix the issues. Very annoying.

Sent from my SM-G975U using Tapatalk

[SamuriHL]

Now that I'm thinking about vbb's idea of ycbcr 444 full more that actually might work best with these foolish panels. The reason is that madvr will embed a limited signal, the gpu will convert it from rgb to ycbcr using dithering, and the oled expects that as a signal from consumer devices anyway so that's likely the format it handles the best internally. Yes it'll still do a conversion but it does it better than on rgb. Interesting.

Sent from my SM-G975U using Tapatalk

[Magik Mark]

Can we use ycbcr 444 10bit in gpu?

Sent from my SM-G988B using Tapatalk

[SamuriHL]

Not at 60hz no.

Sent from my SM-G975U using Tapatalk

[chros]

There's a couple things you're glossing over. Yes madvr dithers. However it still has to get sent through the gpu after that before making it to the tv. And that tends to be a black box as to what it ends up doing to the video. It depends on how it's configured... See below.
I don't think so, maybe you are (?) :)
There's no different config options about the case I talked about: madvr dithers to 8bit, gpu outputs RGB full 8bit.
I'm not telling that it can't happen. I'm asking if it happens that how on earth?! :)

As for full vs limited, people get that confused too. I'm not ever running my tv in full.
I'm not confused, maybe I wasn't clear enough. And that's you, but I use full on the TV (black level high). I talked about only the black level setting on the TV (high/low), nothing else.
What I meant was: if you already calibrated your chain using full range on the TV (black level high) and if you switch to limited range on the TV (black level low) then no matter how correctly your chain is set up, your calibration is obsolete.
The reason behind this is that (as you guys said) the TV works differently with full range vs limited range.

Now that I'm thinking about vbb's idea of ycbcr 444 full more that actually might work best with these foolish panels. The reason is that madvr will embed a limited signal, the gpu will convert it from rgb to ycbcr using dithering, and the oled expects that as a signal from consumer devices anyway so that's likely the format it handles the best internally. Yes it'll still do a conversion but it does it better than on rgb. Interesting.
I'm 100% sure that posterization will be present, there's no way around it. See my post recently about LLDV (https://forum.doom9.org/showthread.php?p=1920202#post1920202) (and that is using "HDMI" limited mode, not even PC mode), that's exactly how PC mode looks in the best case.

[SamuriHL]

Conclusion....LG sucks and we continue to wait for better screens. Which is why I have zero intention of replacing my C8 any time soon. What good is 120hz and HDMI 2.1 when they have the same or similar problems? If they can't get the basics right, I certainly don't trust them to get more advanced features correct.

[Asmodian]

Can we use ycbcr 444 10bit in gpu?

Yes, with the very new 3000 series GPUs. HDMI 2.1 out is finally available!

The 3080 seems to be out of stock at the moment but I hope to get a 3090 by the end of October, at the latest. It also looks like there will be a 3060 and 3070 within a few months, unsurprisingly.

The reason behind this is that (as you guys said) the TV works differently with full range vs limited range.

And I hate them for it. Why can't I update the internal calibration LUTs with Calman for full range? :p

I'm 100% sure that posterization will be present, there's no way around it. See my post recently about LLDV (https://forum.doom9.org/showthread.php?p=1920202#post1920202) (and that is using "HDMI" limited mode, not even PC mode), that's exactly how PC mode looks in the best case.

But it is less than with RGB input. Especially on my CX it is pretty good with 8 bit YCbCr 444 limited in PC mode. But yes, it still has more posterization and banding than in a non-PC mode. :(

Conclusion....LG sucks and we continue to wait for better screens. Which is why I have zero intention of replacing my C8 any time soon. What good is 120hz and HDMI 2.1 when they have the same or similar problems? If they can't get the basics right, I certainly don't trust them to get more advanced features correct.

I will not fault you for this attitude, I would love better processing with for 10 bit RGB 120 Hz input now that I can send it. Someone else will come out with an OLED with good full res chroma processing and full HDMI 2.1 bandwidth in the near future, maybe Sony in a year or two. Who knows, maybe even LG will fix it.

