Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 21st August 2019, 03:14   #161  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Quote:
Originally Posted by Warner306 View Post
Good that you got it sorted. For some reason QBhd's results aren't matching.
They won't. They're using the latest release build of madvr which has had ALL KINDS of fixes in the betas. So the fact that it doesn't match doesn't surprise me now that j82k is using the latest beta.
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 21st August 2019, 03:17   #162  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
I don't have a clue about AMD cards. Do they maybe have similar metadata problems as nvidia had in the past? 696 was the number one would always get with the 1000 nits bogus metadata or windows HDR.

I wasn't using the test builds and still got it to work.

edit:
I can set the target peak nits up to 760 in madVR while keeping the "669" value, at 770 target peak it switches to "696". So 760 target peak nits might be a good value to use for people whose C8 can reach that as I assume that is the point up to where the TV will do no or the least amount of tone mapping.

The fastest way to make the metadata change happen is to disable/enable 'output video in HDR format', so no need to restart the player like I did earlier.

Last edited by j82k; 21st August 2019 at 03:42.
j82k is offline   Reply With Quote
Old 21st August 2019, 04:15   #163  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
AMD cards use the windows hdr which probably does still send fake metadata at 1000.

That's awesome about the max target nit. I can't wait to play this weekend!

Also I really want to thank you for all this awesome information!! This has been a great learning experience for me and I truly appreciate it!

Sent from my SM-G975U using Tapatalk
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED

Last edited by SamuriHL; 21st August 2019 at 04:18.
SamuriHL is offline   Reply With Quote
Old 21st August 2019, 04:45   #164  |  Link
quietvoid
Registered User
 
Join Date: Jan 2019
Location: Canada
Posts: 570
I have also been testing madVR (latest beta) on my LG C8 and obtain similar results.
With dynamic target nits, using 10 000 nits as a maximal target causes clipping when the target peak nits are automatically set to 1000+.

I've ended up using 750 peak and maximal target nits to avoid clipping.
Even using 800 as maximal target has some clipping.

Last edited by quietvoid; 21st August 2019 at 04:49.
quietvoid is offline   Reply With Quote
Old 21st August 2019, 08:18   #165  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
afaik AMD uses AGS for HDR and not the OS else there would be no need to add this API in madVR if the OS HDR is used anyway.
huhn is offline   Reply With Quote
Old 21st August 2019, 12:11   #166  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Hmm I could have sworn there was a discussion about the fact that HDR mode on nvidia had a slight advantage over amd due to the nvidia HDR api vs amd using the windows os HDR. I readily admit I could be 100% wrong though.

Sent from my SM-G975U using Tapatalk
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 21st August 2019, 12:14   #167  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by Warner306 View Post
There is no standard for tone mapping, but almost all current HDR displays start the roll-off point at some point...
Yes, thanks, but do you understand that I've talked about ABL algos?

Quote:
Originally Posted by j82k View Post
So 760 target peak nits might be a good value to use for people whose C8 can reach that
In what window size? 2/5/10/20% ... ?
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config
chros is offline   Reply With Quote
Old 21st August 2019, 14:18   #168  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
Quote:
Originally Posted by chros View Post
In what window size? 2/5/10/20% ... ?
I guess that is up to you what window size to use and if you want to account for some ABL by using a larger one or not. I haven't given this much thought and usually use a 5-8% window when measuring peak brightness. I think the way ABL works is that it just evenly dims the whole picture so it shouldn't cause any clipping in highlight but I'm really not sure on this.
j82k is offline   Reply With Quote
Old 21st August 2019, 17:21   #169  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
I just went back to the good and old 385.28 nvidia driver and guess what....

Metadata changes happen instantly now! I can change madVR's "HDR target peak nits" value and see the C8s second last WB value change immediatley while the movie is playing. So the metadata change happens on the fly without me having to disable/enable HDR.

This is also how I remembered things and is the reason why I was confused and thought madVR didn't change the metadata. Again just an Nvidia driver problem...

I think all the newer Nvidia drivers behave like this, meaning metadata change requires retriggering of the HDR signal whereas with 385.28 it just happens on the fly...
j82k is offline   Reply With Quote
Old 21st August 2019, 18:17   #170  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Fascinating. But not entirely surprising.

Sent from my SM-G975U using Tapatalk
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 21st August 2019, 23:47   #171  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Quote:
Originally Posted by j82k View Post
expert controls - white balance - method: 20 points - code value

There are 20 adjustment points and the second last one changes depending on what metadata the TV receives. A different value here means that the TV uses a different tone mapping curve.

