madVR - high quality video renderer (GPU assisted)

[Tobiwan]

I have a question about BT.2020, how did I know that I can use it? Does the monitor need to support BT.2020 or the GPU? Or can everyone use the new BT.2020 and it works?

[tp4tissue]

Consumer bt2020 displays do not exist.

What we have now at the highest are dci-p3 displays. If you make 3dluts, make a p3 lut if the gamut of your device is larger than Rec709 (which is more/less the same as SRGB).

[tp4tissue]

@fourbigkids

Pixel shift is usually able to sharpen edges, but may blur certain types of textures, because the shift is an overlap of half a pixel

You'll have to test it to find out if you like what it does.

If you output 4K to 4K that's one step with chroma upscaling

If you output 4K to 1080p, that adds a downscaling step, which may require you to lower the chroma upscaling setting.

Either one may be sharper depending on content and viewing distance.

Upscaling anything lower than 1080p to 1080p is a non-issue.

[Tobiwan]

So if I understand it right it doesn't give me any benefit to set BT.2020 instead of BT.709 for my monitor? I never use 3DLUT or yCMS to calibrate, I always use "already calibrated" with BT.709, PPC and 2.40 gamma.

[tp4tissue]

So if I understand it right it doesn't give me any benefit to set BT.2020 instead of BT.709 for my monitor? I never use 3DLUT or yCMS to calibrate, I always use "already calibrated" with BT.709, PPC and 2.40 gamma.


In order to set bt2020, the monitor has to have a bt2020 mode, and be operating In that mode.

There are no BT2020 consumer monitors, so that is not an option.

If you do not have a calibrated monitor, pick disable calibration

[Tobiwan]

Thanks tp4tissue!

[Manni]

In order to set bt2020, the monitor has to have a bt2020 mode, and be operating In that mode.

There are no BT2020 consumer monitors, so that is not an option.

If you do not have a calibrated monitor, pick disable calibration

This isn't correct.

There are no consumer monitor covering 100% of BT2020, but there are many consumer and monitors and projectors offering a BT2020 mode that covers above rec-709, often up to P3 and for some projectors like the RS4500/Z1 beyond P3.

So there is an advantage to selecting a BT2020 mode in the display/projector, even if the projector is unable to cover BT2020 fully. For example, with my rs2000 that covers close to 100% of P3 with its P3 filter, I benefit a lot from selecting the BT2020 calibration (around 100% of P3 with a P3 filter) or the HDR calibration (around 90% of P3 without filter). Both deliver a significantly larger color volume than the rec-709 calibration, for HDR/SDR WCG content of course.

With madVR, especially if you use 3D LUTs, it's usually better to select a DCI-P3 calibration as it tends to be closer to the actual capability of the display and madVR can simply discard the BT2020 container and deliver P3 which is used to grade the majority of current content, even if it's delivered within a BT2020 container. But with a consumer source you don't have the choice, they either output rec-709 or BT2020 so you have to select one or the other if you want to track the correct saturations.

[el Filou]

If you output 4K to 1080p, that adds a downscaling step, which may require you to lower the chroma upscaling setting.Or check "scale chroma separately if it saves performance", which will allow use of better downscaling and/or not compromise on HDR tonemapping quality if you are GPU limited.

[cyberscott]

Does anyone know if the LG C6 is fine for 12 bit output or should I just leave it at 8 or 10 bit?

I own a LG C6 and I've noticed no additional banding or issues with 12 bit enabled in the Nvidia control panel with madVR. Of course, all equipment has variations so best thing to do is a "stare and compare" with a paused video or better yet a .pgn screen capture with 12 bit on and off.
Even so, madVR's 8 bit processing is excellent so even with 12 bit not enable, you'd be hard pressed to see a difference, I'd think.

[ryrynz]

Yeah that's usually the case.. The benefits of using 10/12bit are typically small to non apparent.

[tp4tissue]

@cybscott

Does it show banding even at rgb 444, 12bit @ 23/24 hz ?

Sometimes it's the sharpening algorithm which screw up 4:2:2 conversion, leading to (banding), but it's not actually banding, it's just bad sharpening.

[fourbigkids]

Thanks @tp4tissue and @el Filou. Looks like I have some testing to do! :-)

[cyberscott]

@cybscott

Does it show banding even at rgb 444, 12bit @ 23/24 hz ?

