View Full Version : madVR - high quality video renderer (GPU assisted)
mclingo
10th April 2019, 10:02
the general user wont see this though as they probably wont refresh rate match, it also only happens at sub 60hz resolutions, plus it doesnt happen if I put my TV in PC mode, not sure what that is either.
chros
10th April 2019, 10:54
What's the proper settings in madVR to output SDR content (including HDR2SDR pixelshader) to a DCPI-3P capable display?
- now I checked the "report bt2020 to display (nvidia only)" option for all SDR and HDR content (LG B8)
- set display to calibrated (until I'll create 3dluts)
- set DCPI-3P, pure power curve, 2.20 (in calibration tab) (and changing the gamma if it's needed)
madjock
10th April 2019, 11:56
see Warners signature.
Bl4ze87
10th April 2019, 15:13
Hello, there's a way to change color/dim the brightness of external subtitle while playing hdr file?
chros
10th April 2019, 16:11
see Warners signature.
Thanks and Warner as well (https://forum.kodi.tv/showthread.php?tid=223175&pid=2751243#pid2751243), it seems I thought well:
"Notes:
- When BT.709 is converted to BT.2020 or DCI-P3, it is simply placed in a larger container for the display, which shouldn’t lead to any oversaturation of BT.709 content. So it is safe to choose BT.2020 or DCI-P3 for both SDR and HDR content if your display color profile matches this setting;
- Output in DCI-P3 may require report BT.2020 to display (Nvidia only) is checked under calibration due to the fact most displays will be expecting a BT.2020 input despite only covering all or part of the DCI-P3 color space."
Hello, there's a way to change color/dim the brightness of external subtitle while playing hdr file?
This is not the responsibility of madVR but the player, e.g. MPC-HC / Be (you can do this in them).
Bl4ze87
10th April 2019, 16:25
This is not the responsibility of madVR but the player, e.g. MPC-HC / Be (you can do this in them).
The change doesn't affect hdr file.
I ve already a greysh color for subtitle that works normally with all kind of file, except the hdr one
mclingo
10th April 2019, 17:14
The change doesn't affect hdr file.
I ve already a greysh color for subtitle that works normally with all kind of file, except the hdr one
indeed if I calibrate for BT2020 in MADVR it doesnt change HDR titles at all, only SDR sub 60hz
madjock
10th April 2019, 17:38
https://forum.kodi.tv/showthread.php?tid=259188
Most 4K UHD displays have separate display modes for HDR and SDR. Calibration settings in madVR only apply to the display's default SDR mode. BT.2020 HDR content is passed through unless a special setting in hdr is enabled such as converting HDR to SDR.
VBB
10th April 2019, 21:19
The change doesn't affect hdr file.
I ve already a greysh color for subtitle that works normally with all kind of file, except the hdr one
Assuming your HDR files are REMUX, they all come with PGS subs, which are bitmap and not text based. Unfortunately, you cannot change them. Bothers me as well.
yok833
10th April 2019, 21:32
Hi,
I have an LG OLED 2017 and i would like to know if the last experimentals builds of madvr and measurment file can improve the tone mapping of my tv or is it mainly reserved for videoprojector ?
What would be the right "min target / real display peak nits" for an oled ?
Do I need tick "output video in HDR format" ?
With the settings of my tv do I need to turn off the "dynamic contrast" to let madvr tone mapping do is job ?
VBB
10th April 2019, 21:35
What's the proper settings in madVR to output SDR content (including HDR2SDR pixelshader) to a DCPI-3P capable display?
- now I checked the "report bt2020 to display (nvidia only)" option for all SDR and HDR content (LG B8)
Should be enough to put the OLED in "wide" mode. No need to check "report BT.2020...". "Wide" is the LG's native gamut. Really best to do this with LUTs, though. Calibrate with the TV in "wide", then create a LUT for all available SDR slots in madVR.
Asmodian
10th April 2019, 21:38
I also have a 2017 LG OLED and madVR's tone mapping can improve HDR but it is impossible to turn off LG's tone mapping, if they end up interacting it can cause issues (clipping is the most obvious).
I use ~600 nits (this is close to but not quite a max brightness), and you do need to tick "output video in HDR" if you want the TV to go into HDR mode. If you do see clipping you can turn down madVR's target nits.
Warner306
10th April 2019, 23:51
Are the highlights all turning white (ending up squished), or simply being compressed to a lower brightness? Does it work with brighter movies like Mad Max Fury Road when you set the target nits to something higher like 1,000 nits?
