madVR - high quality video renderer (GPU assisted)

[Manni]

You should be able to go back at any time by clicking on the custom mode, then pressing the "delete" button. At least that's how it's supposed to work.

Thanks, I wasn't sure it would do that or delete the mode, so instead, I did a clean install of the drivers to make sure I started with a clean slate.

Even with no custom res from the standalone experimental tool, the tab was missing.

So I connected the PJ directly to the GPU momentarily, and the tab appeared, so it looks like it doesn't like it when an AVR is in the chain, even if the AVR reports a populated EDID block.

I also tried 3D without FSE without an AVR in the chain and without any custom timings, and it definitely doesn't work here. The picture isn't corrupted/with weird colors as earlier, but it's 2D (no 3D at all) even if the OSD reports 3D windowed.

Hope this helps!

[mitchmalibu]

Thanks for the release madshi.

I did the first optimization run for the 2160p23 resolution. Everything went fine but when using the resolution and the same video file, windowed fullscreen reports a frame repeat every 5 minutes and fullscreen exclusive reports a frame repeat every 65/70 minutes. Clock deviations are the same.

Am I missing something ?

[JarrettH]

Edit: Now understand you need to switch to 60 hz before modifying 59 hz otherwise you can't customize a "standard mode" refresh rate.

madshi, thanks for this new version :sly:

I have a Dell U2417H and modes 50, 59, 60 under the status column are listed as standard modes (the OS reports them). When playing a 23.976 video madvr reports "display 59.933 hz" and "composition rate 59.940 hz". Since I already have these modes listed and not missing, should I further optimize the 59 hz one or is it already perfect? The EDID for 59 hz "results in 59.93939 hz". Just now I played a 23.976 video for 40 minutes uninterrupted resulting in 1 frame repeat every 6.50 minutes.

What do you mean by a 59p video? I've never played a 59 fps video in my life :P

https://c1.staticflickr.com/5/4431/36469621982_c94b68cce2_o.jpg

Thanks for clarifying.:thanks:

[egandt]

custom mode tutorial:

http://madvr.com/crt/CustomResTutorial.html

Also linked in first post of this thread.

I'm trying this, but I have a 4K TV with profiles and a 1080P monitor, however the custom dialog only appears for the 1080P monitor, and not the 4K TV (where it would count), with profiles, I see it appear and then disappear just as fast.

Thanks,
ERIC

[70MM]

Guys now that Im finally on Adaptive and have everything working perfectly I was wondering about dithering for 4K upscaling BDs...
I have the 1080Ti card and all my rendering times are now very very low.
These are my settings:
Chroma upscaling NGU sharp high
Luma upscaling NGU sharp high
render times 15.20ms

Dithering is set to Ordered dithering,
Should I use something higher like Error diffusion 1 or 2? since my render times are so low?
If so which one produces the best image?

[Asmodian]

There is no "best" when it comes to the two error diffusion options. I have been a fan of ED1 on my recent IPS monitors, on my older one I prefered ED2. However, on my new OLED TV I prefer ordered dithering, after a lot of testing. I didn't expect it but I have to believe my eyes. :)

I'm trying this, but I have a 4K TV with profiles and a 1080P monitor, however the custom dialog only appears for the 1080P monitor, and not the 4K TV (where it would count), with profiles, I see it appear and then disappear just as fast.

You need to open it on the TV itself.

[egandt]

There is no "best" when it comes to the two error diffusion options. I have been a fan of ED1 on my recent IPS monitors, on my older one I prefered ED2. However, on my new OLED TV I prefer ordered dithering, after a lot of testing. I didn't expect it but I have to believe my eyes. :)



You need to open it on the TV itself.

I've tried disabling the 1080P monitor (so everything opens on the TV and also setting to the primary in display properties, again so that madvr opens on the TV. On another system with 4 monitors all monitors show up when I select them no matter which it was opened on, but on the HTPC, I can not get custom to display for the TV, only for the monitor, it might have to do with profiles, but I have 8 and do not want to loss them unless there is no other choice.

Thanks,
ERIC

[Asmodian]

You can backup settings.bin and then reset to test it. I am not sure, it shows up on my TV but I don't have any profiles for displays. :o

[AngelGraves13]

Getting random crashes of madvr and lock ups of MPC-HC with the new madvr.

Also, deinterlacing doesn't work with the new DX11 renderer.

Using NGU for Chroma causes it to crash as well. Didn't try the upscaling as I never use anything but Jinc for all my scaling needs.

[QBhd]

Similar here. PotPlayer + madVR + LAV.... Files in play list play fine, but the player crashes (sometimes it's madVR) as soon as it moves on to the next file. I had to get up off the couch and close everything and restart playback again manually for the next file. I rolled back version and all is working fine again.