That said the C9 and CX look good with 10 bit RGB 120 Hz input in a non-PC mode (The CX's PC mode is fine with 120 Hz input too, but it is not 444). LG sucks at 444 chroma but if you drop that everything else is pretty good. :(

HDR in PC mode is way better on the CX than on the C9 but SDR PC mode is about the same. I am not sure this helps anyone but if the banding isn't too bad (e.g. using aggressive dithering when converting to limited 8bit YCbCr 444) you can actually use PC mode for HDR content! HDR in PC mode on the C9 and earlier is absolutely terrible, tone mapping does not work properly so everything is very undersaturated.

All my posts about these TVs get too long... there are always caveats and special cases! :devil:

[SamuriHL]

That's why I, and I suspect others, are frustrated. There's no reason for any of this other than either laziness or stupidity on LG's part. Or they just don't care. I can understand this being a problem on one year's lineup of panels, but, 4 or 5 years in a row with barely an improvement? That's not really all that acceptable. Even things like the 1D lut problems, which portrait displays continues to say is an LG problem that they've promised a fix for, is still a problem after how long? These are basic, basic things that you as a consumer should expect to work out of the box...at least between generations of panels! My C8 is a known quantity and I know its limitations. I don't like them, but, at least I know what they are. Until one of these unicorn panels comes out that just works out of the box for all the things we want to do with it, I will sit on the C8. Not to mention that will buy me time to upgrade my receiver yet again for HDMI 2.1 while I wait. And time to find a 3080. BAHAHAHAHA. Maybe I'll get one for my birthday next year. LOL

[VBB]

Hey, at least I revived the thread! ;)

[SamuriHL]

Gee, thanks for that. LMAO :D LOL!

[Magik Mark]

Turns out 30 series cards have problems with C9 and CX

https://www.forbes.com/sites/johnarcher/2020/09/20/lg-oled-tvs-having-issues-with-latest-nvidia-rtx-30-graphics-cards/#4179cdc0267a

Sent from my SM-G988B using Tapatalk

[SamuriHL]

For crying out loud, LG, can you guys get ANYTHING right? "Oh we'll fix it in a firmware update!" Yea right. Good luck with that.

[Asmodian]

I find the article weird because I get the same issue with no 444 at 120 Hz with my 2080 Ti and a Club3D CAC-1085. That is a problem with the TV and has nothing to do with the 3000 series. The HDMI 2.1 G-sync black screen issue is the 3080 and has nothing to do with the TV (except it is one of the few HDMI 2.1 g-sync compatible displays). Two separate issues are being weirdly conflated, I get why but it is a really sloppy article.

I am also skeptical a firmware update will fix it but I am eternally hopeful. :o

The firmware was really broken when I got the CX, and it is still buggy in other ways, so it seems possible the TV sees 120 Hz input and automatically does something that results in 422 that they can turn off with a firmware change. It might increase latency or have some other negative consequence too. For example, can the CX hardware tone map HDR 120 Hz 444? Since better HDR tone mapping in PC mode is the only difference the CX has with the C9's video chain (that I know of, of course) it might explain why the CX is cannot do 444 at 120 Hz but the C9 can.

[oldpainlesskodi]

I just want to say - thanks for the heads up....I'll stay away from these panels, and stick with what I have (with no issues).

I'll have a dual room setup one day - when they perfect short throw laser projectors contrast issues, and then, TV in the lounge and a PJ in the cinema room.

K

[VBB]

It's really not all as bad as it sounds. I love my OLED, and I don't ever regret buying it, flaws and all. It's still the technology to beat at this point, but I can agree that it's not for everyone and every use case.

[VBB]

Question for the guys with anything newer than a 7-series: In PC mode, is the gamut still locked to WIDE? If so, and if you didn't use a 3D LUT and selected "this display is already calibrated", would it then make sense to set the gamut in madVR to DCI-P3, since the OLEDs come close to covering that in native mode? I've been using 3D LUTs for so long that I've never opted for anything else, but with my meter being from 2013 I don't want to do a new LUT at this time.

[chros]

It can be Yes to your question, or set Auto gamut and bt709 in madvr. Maybe the latter is the better since madvr doesn't have to do anything with SDR content.
On B8 you can select gamut in PC mode (Wide/Extended/Auto) in both SDR and HDR10. I also use 3dlut for SDR and gamut is set to Wide.