Clipping should imply the display is reading the metadata as being lower than it actually is (e.g. 500 nits) and it is cutting off some pixels (e.g. 500-600 nits). I don't think that is supposed to happen with PQ tracking that can hit 600 nits. That would be more like an internal display conversion to some type of gamma curve that is omitting some information.
ryrynz
11th April 2019, 00:52
but it is impossible to turn off LG's tone mapping
It is possible with the 8 series and above though yeah?
chros
11th April 2019, 10:28
Should be enough to put the OLED in "wide" mode. No need to check "report BT.2020...". "Wide" is the LG's native gamut.
Thanks, I didn't get that far yet, I have other issues :) I use Auto now, and it correctly switches to bt.2020 mode with "report BT.2020..." checked.
Really best to do this with LUTs, though. Calibrate with the TV in "wide", then create a LUT for all available SDR slots in madVR.
I don't think that's necessary: I only want to create 1 3dlut for DCI-P3 slot and madVR will do all the necessary conversions. (Actually I'll create 2 for this: 1 for night and 1 for day usage and switching them in a script using hardlinks.)
I have an LG OLED 2017 and i would like to know if the last experimentals builds of madvr and measurment file can improve the tone mapping of my tv or is it mainly reserved for videoprojector ?
No, you can use it with any SDR capable display as well (e.g. with your oled).
Whether it's better than the TV's conversion, that I don't know, you have to judge this. 1 thing for sure, since you want to use your TV in SDR mode, you will treat your TV better (way lower oled light setting).
What would be the right "min target / real display peak nits" for an oled ? Do I need tick "output video in HDR format" ?
Go back couple of pages on that thread, multiple users attached screenshots of their settings (based on their real display nits).
With the settings of my tv do I need to turn off the "dynamic contrast" to let madvr tone mapping do is job ?
In general, you want to turn off every "stupid" TV processing/feature, including the mentioned one as well; but leave RealCinema On.
It is possible with the 8 series and above though yeah?
No, you can't.
If6was9
11th April 2019, 11:50
Is your issue triggered similar to this post:
https://forum.kodi.tv/showthread.php?tid=223175&pid=2788748#pid2788748
It's not a bug (or at least not from AMD), why send 10bit to the display in SDR?
"The more there is the better", it does not always go well ...
Are you sure your panels are 10bit real?
I with my PJ 10bit, after several tests, I chose to always go out in 8bit plus dithering (in madVR and AMD gpu), the work done by madVR with dithering is above to 10bit: No banding
To use 10bit, you must have the whole chain that supports them 100% (HDMI port protocols, HDMI cables with sufficient bandwidth, and first of all the display)
Solitaire01
11th April 2019, 13:24
So I know I may be the outlier here, but I've come to determine the least problematic config for my 2070 is to have madvr set to pass through HDR but disabled passing meta data, followed by manually engaging the PS HDR as necessary. This last step I may script with autohotkey at some point.
The results appear every bit as good as they did prior to this driver debacle -- No perceptible difference in dynamic or colour range.
...?
Cheers!
Doug
huhn
11th April 2019, 13:36
@If6was9
using up to 12 bit input support is a default feature on nearly all TV for over 6 years now.
in a perfect world even if your panel is 6 bit sending 10 bit wouldn't be worse at all. while i agree that blindly using 10 bit is not best for picture quality simply because far more TV produce bending with 10 bit input then not it has to work.
a GPU driver that is changing the colors of the image just by switching between 8 bit and 10 bit renderer is clearly not working properly.
how should someone write a color critical application with that behaviour?
intel and nvidia doesn't do this so they either both workaround an windows bug or the issue is at AMD.
Manni
11th April 2019, 14:10
@If6was9
using up to 12 bit input support is a default feature on nearly all TV for over 6 years now.
in a perfect world even if your panel is 6 bit sending 10 bit wouldn't be worse at all. while i agree that blindly using 10 bit is not best for picture quality simply because far more TV produce bending with 10 bit input then not it has to work.
a GPU driver that is changing the colors of the image just by switching between 8 bit and 10 bit renderer is clearly not working properly.
how should someone write a color critical application with that behaviour?
intel and nvidia doesn't do this so they either both workaround an windows bug or the issue is at AMD.
It's not necessarily the drivers' fault, but nVidia does that with the JVCs.
When sending 8bits, you get the native colorspace in SDR rec-709, or SDR WCG (DCI-P3 or BT2020).
When sending 12bits, the colorspace is NOT the native colorspace (that tracks saturation very well) but a wonky colorspace that needs a lot more corrections. You only get the native colorspace if you select "report BT2020" in MadVR's calibration tab. This, along with the colorspace bug (in 12bits, the JVCs force YCC422 internally when the source sends RGB 12bits or YCC 444 12bits).