QB

[JarrettH]

Edit: Trying something...

[JarrettH]

Edit: I think Asmodian solved this on the next page

Well, I don't know what's wrong now. madvr is forcibly going into 59 hz no matter what. I've uninstalled it, reset default settings, made sure display mode switching is disabled, went back to 0.91.11. This only happens with full screen exclusive after the delay and doesn't revert to 60 hz when the player is closed. Display properties reports 59 hz while the nvidia control panel reports 60 hz. In windowed overlay it stays at 60 hz like it should.

Another observation

d3d9: switches to 59 hz in FSE, returning to windowed stays at 59 hz

d3d11: switches to 59 hz in FSE, returns to 60 hz in windowed

[Oguignant]

wow, this version is amazing...:thanks: "Let madVR decide" HDR mode works perfectly on my TV (active and deactivates great), also, the performances of the render times (with the same configuration as the previous version) are much better, 5 or 7 ms less. Excellent that 10bit option works also in full screen non-exclusive mode too.

The only thing bad, the display of my laptop will not let me add another mode. He says "the GPU driver rejected this mode, for unknown reasons". Is a Intel HD 630...

I have a question, how can I create a 3DLut for HDR? I use CalMAN Home Theater. Someone knows?

[Asmodian]

Well, I don't know what's wrong now. madvr is forcibly going into 59 hz no matter what.

Are you using the "hack Direct3D to make 24.000Hz and 60.000Hz work"? It is under devices -> display -> display modes.

[JarrettH]

Are you using the "hack Direct3D to make 24.000Hz and 60.000Hz work"? It is under devices -> display -> display modes.

OMG I love you Asmodian :D:D:D

I didn't use it because I didn't understand what it's for.

[Oguignant]

Any way to activate the TV HDR from madTPG?

[FDisk80]

Actually it *is* set to be always on top! However, in fullscreen, it's possible that the media player (or even madVR) might try to make sure that the video surface is always on top, and if multiple windows are on top, only one can win.


Yes, it's only when video is set to always on top but every other window that madVR opens doesn't have a problem with this and pops on top even if the video is set to be always on top except this one window.



Strange! I've just tested it here, and modifying the pixel clock works just fine here. Generally, the edit field is dropping back to the old value if interpreting your modified result failed for whatever reason. E.g. did you enter a "," for the decimal instead of a "."?


After deleting the custom mode from Nvidia Control Panel I tried again and this time it worked. Now it's in perfect 23.976hz mode (https://i.imgur.com/CLgGFUi.jpg). With no frame drops or repeats (https://i.imgur.com/yp2w3pf.jpg). Awesomeness! :)

[Asmodian]

I have a question, how can I create a 3dlut for hdr? I use CalMAN Home Theater. Someone knows?

I haven't been able to figure out how to get CalMAN to do it, all the HDR workflows are simply for adjusting display settings and it doesn't seem to pass any metadata to madTPG to trigger HDR. :(

If you enable Windows' HDR mode all the patterns rendered through madTPG are in the SDR range. Given the way HDR works there doesn't seem to be a good way to calibrate an HDR display for madVR with CalMAN.

[70MM]

There is no "best" when it comes to the two error diffusion options. I have been a fan of ED1 on my recent IPS monitors, on my older one I prefered ED2. However, on my new OLED TV I prefer ordered dithering, after a lot of testing. I didn't expect it but I have to believe my eyes. :)


I use a JVC X9500 proj on a 145" diag curved scope screen, I cant see any difference when I test Ordered dithering, ED1 or ED2. Can you explain what I should be looking for when I test which dithering is best to use please?

I can go right up close to the screen to look if I know what I should be looking for?

[Sunset1982]

Are you sure it's really 0.92.1? Did you replace "mvrSettings32/64.dll"? Maybe your browser gave you an old version of the zip?


Did a complete clean install, still the same problem. But I think it is the same problem like the one Manni has, as I am using an AVR between my HTPC and my TV too.

If I connect my htpc directly to my tv the display modes tab appears.

It's a Denon X4200W AVR in passthrough mode btw

[Anima123]

The new custom display modes does not work with Optimus yet?

When I'm trying to apply the modes, there's an error message with "The GPU driver rejected this mode, for unknown reasons". It behave the same way with the Intel's internal GPU.

[Oguignant]

Off topic. Hi, I need to buy a 4 "monitor 27" or more. Not for games, is for movis, image and content in general. What brand and model do you recommend? LG, Samsung, Dell?