Sometimes it's the sharpening algorithm which screw up 4:2:2 conversion, leading to (banding), but it's not actually banding, it's just bad sharpening.

For my HTPC set-up, I exclusively use RGB Full, 444 @ 4k. The C6 HDMI input is set to "PC" I have madVR set to use "switch to matching display mode" for 2160p23, 2160p29, 2160p50, and 2160p59. My current video card is a RTX 2080 TI.
Some of the videos I watch have banding already in the source and some with little to no banding in the source. If I look at a scene that has little to no banding, running the Nvidia control panel at 12 bit with the appropriate display mode, it does NOT add any additional banding or video noise that was already part of the source.
I don't typically use any additional sharpening algorithms beyond what is built into NGU-AA, which I use exclusively for my content.

[Alexkral]

With madVR, especially if you use 3D LUTs, it's usually better to select a DCI-P3 calibration as it tends to be closer to the actual capability of the display and madVR can simply discard the BT2020 container and deliver P3 which is used to grade the majority of current content, even if it's delivered within a BT2020 container. But with a consumer source you don't have the choice, they either output rec-709 or BT2020 so you have to select one or the other if you want to track the correct saturations.

I guess you mean select a DCI-P3 calibration on the screen if it happens to have one, or create it, because otherwise sending DCI-P3 is wrong.

[Manni]

I guess you mean select a DCI-P3 calibration on the screen if it happens to have one, or create it, because otherwise sending DCI-P3 is wrong.

Re-read what I wrote, and that is clearly what I said in the context of my post. Remember we were discussing display modes. I also mentioned the BT2020 container, but maybe that's too technical :)

[j82k]

I suggest everyone running their LG Oleds at 12-bit or in PC-mode to check out Rtings 16-bit gradient test pattern played in madVR.

https://www.rtings.com/images/test-materials/2016/gradient-16-bit.tif

My LG C8 shows noticeable banding, especially visible in the darker green gradient when using either 12-bit or 8/12-bit PC-mode. In 8-bit, non-pc mode it looks fine.
It was the same on my C6 so I don't really understand if something is wrong with my setup or if other people just don't notice the banding.

By the way PC-mode also forces 3:2 pulldown even when outputting 23.976 Hz.
So you get 4:4:4 chroma and lower input lag but you'll also get color banding and forced 3:2 pulldown.

[SamuriHL]

Yup, that's well known for the C8, unfortunately.

[jmone]

So I spent the afternoon trying to get my JVC x7500 working without dropping/repeating frames and I think I've worked it out. The JVC is very picky about what signal it will accept and all the ones offered by madVR outside the normal ones did not work. So I started with the EDID / CTA profile (but it would have a drop/add about 1 per hour) and tweaked that a bit. While madVR's config GUI says this setting will last 7.79 hours, I can confirm it worked perfectly for just under 2 hours. Interestly, is that madVR kept updating the reported display calculation from 23.9758 @ 1 min, to 23.97608 but the end of the movie as well as the expect drop.repeat frames.
https://jmone.org/nextcloud/index.php/s/Yoomp84BYf8gxSP/preview
https://jmone.org/nextcloud/index.php/s/CLsmkCx6JBJnDjT/preview

[arcspin]

@jmone
Can you please tell us what you did to tweak your settings.

I'm in a similar situation and would very much appreciate some guidance.
I have read the tutorial found here:
http://madvr.com/crt/CustomResTutorial.html

Is it something similar you did?



Best regards

[jmone]

@arcspin, yes that is the correct tut but I then manually changed the values as per the 1st pic (just type them in then press "test mode" to see if they work for you).

[hannes69]

While madVR's config GUI says this setting will last 7.79 hours, I can confirm it worked perfectly for just under 2 hours. Interestly, is that madVR kept updating the reported display calculation from 23.9758 @ 1 min, to 23.97608 but the end of the movie as well as the expect drop.repeat frames.

Normal behaviour. The longer you measure physical quantities, the more accurate gets the measurement. This is the case when the value is a constant one in reality.
Here the values (frequencies) aren´t constant, they are e.g. temperature dependant. So when watching a movie and heating up your computer case these values will drift to a certain degree.
So your custom resolution is more or less perfect now (as perfect as you can tune it) when drifting between a couple of hours and some days drop/repeat time.

