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 11th February 2019, 20:36   #54661  |  Link
madjock
Registered User
 
Join Date: May 2018
Posts: 259
Quote:
Originally Posted by el Filou View Post
With HDR the issue is not limited to just colours that look correct, tbh. The whole cinematographic effect of some scenes can radically change with bad tonemapping.
No doubt, but what concerns me is if we/they mess to what they think it should look like, then are they actually seeing what it was meant to be. I am sure like me people read UHD reviews on Bluray.com and the amount that say the colours are this, or uptick in detail here, but how many people actually care what there TV is doing. We are actually buying Blurays or UHDs from reviews when its just a normal TV with no fancy third party application making it perfect ? if there is such a thing and looking at that forum, it is going round in circles with every new clever iteration.
madjock is offline   Reply With Quote
Old 11th February 2019, 20:37   #54662  |  Link
madjock
Registered User
 
Join Date: May 2018
Posts: 259
Quote:
Originally Posted by iSeries View Post
With passthrough there is nothing that should be messed up, but apparently even when just passing HDR through, nvidia is sending bogus metadata to the TV. So this is (may?) be an issue for all madVR nvidia users, regardless of whether they are using madVR to tonemap (and sending as HDR) or straight passthrough
I mean from my side with messing with TV settings, so probably make an ass of it myself let alone nVidia
madjock is offline   Reply With Quote
Old 11th February 2019, 21:07   #54663  |  Link
Raimu
Registered User
 
Join Date: Oct 2015
Location: Oulu, Finland
Posts: 5
Quote:
Originally Posted by el Filou View Post
Yes, I also am seeing something similar on a 1080 screen albeit with fewer steps. The problem is NGU can only double, and (IIRC) SSIM can only scale x and y in the same proportions so if your source is anamorphic like TV, you need another step to do the scaling step from 1,25:1 to 1.78:1
Hey, that makes perfect sense. I didn't recall SSIM only does x and y in lockstep. Thank you!
Raimu is offline   Reply With Quote
Old 11th February 2019, 21:46   #54664  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
Quote:
Originally Posted by madjock View Post
No doubt, but what concerns me is if we/they mess to what they think it should look like, then are they actually seeing what it was meant to be. [...] if there is such a thing and looking at that forum, it is going round in circles with every new clever iteration.
I don't see it that way. From reading that thread over at AVS Forum, they are very carefully comparing the results of madVR's tonemapping with what a full range image would/should look like. It's also madshi's stated goal to create processing that gives results that are visually closer to what a full range HDR display would (highlight recovery is a good example). They spend a lot of time testing with and discussing specific scenes, but it's all done in a pretty scientific way IMHO.
Of course anybody is free to ignore that and just use passthrough especially if you have a good HDR display, but if you only own an SDR display that can output high brightness or an HDR display with bad tonemapping, and you want the best picture possible then it's an interesting development and I don't think it's a case of excessive tweaking.
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 12th February 2019, 00:07   #54665  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by Raimu View Post
This seems like an unusually convoluted scaling path. Noticed this upscaling TV content from 576p.

I would think you Photoshopped that. If not, it just looks ridiculous, to me.
Warner306 is offline   Reply With Quote
Old 12th February 2019, 00:13   #54666  |  Link
Manni
Registered User
 
Join Date: Jul 2014
Posts: 942
Quote:
Originally Posted by el Filou View Post
I don't see it that way. [...]
Thank you
__________________
Win11 Pro x64 b23H2
Ryzen 5950X@4.5Ghz 32Gb@3600 Zotac 3090 24Gb 551.33
madVR/LAV/jRiver/MyMovies/CMC
Denon X8500HA>HD Fury VRRoom>TCL 55C805K
Manni is offline   Reply With Quote
Old 12th February 2019, 00:38   #54667  |  Link
Manni
Registered User
 
Join Date: Jul 2014
Posts: 942
Quote:
Originally Posted by sat4all View Post
@Manni

Does other apps like HDR 3D games also send wrong metadata or it's only related to madVR?
Quote:
Originally Posted by KoKlusz View Post
If you would like to test it, Shadow of the Tomb Raider and Resident Evil 2 have free demos on steam that support HDR.
As far as I can see, once you enable HDR in the OS, it's exactly the same bogus and constant HDR metadata that is sent to the display.

It starts when I enable HDR in the CP, and the metadata remains the same when the game is launched and plays. Of course I've enabled HDR in the game as well, as it was disabled by default.

I couldn't find the demo of RE2 on Steam, but I did find the demo of SOTTB, so that's what I used for testing.