I apologize, but I think here they should know a lot about quality monitors!

thks

[oldpainlesskodi]

@ Madshi - nice work with the latest version!

Anyone else using the latest dslayer 32bit getting a kodi crash popup notice when stopping playback (hdr/non-hdr)?

K

[mparade]

@madshi

do those MKV files have a 3D marker in the MKV header somewhere? Or am I supposed to guess the 3D based on the resolution alone?

Yes, they do in "video stereo mode" tab. Full SBS containers are set to 1 by the software automatically used for encoding to HEVC and remuxing. I think you can use the resolution alone as well because that one is also so unique for 3D SBS, only featuring this stereoscopic mode. The resolution of such files are ranging from 3840x800p to 3840x1080p depending on how much of cropping (due to black bars at the top and bottom) is required before encoding. If madvr had access to file names you could use that one as well, e.g. via the file name itself madvr could suppose the mode of the content: Croods (2013).3D.SBS.mkv. Maybe, the easiest way would be to check the file name first and process accordingly.

In my opinion it should also be considered that such files may be wanted to display/process as 2D (only one side of frame would be processed/displayed by madvr) if required. It would be an option for the user. He/She could watch such videos as 3D/2D depending on the equipment they have by them at the moment.

I would really appreciate your help.

Thank you very much for your efforts in advance.

[Asmodian]

I bet Nvidia's API's don't work on Optimus systems.

I use a JVC X9500 proj on a 145" diag curved scope screen, I cant see any difference when I test Ordered dithering, ED1 or ED2. Can you explain what I should be looking for when I test which dithering is best to use please?

Stick with ordered if you don't see any differences, at 8-bit any differences are very subtle, at 10-bit I have never noticed any differences at all. Also my verbiage for describing what I think the differences are starts sounding like audiophiles talking about sound which I find unhelpful and embarrassing. :o

Or, if you need extra heat in your projector room, set your display to a very low bit-depth, like 4-bit, and decide based on that (with the dithering trade quality for performance option disabled, that feature is much more important at low bit depths). :)

[Siso]

Is image doubling worth for 1.5 upscaling?

[nevcairiel]

Is image doubling worth for 1.5 upscaling?

Turn it on, judge image quality, decide? :)
We can't tell you if you will perceive a difference, only you can figure that out. It can greatly depend on the quality of your screen and your eyes. ;)

In theory there is a difference, but if its "worth" it is up to you.

[madshi]

how about some even more advanced test.

like frame skipping: https://www.testufo.com/#test=frameskipping

or a test to check if a TV has a properly working 23p mode.

you use the same test and check the brightness of the squares. it should be very obvious on a TV that is doing a 3:2 pullup.
Yeah, but the purpose of the tutorial is not to explain everything about everything. How to test if playback is smooth or not is a topic on its own. The tutorial only tries to explain how to create custom modes.

Even with no custom res from the standalone experimental tool, the tab was missing.

So I connected the PJ directly to the GPU momentarily, and the tab appeared, so it looks like it doesn't like it when an AVR is in the chain, even if the AVR reports a populated EDID block.
Ok, will have to look into that.

I also tried 3D without FSE without an AVR in the chain and without any custom timings, and it definitely doesn't work here. The picture isn't corrupted/with weird colors as earlier, but it's 2D (no 3D at all) even if the OSD reports 3D windowed.
OMGosh. I'm not sure if it's the new Nvidia drivers or Windows 10, but I can't get 3D to work *at all* on my PC now, using my Nvidia 1070. Activating 3D in the OS control panel doesn't do squat. Trying to activate 3D in the Nvidia control panel is painful, it opens up a new window, which then crashes. There's a reason why I recommend people to stick to Windows 8.1. Unfortunately nobody seems to be listening to me. Personally, I'd have loved to stay on Windows 8.1, but I had to switch to Windows 10 because of my commercial work.

When using the AMD GPU, 3D still works fine here, even in windowed mode. But in contrast to Windows 8.1, now it only works properly on the primary monitor. In Windows 8.1 is worked fine on secondary, too. Once again, Windows 8.1 beats Windows 10.

I did the first optimization run for the 2160p23 resolution. Everything went fine but when using the resolution and the same video file, windowed fullscreen reports a frame repeat every 5 minutes and fullscreen exclusive reports a frame repeat every 65/70 minutes. Clock deviations are the same.
After how many minutes of measurement do you get these results? These drop/repeat estimates need a long time to become reliable.

I'm not completely sure yet if custom modes are generally effective for D3D11 FSE mode. Have you tried D3D9 FSE mode?