Here's some values I've found so far while playing Godzilla King of Monsters (4000 nits MDL).
Second last adjustment point value:

passthrough: 713
madVR set to 4000: 713
madVR set to 1000: 696
madVR set to 700: 669
I wanted to come back to this. I SOOOOO appreciate this. I have confirmed exactly what you're seeing. madvr up to 760 is using 669, above that 696, and then 713. I used Aquaman to test 4000 nits and A Wrinkle In Time to test 1000 nits and the results were consistent. To be clear to those who may be looking at this information later down the road, this is with dynamic tone mapping on the C8 turned OFF. I am EXTREMELY pleased with the results we're getting. I've yet to watch a whole movie with these settings but the quick tests I've done so far look outstanding. I do want to compare with the UB820 this weekend when I get some time and just take a look at the same movies. To make it a "fair" comparison, I'll probably set madvr target nits to 1000 cause that's what the UB820 uses for OLED tone mapping. Nonetheless, this is great info and really helps us a lot.
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 22nd August 2019, 09:43   #172  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by j82k View Post
I think all the newer Nvidia drivers behave like this, meaning metadata change requires retriggering of the HDR signal whereas with 385.28 it just happens on the fly...
I wanted to tell you this, but I didn't know what GPU you have. Which one?
And indeed, everything applies immediately.

@Manni: a huge thanks to you for your endless driver testing and finding the one (!) that is the best for Pascal cards!!!

Quote:
Originally Posted by j82k View Post
I can set the target peak nits up to 760 in madVR while keeping the "669" value, at 770 target peak it switches to "696". So 760 target peak nits might be a good value to use for people whose C8 can reach that as I assume that is the point up to where the TV will do no or the least amount of tone mapping.
Quote:
Originally Posted by j82k View Post
Here's some values I've found so far while playing Godzilla King of Monsters (4000 nits MDL).
Second last adjustment point value:

passthrough: 713
madVR set to 4000: 713
madVR set to 1000: 696
madVR set to 700: 669

So the TV apparently uses yet another tone mapping curve when receiving 700 nits metadata or rather no tone mapping because it can display 700 nits?
So, here are the second last code values on B8 (DTM off in B8, tested with 4000 and 1000 nits content, madvr latest stable release version):
705 : passthrough 4000
705 : pixelshader 2501 - 10000
692 : passthrough 1000
692 : pixelshader 761 - 2500
669 : pixelshader 100 - 760

So, what does it mean exactly?
- there are slightly different code values on B8 and C8 (I assume due to the max brightness difference)
- the same 760 nits value can be applied in B8 as well!!! (the above mentioned brightness difference doesn't matter)
- I checked and 760 nits value looks the best (vs 500 / 2000 / etc.), I haven't compared to passthrough yet
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config

Last edited by chros; 22nd August 2019 at 09:45.
chros is offline   Reply With Quote
Old 22nd August 2019, 13:26   #173  |  Link
ashlar42
Registered User
 
Join Date: Jun 2007
Posts: 652
Quote:
Originally Posted by ashlar42 View Post
This is curious, as I initially understood things as you described them here to be. But then I read different opinions.
Is it possible to completely disable LG's tone mapping, if madVR does all the work already? In the case of preanalyzed files, it should do that on a frame by frame level, which makes me think it should provide the best results. Better probably than what Dolby Vision can achieve (if that works on a scene by scene basis, as I remember).
At the beginning of the month I asked this but was told it couldn’t be done. Am I right in understanding that it’s more or less what you’ve been discussing in the past few pages? Using madVR for tone mapping in order to avoid LG’s built in functionality?

Also, on a different subject, does using Smooth Motion in madVR improve 24fps material playback with regard to motion resolution? Theoretically the frames updating on the screen would be 120 and not 24, which should make things better for a sample and hold display. And with 120 frames to work with, blurring should be minimal, right?
ashlar42 is offline   Reply With Quote
Old 22nd August 2019, 14:56   #174  |  Link
QBhd
QB the Slayer
 
QBhd's Avatar
 
Join Date: Feb 2011
Location: Toronto
Posts: 697
That might be true if you could do 120 hz... And even then I doubt blending would happen.

QB
__________________

Last edited by QBhd; 22nd August 2019 at 15:01.
QBhd is offline   Reply With Quote
Old 22nd August 2019, 16:18   #175  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Quote:
Originally Posted by ashlar42 View Post
At the beginning of the month I asked this but was told it couldn’t be done. Am I right in understanding that it’s more or less what you’ve been discussing in the past few pages? Using madVR for tone mapping in order to avoid LG’s built in functionality?
That's what we've been discussing, yes, but I want to be clear here. There's no way to disable the tone mapping built in to the lg tv. That being said by sending the tv metadata that uses the max target nit that the tv can display it seems to be avoiding triggering the more aggressive tone mapping curves of the built in tone mapping. This is about the best situation we can hope for. And it looks really good.

Sent from my SM-G975U using Tapatalk
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED

Last edited by SamuriHL; 22nd August 2019 at 16:21.
SamuriHL is offline   Reply With Quote
Old 22nd August 2019, 16:32   #176  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
Found another really strange behavior of the C8 using the HDR color clipping test pattern from here:
https://drive.google.com/drive/folde...DUewbFsSXY1r8p

There is a big difference of clipping points with certain colors depending on the color format output.
It's especially noticeable with red and green. I used passthrough HDR and levels were of course set correctly.