Game isn't a priority so please don't ask for more tests, I'm busy enough with the HDR tonemapping
__________________
Win11 Pro x64 b23H2
Ryzen 5950X@4.5Ghz 32Gb@3600 Zotac 3090 24Gb 551.33
madVR/LAV/jRiver/MyMovies/CMC
Denon X8500HA>HD Fury VRRoom>TCL 55C805K
Manni is offline   Reply With Quote
Old 12th February 2019, 06:54   #54668  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,890
@Raimu
can you do a screen of the full OSD.

madVr did strange things before if up and downscaling is needed at the same time.
but for that i need a full OSD and a sample to reproduce it.
huhn is offline   Reply With Quote
Old 12th February 2019, 09:05   #54669  |  Link
RXP
Registered User
 
Join Date: Aug 2002
Posts: 55
I've just got my PC hooked up with my TV and enjoying using MadVR especially on the JVC RS49.

I'd like to use MadVR's tone mapping with my LG C6. I use pixel shaders and pass the image through as HDR to benefit from the higher peak brightness of my TV in HDR.

My question is: Due to the Nvidia/Windows bugs where bogus meta data of 1000/20 is sent; will the internal tone mapper LG has throw off the results of MadVR? If I set the peak nits of my display to 600, LG's tone mapper should be bypassed as MadVR will report all content as 600 nits or below. But if the OS is saying it's 1000 nits - it'll be applying its own curve?
RXP is offline   Reply With Quote
Old 12th February 2019, 09:44   #54670  |  Link
madjock
Registered User
 
Join Date: May 2018
Posts: 259
Quote:
Originally Posted by RXP View Post
I've just got my PC hooked up with my TV and enjoying using MadVR especially on the JVC RS49.

I'd like to use MadVR's tone mapping with my LG C6. I use pixel shaders and pass the image through as HDR to benefit from the higher peak brightness of my TV in HDR.

My question is: Due to the Nvidia/Windows bugs where bogus meta data of 1000/20 is sent; will the internal tone mapper LG has throw off the results of MadVR? If I set the peak nits of my display to 600, LG's tone mapper should be bypassed as MadVR will report all content as 600 nits or below. But if the OS is saying it's 1000 nits - it'll be applying its own curve?
You could use the drivers it works on 385.28 (using them now). But I have to say this metadata thing news is news to me, I have tried to find other information, and have found the odd quote.

But for some reason in all the guides in using madVR and the likes, there is no mention of use this or this driver for HDR (due to metadata) so it seems like its a well known problem (or is it) thats just getting talked about again ?

https://forums.geforce.com/default/t...data/?offset=2

Last edited by madjock; 12th February 2019 at 09:51.
madjock is offline   Reply With Quote
Old 12th February 2019, 09:58   #54671  |  Link
sat4all
Registered User
 
Join Date: Apr 2015
Posts: 62
Nvidia bug affect both madVR passthrough and Pixel Shader with HDR output enabled, Do some comparaison by playing your HDR content using TV internal media player and you will see what i'm talking about.
But if you ask me, Pixel Shader with HDR output enabled is the best of the worst.
I was digging arround nvidia forums and i couldn't find a single topic about this bug, so time they hear our voice. and maybe madshi could help us by using his NV contact.

https://forums.geforce.com/default/t...-hdr-metadata/

Edit: haha madjock was faster quoting my own post
__________________
ZOTAC MAGNUS EN1060K: Win 10 x64 + Kodi DSPlayer x64
LG OLED65C8 / Denon AVR-X3200W / KEF E305+ONKYO SKH-410 / Synology DS2415+ / Logitech Harmony 950

Last edited by sat4all; 12th February 2019 at 10:00.
sat4all is offline   Reply With Quote
Old 12th February 2019, 10:50   #54672  |  Link
RXP
Registered User
 
Join Date: Aug 2002
Posts: 55
Thanks guys, but does 385.28 mean you can pixel shader with HDR output without issue? Easy for me to downgrade
RXP is offline   Reply With Quote
Old 12th February 2019, 11:09   #54673  |  Link
sat4all
Registered User
 
Join Date: Apr 2015
Posts: 62
Yes, 385.28 work properly with either passthrough or HDR pixel shader.
__________________
ZOTAC MAGNUS EN1060K: Win 10 x64 + Kodi DSPlayer x64
LG OLED65C8 / Denon AVR-X3200W / KEF E305+ONKYO SKH-410 / Synology DS2415+ / Logitech Harmony 950
sat4all is offline   Reply With Quote
Old 12th February 2019, 11:46   #54674  |  Link
iSeries
Registered User
 
Join Date: Jan 2009
Posts: 625
Anyone know what the last nvidia driver to work properly for hdr passthrough is?
iSeries is offline   Reply With Quote
Old 12th February 2019, 11:59   #54675  |  Link
sat4all
Registered User
 
Join Date: Apr 2015
Posts: 62
The bad news, is nvidia's hdr api is deprecated and pending removal.
As nevcairiel said, maybe madshi should consider OS HDR API instead of NV.
__________________
ZOTAC MAGNUS EN1060K: Win 10 x64 + Kodi DSPlayer x64
LG OLED65C8 / Denon AVR-X3200W / KEF E305+ONKYO SKH-410 / Synology DS2415+ / Logitech Harmony 950
sat4all is offline   Reply With Quote
Old 12th February 2019, 13:07   #54676  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
So, because of all this bogus metadata talk I compared nvidia drivers 385.28 (red) and 418.81 (yellow) simply by running a 4000 nits metadata grayscale sweep in HCFR.
TV is a LG C8.