The second issue is the JVCs fault because it also happens with a standalone player, however I'm not sure if the first one is a driver issue or a display issue.
In any case, the colors are NOT the same with nVidia on the JVCs using 8bits or 12bits, which clearly is a problem and makes calibration more challenging as you have to activate BT2020 for SDR WCG and disable it for SDR rec-709 to get the optimal gamut.
For this reason, I use RGB 8bits: no colourspace issues and easier calibration.
kostik
11th April 2019, 14:20
New NVIDIA driver is out : 425.31 , can someone check if HDR metadata working correctly?
If6was9
11th April 2019, 14:33
@If6was9
using up to 12 bit input support is a default feature on nearly all TV for over 6 years now.
in a perfect world even if your panel is 6 bit sending 10 bit wouldn't be worse at all.
Instead I think at that date, most consumer panels are not even real 10bit, but 8bit real + 2 fake, not limited to the one written on the brochures.
So sending 10bit to these panels, both in SDR and HDR, worsens the image quality.
And even with native 10bit panels, you have to have the whole chain capable of conveying the 10bit, otherwise you create the bottleneck, which generates artifacts on the color palette, or band saturation, or color conversions busted
Seeing is believing, there are tests with specific patterns,
try sending 10bit, and then 8bit with dithering,
is simple
chros
11th April 2019, 14:42
as you have to activate BT2020 for SDR WCG and disable it for SDR rec-709 to get the optimal gamut.
Can't you just use BT2020 on Projector and DCI-P3 in madvr for every case?
So sending 10bit to these panels, both in SDR and HDR, worsens the image quality.
Seeing is believing, there are tests with specific patterns,
try sending 10bit, and then 8bit with dithering, is simple
Yep, huhn knows this, and many LG OLED owners as well, that's why they use 8bit output :)
huhn
11th April 2019, 14:52
@Manni
as long as no other user group reproduces the problem you should assume your end device first.
more then try to reproduce is is not possible for me.
i blamed mclingos screen first too.
@If6was9
i'm over this topic
i'm using 8 bit myself. i'm recommending it for quite sometime but...
if TV processing would be flawless we wouldn't get any issues.
every TV in theory can do the same as madVR. that doesn't change that 10 bit has to work properly.
edit: AFAIK the AMD HDR API needs 10 bit input. that doesn't stop you from sending 8 bit to the display but you have to use 10 bit rendering which creates this bug if i see this correctly.
@Sunset1982
can you please try the "compatibility" FSE mode.
mpc-hc64.exe -> right click properties -> compatibility -> disable full-screen optimisations. just as a warning my TV/driver/windows version combination was happy about this to say it friendly so use it at your own risk... my desktop PC was fine with that.
tp4tissue
11th April 2019, 15:17
Is madvr hdr_sdr conversion EXPECTING a rec709 space after the conversion ?
If I run the monitor in (AdoberRGB)/ P3 space on a wide gamut pc monitor, is that oversaturation ??
If6was9
11th April 2019, 15:17
@Manni
as long as no other user group reproduces the problem you should assume your end device first.
more then try to reproduce is is not possible for me.
i blamed mclingos screen first too.
@If6was9
i'm over this topic
i'm using 8 bit myself. i'm recommending it for quite sometime but...
if TV processing would be flawless we wouldn't get any issues.
every TV in theory can do the same as madVR. that doesn't change that 10 bit has to work properly.
edit: AFAIK the AMD HDR API needs 10 bit input. that doesn't stop you from sending 8 bit to the display but you have to use 10 bit rendering which creates this bug if i see this correctly.
8bit in madVR and gpu,
in this way we will be safe from any problems,
the work done by madVR with dithering is impeccable.
I trust anyone to notice the difference between 10bit and 8bit + dithering ....
Conversely, errors in color conversion or band saturation with the 10bit are clearly visible
If I send HDR 8bit, through AMD RX580 to PJ, I see no bug
If6was9
11th April 2019, 15:20
Is madvr hdr_sdr conversion EXPECTING a rec709 space after the conversion ?
If I run the monitor in (AdoberRGB)/ P3 space on a wide gamut pc monitor, is that oversaturation ??
madVR does not support AdoberRGB
huhn
11th April 2019, 15:37
Is madvr hdr_sdr conversion EXPECTING a rec709 space after the conversion ?
no.
If I run the monitor in (AdoberRGB)/ P3 space on a wide gamut pc monitor, is that oversaturation ??
the output of madVR is converted to BT709 if nothing is changed under calibration.
i ignore HDR here.