When playing a 23.976 video madvr reports "display 59.933 hz" and "composition rate 59.940 hz". Since I already have these modes listed and not missing, should I further optimize the 59 hz one or is it already perfect? The EDID for 59 hz "results in 59.93939 hz". Just now I played a 23.976 video for 40 minutes uninterrupted resulting in 1 frame repeat every 6.50 minutes.
If your display can't do 23p, then you're probably using smooth motion? In that case custom modes are not needed for you, neither is Reclock.

I'm trying this, but I have a 4K TV with profiles and a 1080P monitor, however the custom dialog only appears for the 1080P monitor, and not the 4K TV (where it would count), with profiles, I see it appear and then disappear just as fast.
Does the "identification" tab for the 4K TV contain an EDID block with raw data? If so, is it directly connected to your PC, or is there an AVR in between?

on the HTPC, I can not get custom to display for the TV, only for the monitor, it might have to do with profiles, but I have 8 and do not want to loss them unless there is no other choice.
Are the profiles inside of the "device" section? Are they covering the "diaplay modes" tab?

Getting random crashes of madvr and lock ups of MPC-HC with the new madvr.

Also, deinterlacing doesn't work with the new DX11 renderer.

Using NGU for Chroma causes it to crash as well. Didn't try the upscaling as I never use anything but Jinc for all my scaling needs.
Similar here. PotPlayer + madVR + LAV.... Files in play list play fine, but the player crashes (sometimes it's madVR) as soon as it moves on to the next file.
Anyone else using the latest dslayer 32bit getting a kodi crash popup notice when stopping playback (hdr/non-hdr)?
Ok guys, but I can't magically fix crashes if I can't reproduce them, and if you don't provide any crash information! Do you get a crash box like this?

http://madshi.net/exc-ss3.gif

If so, please upload the crash report for me to look at. You can do that by pressing the "show bug report" button, then press "Ctrl+C" to copy the bug report to the clipboard.

"Let madVR decide" HDR mode works perfectly on my TV (active and deactivates great), also, the performances of the render times (with the same configuration as the previous version) are much better, 5 or 7 ms less. Excellent that 10bit option works also in full screen non-exclusive mode too.
Glad to hear that!

The only thing bad, the display of my laptop will not let me add another mode. He says "the GPU driver rejected this mode, for unknown reasons". Is a Intel HD 630...
Intel GPU driver bug, probably.

Any way to activate the TV HDR from madTPG?
Not yet.

Yes, it's only when video is set to always on top but every other window that madVR opens doesn't have a problem with this and pops on top even if the video is set to be always on top except this one window.
Strange. Will check if I can reproduce that.

After deleting the custom mode from Nvidia Control Panel I tried again and this time it worked. Now it's in perfect 23.976hz mode (https://i.imgur.com/CLgGFUi.jpg). With no frame drops or repeats (https://i.imgur.com/yp2w3pf.jpg). Awesomeness! :)
Great!

Did a complete clean install, still the same problem. But I think it is the same problem like the one Manni has, as I am using an AVR between my HTPC and my TV too.

If I connect my htpc directly to my tv the display modes tab appears.

It's a Denon X4200W AVR in passthrough mode btw
But the "display modes" is not listed twice, anymore?

It seems the custom modes tab currently doesn't like AVRs. Will have to fix that for the next build.

The new custom display modes does not work with Optimus yet?

When I'm trying to apply the modes, there's an error message with "The GPU driver rejected this mode, for unknown reasons". It behave the same way with the Intel's internal GPU.
I think Optimus always runs through the Intel GPU for final output, so this will only work if the Intel GPU drivers work correctly. Unfortunately they're mostly broken for custom resolutions. I'm trying to get them fixed, though.

Yes, they do in "video stereo mode" tab. Full SBS containers are set to 1 by the software automatically used for encoding to HEVC and remuxing.
Ok, that should make it easier. Well, your feature request has already been on my to do list for a long time, and of course it's still there. It's just not a high priority atm.

Can you provide me with a small sample? Then I'll have something to test with, whever I get around to implementing support for that.

[oldpainlesskodi]

"Ok guys, but I can't magically fix crashes if I can't reproduce them, and if you don't provide any crash information! Do you get a crash box like this?

http://madshi.net/exc-ss3.gif"

Yes, showing Kodi - extract from crash log for Kodi:


date/time : 2017-08-18, 09:47:04, 834ms
computer name : DESKTOP
user name : k
operating system : Windows 10 x64 build 14393
system language : English
system up time : 1 day 21 hours
program up time : 19 seconds
processors : 4x Intel(R) Core(TM) i5-2500K CPU @ 3.30GHz
physical memory : 11270/16351 MB (free/total)
free disk space : (C:) 230.43 GB
display mode : 3840x2160, 32 bit
process id : $1fbc
allocated memory : 624.13 MB
largest free block : 129798.70 GB
executable : kodi.exe
current module : MADHCNET64.DLL
module date/time : 2015-11-18 21:01
version : 1.0.17.12
compiled with : Delphi XE7
madExcept version : 4.0.13
callstack crc : $2246ee6b, $486a5104, $51cf3013
count : 10
madVR version : 0.92.1
exception number : 1
exception class : Exception
exception message : Zugriffsverletzung bei Adresse $4a4cd55e in Modul 'madVR64.ax'. Lesen von Adresse $13424a09244.