As can be seen in the attached picture, below 450 nits there is no difference but above that the roll-off is much smoother with the correct metadata (385.28).
Using windows HDR produces the same bogus metadata result as 418.81.




I really hope this will get fixed soon.
j82k is offline   Reply With Quote
Old 12th February 2019, 13:16   #54677  |  Link
iSeries
Registered User
 
Join Date: Jan 2009
Posts: 625
Quote:
Originally Posted by j82k View Post
So, because of all this bogus metadata talk I compared nvidia drivers 385.28 (red) and 418.81 (yellow) simply by running a 4000 nits metadata grayscale sweep in HCFR.
TV is a LG C8.
Hi j82k, thanks for this - I know you're likely to be busy but any way you could test a couple other driver versions? (I loaned my meter to someone). Specifically 398.11, which is apparently the last driver that is working fine for HDR passthrough (and conveniently the driver that brought improved 23p playback...I never could get any good timings with 385.28)

Last edited by iSeries; 12th February 2019 at 13:45. Reason: Added thanks
iSeries is offline   Reply With Quote
Old 12th February 2019, 13:23   #54678  |  Link
sat4all
Registered User
 
Join Date: Apr 2015
Posts: 62
@j82k
Thanks, that very helpfull, would you please install the latest test build along with nvidia 418.81 drivers and take measurements with those settings:

tone map HDR using pixel shaders
output video in HDR format = enabled
target / real display peak nits = 800
apply dynamic clipping = enabled
apply target nits selection = disabled
LG dynamic tone mapping = off

Because it seems ok for me and almost matchs correct playback using my tv embedded media player.

Btw, did you take the previous measurements with your tv's dynamic tone mapping on or off?

__________________
ZOTAC MAGNUS EN1060K: Win 10 x64 + Kodi DSPlayer x64
LG OLED65C8 / Denon AVR-X3200W / KEF E305+ONKYO SKH-410 / Synology DS2415+ / Logitech Harmony 950

Last edited by sat4all; 12th February 2019 at 14:19.
sat4all is offline   Reply With Quote
Old 12th February 2019, 13:28   #54679  |  Link
madjock
Registered User
 
Join Date: May 2018
Posts: 259
@j82k

Thanks for clarifying and the explanation.
madjock is offline   Reply With Quote
Old 12th February 2019, 14:06   #54680  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
Quote:
Originally Posted by iSeries View Post
Hi j82k, thanks for this - I know you're likely to be busy but any way you could test a couple other driver versions? (I loaned my meter to someone). Specifically 398.11, which is apparently the last driver that is working fine for HDR passthrough (and conveniently the driver that brought improved 23p playback...I never could get any good timings with 385.28)
I'm already back to 385.28.

But it's easy to test even without a meter. I explained this before, don't know if you missed it. On the LG C8 the 20-point white balance adjustment points are different when the TV receives 1000 or 4000 nits metadata.

Just open madTPG and enable HDR. Go into your TVs 20 point white-balance settings and check the second last adjustment point. With 4000 nits metadata it should be 713. The number will change in real time when you switch between 1000 and 4000 nits in madTPG. If it doesn't show 713 when madTPG is set to 4000 it means the TV is getting the bogus metadata.

As for timings for 23p. On my 1050ti all I have to do in madVRs custom modes tool is pick EDID/CTA and change the pixel clock from 296.70 to 296.69 which gives me almost perfect 23.976 Hz. But of course that could be different on other GPUs.


Quote:
Originally Posted by sat4all View Post
@j82k
Thanks, that very helpfull, would you please install the latest test build along with nvidia 418.81 drivers and take measurements with those settings:

tone map HDR using pixel shaders
output video in HDR format = enabled
target / real display peak nits = 800
apply dynamic clipping = enabled
apply target nits selection = disabled

Because it seems ok for me and almost matchs correct playback using my tv embedded media player.

I tried messing with madVRs tone mapping in passthrough and it didn't make a difference on the grayscale sweeps. I was still on madVRhdrMeasure43 though.
But I think displaying single colored squares isn't really a good way to test dynamic tone mapping.
j82k is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

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 11:45.


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