@If6was9
you know you can dither to 10 bit too... and there are more then enough good working professional PC screens with 10 bit.
so there are still user using passthrough to a PJ?
chros
11th April 2019, 16:34
If I run the monitor in (AdoberRGB)/ P3 space on a wide gamut pc monitor, is that oversaturation ??
In theory it shouldn't, but it's hard to tell without measuring the actual display with the given colorspace combinations.
SamuriHL
11th April 2019, 16:54
New NVIDIA driver is out : 425.31 , can someone check if HDR metadata working correctly?
NO. That driver is useless and doesn't fix anything. Nev said it's not part of the new major release yet. They just ran out of numbers from what he said.
Manni
11th April 2019, 17:37
Can't you just use BT2020 on Projector and DCI-P3 in madvr for every case?
No, if you use DCI-P3 in madVR the PJ also has to be calibrated to DCI-P3, same for BT2020.
You can use the same P3 or BT2020 calibration for everything with a flat panel, but not with a projector if you care about picture quality. Using the same calibration for HDR and SDR raises the black floor and lowers the contrast in SDR unnecessarily. Not a problem with OLED, but an issue with projectors.
Point is, I use an SDR rec-709 calibration with DVD/bluray for which "report BT2020" has to be switched off, and an SDR-DCI-P3 for HDR content for which "report BT2020" has to be switched on. This doesn't mean that it's using BT2020 primaries as I use a DCI-P3 calibration in the PJ, just that the WCG calibration will be selected automatically by the Vertex/Maestro in the PJ and the PJ will use the correct (linear) colorspace, not the wonky one.
If6was9
11th April 2019, 18:03
@If6was9
you know you can dither to 10 bit too... and there are more then enough good working professional PC screens with 10 bit.
so there are still user using passthrough to a PJ?
I don't think anyone here is watching movies on a PC monitor
Only those who do not have a PC, because the tone mapping performed by madVR is superior to any other consumer device today on the market.
And applying the 10bit to SDR material makes no sense
huhn
11th April 2019, 19:08
actually 10 bit helps more on gamma output then PQ output.
and madVR tonemapping is problematic with LCD.
Siso
11th April 2019, 19:22
I don't think anyone here is watching movies on a PC monitor
I'm watching movies on a pc monitor ;) :D
Asmodian
11th April 2019, 19:39
I don't think anyone here is watching movies on a PC monitor
Then you would be wrong. madVR is used on many PC monitors.
VBB
11th April 2019, 21:01
I don't think that's necessary: I only want to create 1 3dlut for DCI-P3 slot and madVR will do all the necessary conversions. (Actually I'll create 2 for this: 1 for night and 1 for day usage and switching them in a script using hardlinks.)
What are you trying to accomplish? With DisplayCal, it takes just a few minutes to calculate the additional LUTs. All you need to do is run a calibration in the native color space ("wide" for the LGs). Once that's done, in the 3D LUT tab you select the color spaces you want to calculate for your individual LUTs. DisplayCal will put them all in the correct slots in madVR, and madVR will automatically pick the correct one for the content you're playing, even for HDR, as long as you're converting and not passing through.
mclingo
11th April 2019, 21:38
re the colour saturation issue on AMD cards, if you set MADVR to 8bit it fixes it as you know but you lose HDR switch, would a good workaround be for MADSHI to create a force 8 bit toggle switch for all SDR, just for AMD users. This would force 8 bit on all SDR material but leave HDR as 10 bit .
Should be possible?
Asmodian
11th April 2019, 21:43
This is easily achieved with a profile, madshi does not need to be making workarounds for bad drivers that will hopefully be fixed. :)
tp4tissue
11th April 2019, 22:27
no.
the output of madVR is converted to BT709 if nothing is changed under calibration.
i ignore HDR here.
So, the DCI-p3 colors of a typical 4KHDR movie, is compressed down to rec 709, by madvr hdr_sdr..
Then using an srgb monitor space mode, it's color correct ?
If using an A-rgb or p3 space mode, then it's oversaturated ?
jkauff
11th April 2019, 22:35
Then you would be wrong. madVR is used on many PC monitors.
Indeed. madVR looks great on my 27" 4K PC monitor. My 4K TV has to do its own upscaling, because it's in another room and I don't have an HTPC. Fortunately, it does almost as good a job.
Asmodian
11th April 2019, 22:55
If using an A-rgb or p3 space mode, then it's oversaturated ?
Yes, if you set "disable calibration controls for this display" madVR assumes BT.709 (similar to sRGB for our use case).