UpdateHdrPresentationThread ($153c):
4a4cd55e +1e madVR64.ax settings.cpp 4124 +2 CSettings.UpdateHdrPresentation
4a40f507 +67 madVR64.ax direct3d.cpp 4966 +8 UpdateHdrPresentationThread
7ff859ac +0e KERNEL32.DLL BaseThreadInitThunk
7ff85a63 +1b ntdll.dll RtlUserThreadStart



Using Google Translate, the error appears to be "Access violation at address $ 4a4cd55e in module 'madVR64.ax'. Read address $ 13424a09244"

Hope it helps, and I have tested on both 32 and 64 bit (64bit log extract above)

K

[mitchmalibu]

After how many minutes of measurement do you get these results? These drop/repeat estimates need a long time to become reliable.

I'm not completely sure yet if custom modes are generally effective for D3D11 FSE mode. Have you tried D3D9 FSE mode?

I let the movie run for 2/3min. The thing is that, while the movie is playing, if I switch from wfs to fse in the settings, the frame repeat/drop stat immediatly updates to a value coherent with the optimization done previously and with the clock deviation. It's as if the wfs mode wasn't taking into account the new custom mode, even if the madvr settings indicates it is currently active. I did not tinker with custom modes in the nvidia cpl, I'm currently on a fresh driver install.I do however have an AVR in my display chain, not sure if this could have an impact ..

I did not try to use D3D9 FSE. I will test further this weekend and report my findings.

[AngelGraves13]

* Easy way to reproduce player lockup is to skip to different timecodes of a file with the mouse over the timeline in MPC-HC and eventually the player locks up, and I have to close it from Task Manager or wait for it to stop responding. Latest Nightly MPC-HC and madvr.
* madvr crashed when selecting NNEDI3 Chroma. It said it uploaded when it opened the browser. Does it crash when you select that form of Chroma upscaling?
* DXVA11 and madvr may be having issues with each other
* DVXA11 (LAV) and Madvr doesn't support deinterlacing? It never turns on. As a result, there's also no IVTC.

Can you reproduce these madshi?

Thanks for your continued support!

[Oguignant]

I haven't been able to figure out how to get CalMAN to do it, all the HDR workflows are simply for adjusting display settings and it doesn't seem to pass any metadata to madTPG to trigger HDR. :(

If you enable Windows' HDR mode all the patterns rendered through madTPG are in the SDR range. Given the way HDR works there doesn't seem to be a good way to calibrate an HDR display for madVR with CalMAN.


Calman has the option to calibrate color space bt.2020 HDR and bt.2020 SDR and gammut formula ST.2084 hdr. Maybe using the source pattern of Calman can work. What you think?

[Oguignant]

Intel GPU driver bug, probably.


Is driver but modified by alienware ... maybe that is the problem. Do not try to run it with administrator privileges.

[Manni]

I haven't been able to figure out how to get CalMAN to do it, all the HDR workflows are simply for adjusting display settings and it doesn't seem to pass any metadata to madTPG to trigger HDR. :(

If you enable Windows' HDR mode all the patterns rendered through madTPG are in the SDR range. Given the way HDR works there doesn't seem to be a good way to calibrate an HDR display for madVR with CalMAN.

As long as you're able to switch your display to HDR manually (or using a device like the HD Fury Linker/Integral/Vertex) it's fine to use even 1080p SDR patterns to calibrate HDR with MadVR, as long as the correct gamut/gamma is selected in the PC (BT2020/ST2084) and the correct standards are selected in the software (BT2020/ST2084). I use Calman for this and can measure/calibrate HDR on my JVC RS500 without any problem. Of course you need to use a multiplier for the graphs to make sense, usually 6 to 8 works fine. I set this multiplier in the screen offset of Calman, it should work the same with other software.

When MadVR supports HDR, it will simply force the display to switch to its HDR mode and apply BT2020/ST2084, but there is nothing "magic" about the patterns as long as the software asks for the right ones and the display is set to the correct mode. Even 10bits or UHD isn't necessary for calibration, 1080p 8bits works fine, at least in my experience.