[RXP]

I own a LG C6 and I've noticed no additional banding or issues with 12 bit enabled in the Nvidia control panel with madVR. Of course, all equipment has variations so best thing to do is a "stare and compare" with a paused video or better yet a .pgn screen capture with 12 bit on and off.
Even so, madVR's 8 bit processing is excellent so even with 12 bit not enable, you'd be hard pressed to see a difference, I'd think.


Thanks - I'll keep it at 8 bit so when I have my JVC x9000 on it'll work with that too.

[tp4tissue]

I suggest everyone running their LG Oleds at 12-bit or in PC-mode to check out Rtings 16-bit gradient test pattern played in madVR.

https://www.rtings.com/images/test-materials/2016/gradient-16-bit.tif


By the way PC-mode also forces 3:2 pulldown even when outputting 23.976 Hz.


Are we sure that this test pattern is Valid and that if it shows banding, it means our 10bit HDR remuxes would suffer the same ?

Are you talking about PC-mode on the LG, or pc-mode in NvidiaCP (hdmi intent) ?

:scared:

[j82k]

I'm of course talking about PC-Input mode on the LG Oled, as that is the only way to get 4:4:4 chroma.

I also notice the banding on HDR gradient test pattern when using pc-mode or 12-bit so it's not a pattern issue.

I'm not trying to prove anything here, use whatever you like. I was just a bit baffled that there are quite a few people using such a subpar setup.
4:4:4 chroma is nice but not at the cost of increased banding and 3:2 pulldown judder...

[SamuriHL]

I personally agree with you. I shut off PC mode because of the banding issues. I haven't looked back since.

[el Filou]

I was just a bit baffled that there are quite a few people using such a subpar setup.
4:4:4 chroma is nice but not at the cost of increased banding and 3:2 pulldown judder...Out of curiosity, do those issues also happen if using YCbCr 4:4:4 as input, or just with RGB?
I'm disappointed that UHD TVs that are supposed to be high end can't properly process 4:4:4 @ 4K without suffering those kind of quirks. This shows they have not yet reached the same maturity as the late high end 1080p displays (maybe in the 2019 models it will be fixed?).

[j82k]

PC input mode has always been somehow fu*ked up on LG Oleds. Some time ago I pretty much tried all the color/bit-depth combinations in pc-mode and while some were better than others, all of them produced worse gradients than the standard hdmi mode, especially in darker colors. Output was set to 24Hz, so surely this can't be a HDMI bandwidth issue.

[SamuriHL]

It's very likely caused by their video processor (the a9 or whatever it's called). I have my doubts about the 2019 models fixing 12 bit banding issues or banding issues with PC mode. But we'll know in a few weeks when the first models are released I guess. I'm certainly not upgrading again any time soon.

[Asmodian]

4:4:4 chroma is nice but not at the cost of increased banding and 3:2 pulldown judder...

I notice this as well, in PC mode. I need to use smooth motion at 60 Hz instead of 24 Hz. :(

However, I do like smooth motion more than 4:2:2 chroma. 4:2:2 chroma bothers me really a lot.

I personally agree with you. I shut off PC mode because of the banding issues. I haven't looked back since.

I notice PC mode causes banding issues for HDR on my C7 but not for SDR. Does SDR seems to have more banding in PC mode for you?

Also the input lag in non-PC mode is way too high for me, I need to start worrying about adding an audio delay, the mouse feels laggy, etc.

[SamuriHL]

I didn't really test a lot of SDR content so I'm not really sure. I noticed it in HDR and ended up having to turn it off. I don't have the input lag issues you're seeing, though. But I have almost all processing turned off.

[j82k]

In SDR the difference is more subtle than in HDR and can only be seen on test pattern or very specific scenes but I absolutely hate color banding so PC-mode is out of the question, at least for movies.

[el Filou]

Try NGU Sharp for upscaling.

[cyberscott]

I suggest everyone running their LG Oleds at 12-bit or in PC-mode to check out Rtings 16-bit gradient test pattern played in madVR.

https://www.rtings.com/images/test-materials/2016/gradient-16-bit.tif

My LG C8 shows noticeable banding, especially visible in the darker green gradient when using either 12-bit or 8/12-bit PC-mode. In 8-bit, non-pc mode it looks fine.
It was the same on my C6 so I don't really understand if something is wrong with my setup or if other people just don't notice the banding.

By the way PC-mode also forces 3:2 pulldown even when outputting 23.976 Hz.
So you get 4:4:4 chroma and lower input lag but you'll also get color banding and forced 3:2 pulldown.