So, did you manage to test HDR10 modes during the weekend? What did you end up with? :)

[VBB]

I just downloaded the Mehanik patterns and your sample last night, so I haven't had a chance to look at those in the dark. Watched a bunch of things in both SDR and HDR, and I haven't observed any banding or posterization that would be immediately obvious (outside of specific test patterns). This is all with RGB full 8-bit, madVR 16-235 8-bit, TV in PC mode black level LOW.

I'm just happy I got something to mess with again, but I will go back to normal mode if I find something I don't like. On the plus side, I don't have to use smooth motion any longer, and that alone shaves off a bunch of ms. Enough, in fact, to go back to chroma NGU high, luma NGU high for all my 1080p@23fps upscales. With smooth motion, I was only able to use chroma NGU low and luma NGU med. Looks like I'm gonna hold on to my trusty 960 a little while longer :D

[chros]

:) I'm happy for you, yes SM costs a lot. But you could have used CRU before just like us :) That's why I'm telling that our cards are perfect for madvr: low power consumption.
Check those samples if you have time, I'll be curious what you will choose :)

[QBhd]

I'm still in the Full-RGB Full-Full camp but without the PC mode... Sure I loose 4:4:4 but it has the least banding of any option in HDR

QB

[VBB]

For me and my C7, YCbCr 4:2:2 10-bit has the least amount of banding by a long shot (in normal mode and both SDR and HDR), but I'm trying to get away from using that combo.

[Magik Mark]

Do you have any info if Panasonic or Sonly Oleds are having problems processing RGB from GPU? Should we switch?

[VBB]

We don't really know, because nobody here has either one, and there aren't any owners over at AVS who care or would know what to look for, to be honest. I'm being a bit sarcastic, but it's pretty much like that.

[SamuriHL]

This is why I continue to wait for better screen tech. Eventually, some day, one of these damn technologies will get enough of it right that it'll be worth upgrading. I LOVE my C8, even with all its flaws. The PQ is still out of this world when properly calibrated. And I came from a Panny plasma (which is sitting in my bedroom LOL). You couldn't pay me to go to whatever quantum dot LED technology Samsung's got going on. I'm sure it's fine enough, but, I LIKE my OLED. Perhaps MicroLED will replace it some day or whatever but right now, OLED rocks. Sony is over priced and still has issues (although maybe different issues than LG...as VBB says, we don't really know). Panny is on my s*** list for dropping the US market, although I believe Nev has said those have issues, too. So, really, we're still waiting.

[huhn]

you have to differentiated between a panel and processing issue.

if a TV has banding with RGB but not with YCbCr input than the panel has zero to do with that all panels run RGB/WRGB. so if a panel has no banding with YCbCr it is 100% possible to do the same with RGB. it's just incompetence.

sonys RGB processing is pretty much as good as it gets that doesn't mean sony TV are even close to be perfect they have there own issue they pretty much can't do dolby vision they just pretend they can. so if you get a sony TV and want to use dolby vision content and you know about the "quality" they "handle" it than i simply have no word for you.

Panasonic has multiply issues in RGB processing is not truely "RGB" is not subsampled but it does some crazy things by using some usual patterns you can make it visible you will not notice this with real content but ohh boi...
panasonic is part of the "real cinema" group that may intentionally display a refreshrate wrong or been unable to do so correctly (PC modes makes 23p useless it was again confirmed by HDTVtest) there is no way to force a sony to to do a 3:2 pulldown.
panasonic also has a force processing issue where it has 100 levels of sharpness and up to this day i have no freaking clue what is neutral it'S between 0-50 somewhere there isn't a mode that just disables it like all the PC modes on all the other brands.

the points about to little processing power and such are just that... the majority of sony TV have 2 processor just to show you a simple way to fix it if you are unable to to something a board could do with 1/20 of the processing power of your current board.

[VBB]

Screw it all! :D

I'm back to my original settings from three years ago: normal HDMI mode, RGB full 8/12-bit, madVR 0-255, TV black level high. While PC mode works seemingly well with some very particular settings, several torture tests made me reconsider. As is absolutely right: anything RGB looks significantly better than 4:2:2. Chroma pattern will also show this, of course. Makes me wonder now how I used 4:2:2 for so long...