Ok, will have to look into that.


Did more tests, I think I found the reason.

If you specify that the AVR is an AVR, it then asks you to specify the display connected to it. This is what I usually do, and in that case the custom modes don't appear.

If I lie and tell MadVR that the AVR is a display, then the display modes appear twice (bug already reported) and the custom modes tab appears. I'm going to stay like that temporarily in order to test the custom modes.


OMGosh. I'm not sure if it's the new Nvidia drivers or Windows 10, but I can't get 3D to work *at all* on my PC now, using my Nvidia 1070. Activating 3D in the OS control panel doesn't do squat. Trying to activate 3D in the Nvidia control panel is painful, it opens up a new window, which then crashes. There's a reason why I recommend people to stick to Windows 8.1. Unfortunately nobody seems to be listening to me. Personally, I'd have loved to stay on Windows 8.1, but I had to switch to Windows 10 because of my commercial work.

When using the AMD GPU, 3D still works fine here, even in windowed mode. But in contrast to Windows 8.1, now it only works properly on the primary monitor. In Windows 8.1 is worked fine on secondary, too. Once again, Windows 8.1 beats Windows 10.


Glad to see I'm not the only one experiencing these issues :)

To save you time, I had the same issues with 378.92, which is as far back as I went (my 1080 Ti is my first nVidia card in a long time).

I also experience the crash with the 3D wizard, but I do manage to enable stereoscopic 3D after a few attempts usually. At some point, it crashes but stereoscopic 3D is enabled, and it works until it disables itself and you have to start again.

I was able to run the wizard fully once or twice, but most of the time it doesn't and simply crashes. I can't isolate why it's working sometimes and why most of the times it doesn't. I tried going back to 1080p23 before running the wizard, to use standard parameters instead of nVidia's (to get 12bits RGB Full etc), it doesn't seem to make any difference.

The OS 3D switch has to be enabled in order to get the wizard to run, but as you say that doesn't enables stereoscopic 3D in the driver.

I agree that 8.1 is a much better choice at this point in time for pure HTPC use, but losing DirectX12 is a big compromise for gamers (some games are exclusive to DirectX12, hence only work with Windows 10), so hopefully MS and nVidia will sort this out.

3D was much simpler to use with AMD, but I had issues with my HD7870, half the time 3D wouldn't work, both with PowerDVD and MPC-BE/LAV/MadVR, which was one of the reasons I decided to upgrade, beyond the lack of GPU power.

[nevcairiel]

* DVXA11 (LAV) and Madvr doesn't support deinterlacing? It never turns on. As a result, there's also no IVTC.

Both of those were mentioned in the release notes of madVR 0.92. Please do try to read at least those when testing a new version! :)

[el Filou]

To people reporting deinterlacing or IVTC not working with DX11VA: madshi stated in the release notes that he didn't have time to implement that yet.

Edit: replied at the same time as nevcairiel, a mod can delete my message if you want.

[Oguignant]

As long as you're able to switch your display to HDR manually (or using a device like the HD Fury Linker/Integral/Vertex) it's fine to use even 1080p SDR patterns to calibrate HDR with MadVR, as long as the correct gamut/gamma is selected in the PC (BT2020/ST2084) and the correct standards are selected in the software (BT2020/ST2084). I use Calman for this and can measure/calibrate HDR on my JVC RS500 without any problem. Of course you need to use a multiplier for the graphs to make sense, usually 6 to 8 works fine. I set this multiplier in the screen offset of Calman, it should work the same with other software.

When MadVR supports HDR, it will simply force the display to switch to its HDR mode and apply BT2020/ST2084, but there is nothing "magic" about the patterns as long as the software asks for the right ones and the display is set to the correct mode. Even 10bits or UHD isn't necessary for calibration, 1080p 8bits works fine, at least in my experience.


BT2020 HDR or BT2020 SDR?

To switch to HDR I play a movie with mpc-hc/madvr and minimize the window ... leave activated hdr on my LG TV

Of course you need to use a multiplier for the graphs to make sense

Multiplier for the graphs ?? what is this? An option in Calman?

[Manni]

BT2020 HDR or BT2020 SDR?

To switch to HDR I play a movie with mpc-hc/madvr and minimize the window ... leave activated hdr on my LG TV



Multiplier for the graphs ?? what is this? An option in Calman?

Both HDR BT2020 and SDR BT2020, you just need to select the correct gamut/gamma curve in the software and the correct user mode in the display manually. MadTPG works fine with both, even if the HTPC where MadVR is installed is set to 1080p SDR.