Thanks for the test pattern!

I tried the test pattern and I do see what you mentioned about the banding in PC mode on the C6. I goes away disabling PC mode and switching to the YCbCr444 color space in the control panel.
To my eyes, the banding is slight and with some madVR setting tweaks, it can be pretty much eliminated in videos but at the cost of more processing power.
I can deal with it but hopefully TV manufactures can sort out PC mode better in the near future.

[tp4tissue]

Thanks for the test pattern!

nip nip


Do you guys think, it might be that the Banding is actually gone in other modes, because the LG's internal processor is applying its own debanding?

So with PC mode, it's actually doing the raw input without internal deband, and thereby exhibiting the natural output ?

It doesn't seem right that the banding comes back so consistently on multiple products, unless something internal has been set intentionally.

[cyberscott]

Do you guys think, it might be that the Banding is actually gone in other modes, because the LG's internal processor is applying its own debanding?

So with PC mode, it's actually doing the raw input without internal deband, and thereby exhibiting the natural output ?

It doesn't seem right that the banding comes back so consistently on multiple products, unless something internal has been set intentionally.
It is entirely possible. Pretty much all picture enhancements are disabled in PC mode in order to get RGB 4:4:4 and the lower latency.
All the other modes have some or all of the enhancements available.
I enjoy the over all picture quality of C6 in conjunction with madVR in PC mode. Best thing to do is just experiment with the other modes and choose what looks best for YOU.

[ryrynz]

No need to quote an entire post directly above yours, thanks.

[madjock]

Having a play with one of the latest test builds of HDR->SDR now I have a graphics card that can handle it. I think it looks really nice, but unsure if this is because my SDR is calibrated more for blacks and whites (using test patterns and nothing fancy) and going from HDR to an HDR -> SDR comparison is hard to compare as I cannot really calibrate the HDR side, so the SDR looks brighter and punchier for an initial comparison.

Any expert or more experienced person done a more in depth analysis with an LED TV and have any thoughts ?

[j82k]

So I just measured a 50 point saturation sweep using 20% pattern stimulus on my LG C8 in PC HDMI input mode (4:4:4 chroma) vs standard HDMI input mode.
All picture enhancement settings were disabled. GPU was set to RGB Full 8-bit.

https://i.imgur.com/5O5jQPi.gif

That pretty much confirms the banding I'm seeing when using pc-mode. :scared:
I didn't even bother to check other colors.

[el Filou]

So with PC mode, it's actually doing the raw input without internal deband, and thereby exhibiting the natural output ?But you should not see banding with madVR if dithering is configured correctly, so "raw input" on the TV should not show it either. IMHO it's caused by multiple low-quality conversions inside the TV because some of the image processing pipeline cannot handle RGB. The fact PC mode adds other issues (3:2 judder on 24p) is weird too.
If you want to check if the LG is running a debanding algorithm on its inputs when not in PC mode, you can disable dithering in madVR and then compare PC and standard mode again.

[j82k]

The fact PC mode adds other issues (3:2 judder on 24p) is weird too.


Not really weird. Even in standard HDMI mode you need to enable "real cinema" to get proper 24p cadence but that also adds input lag and in pc-mode this option is greyed out (permanently disabled).

Though I've successfully created a 48Hz custom refresh rate (using 50Hz as a base) as a workaround and that way you can get proper cadence for movie content even in pc-mode.

[Warner306]

Having a play with one of the latest test builds of HDR->SDR now I have a graphics card that can handle it. I think it looks really nice, but unsure if this is because my SDR is calibrated more for blacks and whites (using test patterns and nothing fancy) and going from HDR to an HDR -> SDR comparison is hard to compare as I cannot really calibrate the HDR side, so the SDR looks brighter and punchier for an initial comparison.

Any expert or more experienced person done a more in depth analysis with an LED TV and have any thoughts ?

I have set-up an LED TV like this:

2.40 (madVR) -> 2.20 (display)

Because the image is actually getting brighter at the display, I increased the real display peak nits / lower limit in madVR to 225 nits to prevent the image from appearing washed out.

Set like above, the difference in brightness between HDR and SDR is actually reasonably close when the dynamic target nits in the test builds is making the decision on the end display brightness as the movie plays. UHD Blu-ray actually looks better than SDR Blu-ray in most cases, with some exceptions where SDR Blu-ray is noticeably brighter.