One thing I found when I reset madVR during my testing was that "colored noise" in the dithering options seems to add significant banding to 10-bit sources. Could this be a bug?

[chros]

That was quick :)

I'm back to my original settings from three years ago: normal HDMI mode, RGB full 8/12-bit, madVR 0-255, TV black level high.
So like I use it, except for PC mode :D

Makes me wonder now how I used 4:2:2 for so long...
:D

One thing I found when I reset madVR during my testing was that "colored noise" in the dithering options seems to add significant banding to 10-bit sources. Could this be a bug?
Is it with Ordered dithering? Which samples did you use for this test?

[QBhd]

Screw it all! :D

I'm back to my original settings from three years ago: normal HDMI mode, RGB full 8/12-bit, madVR 0-255, TV black level high. ...

It's funny, no one listened to me long ago and I even got a little push back when I said that using YCbCr seemed wrong on many levels. Welcome back to reality, your chain above is the one I have been using ever since I watched Game of Thrones Season 8 Episode 3 "The Long Night" in HDR... this is truly a torture test for banding and sealed the deal for me with regards to the PC Mode or HDMI Mode debate. A Full RGB chain has never been a question for me :)

QB

[Magik Mark]

Screw it all! :D

I'm back to my original settings from three years ago: normal HDMI mode, RGB full 8/12-bit, madVR 0-255, TV black level high. While PC mode works seemingly well with some very particular settings, several torture tests made me reconsider. As is absolutely right: anything RGB looks significantly better than 4:2:2. Chroma pattern will also show this, of course. Makes me wonder now how I used 4:2:2 for so long...

One thing I found when I reset madVR during my testing was that "colored noise" in the dithering options seems to add significant banding to 10-bit sources. Could this be a bug?This was initial setting. My problem with this is introduces banding in Hdr. We're you able to find a work around?

Sent from my SM-G988B using Tapatalk

[Asmodian]

One thing I found when I reset madVR during my testing was that "colored noise" in the dithering options seems to add significant banding to 10-bit sources. Could this be a bug?

That is quite odd. I always disable colored noise because it looks weird to me. I expect it is more an artifact of the less than perfect 10 bit processing. Trying to figure out good settings for the W, R, G, and B subpixels when neighboring pixels have totally different hues is probably tricky.

That is something to remember, these TVs cannot do native RGB color like a normal display. Theoretically they could do pixel accurate rendering but the panels exhibit some artifacts that make me think you do have to consider the state of the surrounding pixels when setting voltages for subpixels (crosstalk is an issue). I don't know how they mix the white pixel in but the display will have to do more sophisticated / different image processing than one with proper RGB subpixels, with any input format.

[VBB]

@chros - I don't know how you use PC mode with that combo, especially after reading all your posts here and at AVS ;). PC mode is definitely useable with RGB full 8-bit, black level low, though. I always use ordered dithering, yeah. I could clearly see the banding added by "colored noise" in the S&M rotating ramps 8/10-bit pattern. The 10-bit side is usually perfectly smooth, and in this case it looked exactly like the 8-bit side. I was able to reproduce this several times. Weird.

@QBhd - I have always been a been a true believer in a full RGB chain, but for a while I was after the best possible banding scenario. And I don't know if it was the latest Nvidia driver that changed something, but after my five days of testing, I have to say that RGB never looked this good before. It used to show way more banding than YCbCr, or at least that's how I remember it :D

@Magic Mark - Are you talking about normal HDMI mode or PC mode? In normal mode you should not see any obvious banding with the RGB full chain. We're not talking about games here, are we?

@Asmodian - Same here. I have both dithering options disabled, and I would have never noticed this if it hadn't been for the reset.

[quietvoid]

Always had slight color banding in madVR over to the C8, mostly in shadows (hair line, jaw lines). Will definitely try disabling colored noise.

[Magik Mark]

@VBB Oops. I had my black level set to low. Now is ok.

May I ask do you still use 8bit or 10bit in madvr?