Discussing how to use a multiplier for projectors with HDR calibration is OT here, see the first link in the PQ Gamma calibration section in the first post of this thread:
http://www.avsforum.com/forum/24-digital-hi-end-projectors-3-000-usd-msrp/2246658-jvc-calibration-software-v6-2015-models-x9000-x7000-x5000-rs400-rs500-rs600.html
You don't need to use a multiplier if your TV is able to output at least 500nits or so. The multiplier is only for projectors, that often top at 200nits or so max.

[Manni]

Did more tests with the AVR as a display, and sadly the custom modes tab seems to reject most relevant modes (GPU rejects the custom mode), especially 2160p23, which work fine with the standalone experimental tool when the AVR is set to AVR and the PJ as a PJ.

I noticed that the only reported difference when the AVR is set to a display is that the native resolution is 720p (vs 2160p when the JVC is set to be the display). This leads the custom modes tab to only display 720p modes until you uncheck the display native modes only. Otherwise, all the other attributes are the same.

[AngelGraves13]

Ok...I'll have to read the full release notes rather than what's up on the front page then...

[Oguignant]

Both HDR BT2020 and SDR BT2020, you just need to select the correct gamut/gamma curve in the software and the correct user mode in the display manually. MadTPG works fine with both, even if the HTPC where MadVR is installed is set to 1080p SDR.

Discussing how to use a multiplier for projectors with HDR calibration is OT here, see the first link in the PQ Gamma calibration section in the first post of this thread:
http://www.avsforum.com/forum/24-digital-hi-end-projectors-3-000-usd-msrp/2246658-jvc-calibration-software-v6-2015-models-x9000-x7000-x5000-rs400-rs500-rs600.html
You don't need to use a multiplier if your TV is able to output at least 500nits or so. The multiplier is only for projectors, that often top at 200nits or so max.


Yeah, I realized I was talking about projectors. My TV is OLED, close to 500 nits. Luckily, I said to myself: One more thing I have to learn .... "Just when I thought I was out, they pull me back in." :)

[mv]

I haven't been able to figure out how to get CalMAN to do it, all the HDR workflows are simply for adjusting display settings and it doesn't seem to pass any metadata to madTPG to trigger HDR. :(

If you enable Windows' HDR mode all the patterns rendered through madTPG are in the SDR range. Given the way HDR works there doesn't seem to be a good way to calibrate an HDR display for madVR with CalMAN.

I've compared Calman HDR10 Workflow readings with profiled i1D3 between madTPG+manual HDR switch (launched a HDR video file and then collapsed the player window) and calibration patterns from R.Masciola's HDR-10 UHD Test Patterns disk. Result were pretty close not 100% identical though.

Even took a crazy attempt to create a 3DLUT but of course it was a total failure because of the display tonemapping (LG OLED) which couldn't be disabled at the moment.

[Manni]

Went as far back as 378.78, which is the first driver supporting my 1080 Ti, after a clean uninstall with DDU and a clean install of the drivers.
Same crash with the 3D Vision wizard. So it doesn't seem to be a driver issue, maybe an OS update issue? I hope you can get your nVidia contact to look at this, as you're able to reproduce in your system.

[madshi]

Yes, showing Kodi - extract from crash log for Kodi
Thanks, I have a guess now what could cause this issue.

I let the movie run for 2/3min. The thing is that, while the movie is playing, if I switch from wfs to fse in the settings, the frame repeat/drop stat immediatly updates to a value coherent with the optimization done previously and with the clock deviation. It's as if the wfs mode wasn't taking into account the new custom mode, even if the madvr settings indicates it is currently active. I did not tinker with custom modes in the nvidia cpl, I'm currently on a fresh driver install.I do however have an AVR in my display chain, not sure if this could have an impact ..

I did not try to use D3D9 FSE. I will test further this weekend and report my findings.
Hmmm... So you're saying the custom mode seems to be effective for D3D11 FSE, only, but not for fullscreen windowed mode? How's non-fullscreen windowed mode? Would be nice if you could test all the variants and report which modes are effected by the custom mode and which are not. Please reboot the OS, just to be safe, before you do the tests. And it might make sense to make the custom mode extra weird, to get totally clear results. E.g. make it 23.700Hz instead of 23.976Hz, or something like that (if your display syncs to 23.700). Thanks!

* Easy way to reproduce player lockup is to skip to different timecodes of a file with the mouse over the timeline in MPC-HC and eventually the player locks up, and I have to close it from Task Manager or wait for it to stop responding. Latest Nightly MPC-HC and madvr.
First of all please don't make all your post bold. That's very annoying. I almost considered not replying because of that.