For a fair comparison, use full-bitrate UHD rips and not any of the UHD torrents out there that suffer from visual loss of quality.

[tp4tissue]

So I just measured a 50 point saturation sweep using 20% pattern stimulus on my LG C8 in PC HDMI input mode (4:4:4 chroma) vs standard HDMI input mode.
All picture enhancement settings were disabled. GPU was set to RGB Full 8-bit.

https://i.imgur.com/5O5jQPi.gif

That pretty much confirms the banding I'm seeing when using pc-mode. :scared:
I didn't even bother to check other colors.


SO they make the panel, they probably have to do a sweep / precalibration of some sort

My point is maybe that table gets disabled in PC mode to trade for response time or something.

So you're getting a more raw version of what the panel should look like .

This may actually be a better mode to calibrate under, because then correction is not happening twice, once by chip, once by gpu.

How do calibrated luts look on this mode ? still banding ?

[j82k]

What I assume could be happening is that the TVs internal lut works at less precision in pc-mode. If it is input lag related then the game preset which has the same input lag as pc-mode should also suffer from increased banding. Gonna check that out later.

I did create a 3dlut calibration in pc-mode once but it didn't turn out good, which isn't surprising. To correct all of these banding errors (they are everywhere in the low end range) would require an insane amount of corrections.

[XMonarchY]

Many of us have got around that problem by using the CRU utility to create custom resolutions, then using madVR to optimize them. CRU apparently adds the custom settings at the OS level rather than the GPU level which protects them from driver-level changes.

Could you help me out with that? I could never figure out on how to add the same exact custom resolution that madVR adds via CRU...

[jkauff]

Although you can enter any values you want into CRU, it's generally better to create a custom resolution with the default EDID settings in CRU, then use madVR to create the optimizations.

[chros]

Could you help me out with that? I could never figure out on how to add the same exact custom resolution that madVR adds via CRU...

https://forum.doom9.org/showthread.php?p=1868998#post1868998

[thighhighs]

What I assume could be happening is that the TVs internal lut works at less precision in pc-mode. If it is input lag related then the game preset which has the same input lag as pc-mode should also suffer from increased banding. Gonna check that out later.

I did create a 3dlut calibration in pc-mode once but it didn't turn out good, which isn't surprising. To correct all of these banding errors (they are everywhere in the low end range) would require an insane amount of corrections.
Game mode is specifically designed to reduce input lag for games :) Changing the input to "PC", but without turning on the game mode, you will get the PC-Desktop mode with improvements to read the text. I don't have a C8, but I have two LG TVs and dont think LG has made new logic.
If you looking for an accurate picture mode for madvr and movies, try isf Expert mode first, this made for it. Game mode is alternative for people who want chroma 4:4:4.

edit: Your input always should be set to PC, otherwise you trash Windows and apps quality. In this mode TV internally has fps-based source detection with different picture settings, so no reasons to switch input, if you dont watch tons of high framerate movies, of course.

[TheProfosist]

anyone have a screenshot of what the madVR OSD shows when madVR is in fullscreen exclusive mode?

I ask because mine just says full screen windowed. Also there isn't that delay when going fullscreen. Manually triggering it via shortcut doesn't seem to do anything either.

[j82k]

@thighhighs
Wow, there is so much wrong with what you just said and I'm not even gonna bother correcting your statements. It seems like you didn't even read (or understand) my previous posts....

edit: also the issues discussed were specific to watching movies via madVR on LG Oleds and not about windows usage.

[thighhighs]

anyone have a screenshot of what the madVR OSD shows when madVR is in fullscreen exclusive mode?

I ask because mine just says full screen windowed. Also there isn't that delay when going fullscreen. Manually triggering it via shortcut doesn't seem to do anything either.
https://miku.hatsune.my/potplayer-with-seeking-gui-in-madvr-exclusive-mode.jpg Screenshot from google. Try tick\untick some checkboxes from general settings section. Or if you on Win10 just use windowed mode and dont care about it.

@thighhighs
Wow, there is so much wrong with what you just said and I'm not even gonna bother correcting your statements. It seems like you didn't even read (or understand) my previous posts....
When watch movies you use game or desktop modes with 4:4:4 chroma, because 4:4:4 chroma. And reported about bad picture (banding). I miss something?

also the issues discussed were specific to watching movies via madVR on LG Oleds and not about windows usageWho care about properly connection... :D