[VBB]

I switched back to 10-bit, because I ended up using RGB full 12-bit for the lower refresh rates. Don't forget to also change madVR back to 0-255, if you use black level HIGH.

[Magik Mark]

I switched back to 10-bit, because I ended up using RGB full 12-bit for the lower refresh rates. Don't forget to also change madVR back to 0-255, if you use black level HIGH.

I'm using 4k 60hz 8bit in GPU. Is 10bit in madvr still advisable?

[VBB]

If your entire chain is 8-bit and does not switch to something greater at the lower rates, then I would advise to set madVR to 8-bit.

[Magik Mark]

[emoji106]

Sent from my SM-G988B using Tapatalk

[Magik Mark]

Does anybody know if you will get HDMI 2.1 features using PCI 3 interface?

[SamuriHL]

They have nothing to do with each other. The answer is yes.

[Magik Mark]

Glad to know that. I thought I'll be needing a new MOBO with PCIE 4 support

Sent from my SM-G988B using Tapatalk

[Asmodian]

Apparently LG fixed their 120 Hz issues on the CX with firmware 03.11.25. Now there is no chroma subsampling in PC mode at 120 Hz. :)

I was worried it wasn't going to be an easy fix.

[scooterlord]

Hello from me again. Long time since I've last commented in this forum - although I've been silently lurking around :)

So I got my LG CX OLED this week, and I've been having some issues with the colors, etc. So, taking the suggestions in this thread under consideration I ended up with the following settings, but have some issues I want help with.

- On the TV I have set HDR - Standard (yep, not cinema), more or less all options close to defaults (demos are looking gorgeous!)
- Windows 10 w/ HDR enabled, RGB, 12-Bit, Full color, 23Hz

Most of the content I am going to see is h265, 10-bit, HDR. Using Potplayer with MadVR - the settings for my content that suit best require some processing, so I am using the following settings:
- Reduce banding artifacts med, med,
- Reduce random noise

My pc can handle it - i5 9600kf w/2060 RTX Super. Th problem is that this processing pushes GPU utilization to ~55% - essentially this means that it turns the fans on, which is annoying.
Switching to FFMpeg decoder, quality looks more or less the same, with ~5% GPU utilization and ~10% CPU utilization, which is ideal... EXCEPT the colors. Which make a huge difference! It's like HDR is not utilized at all or something. Tried all kinds of video renders but the result can only get worse that EVR.

Coming to questions:
- Is there a way to force MadVR do its processing on the CPU? Or was it not designed for such cases?
- What could be the issue for the huge difference in colors when using MadVR? Is it doing something special about the colors that I could most likely fake with a software decoder? Any suggestions or pointers on where to look?

Thanks!

[VBB]

@As - Have you installed the new firmware yet? If so, how does it look now as far as chroma, and does it look any better at 120Hz than at 60Hz or lower (in PC mode), especially when it comes to banding?

[Asmodian]

Sorry for the slow response. Yes, I have installed 03.11.25 on my LG CX.

For PC mode:
Chroma is perfect now, no chroma artifacts at all with ChromaRes.png. I am very happy with 4K120 Hz PC mode, at least I never expected it to be better than it is now. 120 Hz is better than 60 Hz in all ways.

With 10 bit out from madVR posterization is still noticeably worse, I tested 8 bit full range RGB. I think 120 Hz has less posterization in the midrange and perhaps near black as well but it seems a little softer rather than actually fewer bands, or maybe only slightly fewer. I wasn't expecting to see a difference. I especially noticed an improvement with very dim pixels becoming slightly less dim, at 60 Hz this can make a bright artifact line that moves with the gradient but disappears if the pattern is paused. At 120 Hz it doesn't seem to be an issue. That was the most annoying artifact for me with full range input and dithering to limited range YCbCr would mitigate it in a similar way. Switching back and forth is a bit annoying with the CAC-1085 due to needing to power cycle it so take all this as tentative. I did switch back and forth a lot but there was always some time between and I knew if it was 60 or 120 Hz, etc. :o

Oddly I don't seem to be able to send YCbCr 444 120 Hz to the CAC-1085 using driver 456.71 and a 2080 Ti.