Red:
RGB full/limited - clips at ~590
YCbCr 444/422 - clips at ~820

Green:
RGB full/limited - clips at ~780
YCbCr 444/422 - clips at ~590

With other colors the difference wasn't that big but what's really strange is that red clips much earlier with RGB output but with green it's the opposite, it clips much earlier with YCbCr. These big differences can't be minor color conversion rounding errors right? Anyone able to confirm or explain this??


Quote:
Originally Posted by chros View Post
I wanted to tell you this, but I didn't know what GPU you have. Which one?
And indeed, everything applies immediately.
I have a 1050ti.

Quote:
Originally Posted by chros View Post
So, here are the second last code values on B8 (DTM off in B8, tested with 4000 and 1000 nits content, madvr latest stable release version):
705 : passthrough 4000
705 : pixelshader 2501 - 10000
692 : passthrough 1000
692 : pixelshader 761 - 2500
669 : pixelshader 100 - 760

So, what does it mean exactly?
- there are slightly different code values on B8 and C8 (I assume due to the max brightness difference)
- the same 760 nits value can be applied in B8 as well!!! (the above mentioned brightness difference doesn't matter)
- I checked and 760 nits value looks the best (vs 500 / 2000 / etc.), I haven't compared to passthrough yet
I wouldn't rely on rtings peak measurements. When you look at their C7 burn-in test there are peak differences of more than 100 nits between the 6 TVs even though they're all the same model. Panel variance I guess.

I assumed a different WB value shown means that the TV uses a different tone mapping curve but I could be wrong. Looking at HDR white clipping pattern I can't detect a change when going from 760 to 761 in madVR even though that causes the TV to show a different WB value.
j82k is offline   Reply With Quote
Old 22nd August 2019, 16:35   #177  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by SamuriHL View Post
That's what we've been discussing, yes, but I want to be clear here. There's no way to disable the tone mapping built in to the lg tv. That being said by sending the tv metadata that uses the max target nit that the tv can display it seems to be avoiding triggering the more aggressive tone mapping curves of the built in tone mapping.
That's correct. As we saw above, there are 3 tone curves implemented in 8 series (just like the 3 modifiable ones in 9 series). And by setting 760 nits in madvr the less aggressive is used but what it means in reality is another question. Not to mention the tons of options that can be changed in madVRhdrMeasure86

I'll upload some selected clips later for those who wants to compare the result with pure passthrough (+ DTM).
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config
chros is offline   Reply With Quote
Old 22nd August 2019, 17:59   #178  |  Link
ashlar42
Registered User
 
Join Date: Jun 2007
Posts: 652
Quote:
Originally Posted by QBhd View Post
That might be true if you could do 120 hz... And even then I doubt blending would happen.
Well, clearly we need to wait for HDMI 2.1 videocards but I don't understand why you say blending wouldn't happen. There is the "always" option for smooth motion, isn't there?
ashlar42 is offline   Reply With Quote
Old 22nd August 2019, 18:05   #179  |  Link
QBhd
QB the Slayer
 
QBhd's Avatar
 
Join Date: Feb 2011
Location: Toronto
Posts: 697
24x5=120... So nothing to blend... As far as I understand smoothmotion

QB
__________________
QBhd is offline   Reply With Quote
Old 22nd August 2019, 19:09   #180  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by QBhd View Post
24x5=120... So nothing to blend... As far as I understand smoothmotion
It depends on the refresh rate of the TV you use:
- in your example: 23p -> you're right
- but the guys are experimenting in 60p (!) -> blending

Quote:
Originally Posted by j82k View Post
There is a big difference of clipping points with certain colors depending on the color format output.
It's especially noticeable with red and green. I used passthrough HDR and levels were of course set correctly.
...
With other colors the difference wasn't that big but what's really strange is that red clips much earlier with RGB output but with green it's the opposite, it clips much earlier with YCbCr. These big differences can't be minor color conversion rounding errors right? Anyone able to confirm or explain this??
Sorry, but I don't understand the test case: you mentioned passthrough as well. So, does it only happen with madvr's pixeslhader?

Quote:
Originally Posted by j82k View Post
I wouldn't rely on rtings peak measurements. When you look at their C7 burn-in test there are peak differences of more than 100 nits between the 6 TVs even though they're all the same model. Panel variance I guess.
Good point, I guess I have to dig out my meter during the weekend.

Quote:
Originally Posted by j82k View Post
I assumed a different WB value shown means that the TV uses a different tone mapping curve but I could be wrong. Looking at HDR white clipping pattern I can't detect a change when going from 760 to 761 in madVR even though that causes the TV to show a different WB value.
I'm pretty sure you are right. There are 3 different tone curves implemented in 8 series and the one that belongs to 100-760 nits in madvr is the less aggressive.
I also can't notice the difference between 760 and 761 (maybe with the test clips below).
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config
chros is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 15:18.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.