If you are using P3 set "this display is already calibrated" to DCI-P3. Sadly madVR does not have an Adobe RGB option for "This display is already calibrated" so you need to use another calibration option to get correct colors on an Adobe RGB screen. :(
Adobe RGB is a color space useful for professional print work so in that sense it makes sense to leave it out of video options but there are a lot of Adobe RGB wide gamut displays out there so it would still be nice to have it.
huhn
12th April 2019, 01:25
just use a synthetic 3D LUT.
and again HDR-SDR has nothing todo with this.
mclingo
12th April 2019, 01:41
This is easily achieved with a profile, madshi does not need to be making workarounds for bad drivers that will hopefully be fixed. :)
... you mean like adding CRU solely for NVIDIA cards :)
huhn
12th April 2019, 01:58
blocking 10 bit for SDR would be ok if the issue would be present all the time like it was blocked on nvidia because nvidia broke it and there was no except everyone got the problem but as far as i can see it right now that's not the case her.
custom rates is useful for AMD and intel user too and you know that...
tp4tissue
12th April 2019, 02:17
just use a synthetic 3D LUT.
and again HDR-SDR has nothing todo with this.
Why does HDR to SDR have nothing to do with this ?
If the 4Khdr material contains a color that's only in p3 space (outside of 709 space).
Does hdr_sdr process not clip/map that color to something in the 709 space ?
huhn
12th April 2019, 02:25
because madVR can output what ever you want the HDR->SDR doesn't change this.
if an bt2020 SDR file is played madVR will do the "same".
i have no clue if the madVR gamut mapping (not HDR related) using a roll of or is clipping or something else and i don't know that for 3D LUT too. i guess it's clipping.
tp4tissue
12th April 2019, 05:47
Yes, that is my question.
How is Madvr mapping the gamut from the P3 colorset embedded in BT2020 down to 709.. Which is why I asked originally, if Madvr 's hdr-sdr process is expecting a certain end colorspace.
suisenken
12th April 2019, 08:04
H all,
after reading all the post available everywhere in the web i'm really surprise that its seen so little have issue with those SO annoying stuttering with RTX new card. Coming from a simple GTX 1060 with no issues at all i upgrade to RTX 2060 AMP from Zotac and start with the nightmare to be unable to watch smooth playback, so i follow advice from some of your, going to FSE that i have never need to use it before on GTX card, testing all scaling algorithm and getting nothing better, the only improvement but still not ideal its to add with FSE a lower GPU queue, to its minimum and do not present to much frame in advance, it works for some and look like it really depend from the source, on some video it will have a minimal impact or worst not at all, and im taking HD UHD on MKV AND BD ISO M2TS
i really start to feel a bit part of the few or even alone with this issue, it look like it work fine for so many of RTX owner, nothing on that in the bug report, i thought that odd. Only have this issue with video and not in game.
@madshi if you read this by any luck i know you are very very busy on the render improvement HDR (great work for all the member involved that some brilliant brain storming over there) and also your new project but i'm wonder if that stuttering issue on RTX card is it something that you are willing to take into consideration cause there is something going there with those RTX . Thx.
Seriously this make me mad without the vr , stating to think maybe to ask a card replacement is that only last option ?
if there is some light there from someone facing that same issue ? sharing some thought please let me know.
My Equipements and playback tools: mardvr 0.92.17 MPC HC MPC BE jriver Media 25 ( fse in jriver and HDR on jvc its a pain)
I5 8600k 4.20GHZ 32MB RTX2060 1000 watt power su / Panasonic fz800 JVCdla 7900 marantz 7010
Thanks.
If6was9
12th April 2019, 11:35
re the colour saturation issue on AMD cards, if you set MADVR to 8bit it fixes it as you know but you lose HDR switch, would a good workaround be for MADSHI to create a force 8 bit toggle switch for all SDR, just for AMD users. This would force 8 bit on all SDR material but leave HDR as 10 bit .
Should be possible?
I with the RX580 in 8bit quietly send the HDR to the PJ
I see that you use full 4:4:4 RGB color space at 60hz,
maybe that's the problem, and not the AMD drivers.
There are very few chains that can manage 4K HDR in full RGB 60hz, let alone 10bit,
but it's not an AMD problem
huhn
12th April 2019, 11:38
here you go https://forum.doom9.org/showthread.php?p=1870305#post1870305
this help most RTX user.
BTW is well known the issue. just hope the next major nvidia driver does something.
DMU
12th April 2019, 11:44
I with the RX580 in 8bit quietly send the HDR to the PJ
Your projector goes into HDR mode at 8bit output madVR settings? :eek:
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.