Anyway. Nobody else has reported this specific problem yet, as far as I remember. Is it a new issue with v0.92.1? Or did it occur previously, too? Is this with the new D3D11 DXVA decoder? Or is it independent of which decoder you're using in the LAV settings?

* madvr crashed when selecting NNEDI3 Chroma. It said it uploaded when it opened the browser.
Did you get this crash window?

http://madshi.net/exc-ss3.gif

If so, I need the crash report.

Is driver but modified by alienware ... maybe that is the problem. Do not try to run it with administrator privileges.
If you look at the bottom of the tutorial, you'll see that the Intel drivers generally have pretty big issues with custom modes atm.

If you specify that the AVR is an AVR, it then asks you to specify the display connected to it. This is what I usually do, and in that case the custom modes don't appear.

If I lie and tell MadVR that the AVR is a display, then the display modes appear twice (bug already reported) and the custom modes tab appears. I'm going to stay like that temporarily in order to test the custom modes.
Ah, good to know! Will test that...

Did more tests with the AVR as a display, and sadly the custom modes tab seems to reject most relevant modes (GPU rejects the custom mode), especially 2160p23, which work fine with the standalone experimental tool when the AVR is set to AVR and the PJ as a PJ.
That is *VERY* surprising to me because the basic code to create a custom mode hasn't changed at all. Are you sure that the standalone tool still works? FWIW, the standalone tool doesn't even look at any madVR settings, so it doesn't care if you have the AVR set as AVR or as a display.

I've found that sometimes I was able to create 23p overrides, but then at some point it stopped working for me completely. So I wonder if now maybe your driver got into a state where you can't define 23p overrides, anymore, either? At all, I mean? So please double check if the standalone tool really still works. Also interesting would be to check if the Nvidia custom res control panel works. In my situation, whenever the "custom modes" tab isn't able to create a custom mode, the Nvidia custom res control panel fails, too. Well, it doesn't report any errors, but it also doesn't create any custom modes.

I noticed that the only reported difference when the AVR is set to a display is that the native resolution is 720p (vs 2160p when the JVC is set to be the display). This leads the custom modes tab to only display 720p modes until you uncheck the display native modes only. Otherwise, all the other attributes are the same.
That means that madVR is somehow interpreting a different EDID block. Does that make any sense to you? Are there more than just one "identification" cards for your AVR and/or PJ?

Could this maybe be the EDID block your AVR is sending out when no display is connected? Maybe the AVR then reports 720p as the native resolution? If you click on the "identification" card, you should see detailed information about what information the EDID contains.

P.S: Now that you declared your AVR to be a display, you might actually create custom modes in the wrong place, but I'm not completely sure. Please make sure that you edit the display which is "active" (should be shown in black in the settings dialog, while non-active displays are shown in gray). Also please double check that the EDID mentions JVC instead of the AVR manufacturer. You should get an JVC EDID block if you connect the JVC directly to the HTPC. I know, it's all a bit confusing. Hopefully it will get better when I fix the problem with the custom mode tab disappearing when you use an AVR.

Went as far back as 378.78, which is the first driver supporting my 1080 Ti, after a clean uninstall with DDU and a clean install of the drivers.
Same crash with the 3D Vision wizard. So it doesn't seem to be a driver issue, maybe an OS update issue? I hope you can get your nVidia contact to look at this, as you're able to reproduce in your system.
My Nvidia contact is only replying once in a while, and I'm afraid of asking too much of him. He seems to be mostly responsible for NVAPI kind of stuff, so 3D might not be his area of interest. I'd rather not bother him with any new problems, at least until the custom mode problems are solved.

[Oguignant]

In some HDR movies, the 3DLut file does not take effect. Even if I press "ctrl + alt + shift + 3" the screen is not divided into 2. Why?

movies whit matrix & primaries like this
http://i.imgur.com/VB362uS.png

[Neo-XP]

Getting random crashes of madvr and lock ups of MPC-HC with the new madvr.

Similar here. PotPlayer + madVR + LAV.... Files in play list play fine, but the player crashes (sometimes it's madVR) as soon as it moves on to the next file. I had to get up off the couch and close everything and restart playback again manually for the next file. I rolled back version and all is working fine again.

exception message : Zugriffsverletzung bei Adresse $4a4cd55e in Modul 'madVR64.ax'. Lesen von Adresse $13424a09244.

Same here, madVR v0.92.1 is crashing randomly when loading next file in MPC-BE. I have never had this issue with versions < v0.92.0.

[chros]

If you look at the bottom of the tutorial, you'll see that the Intel drivers generally have pretty big issues with custom modes atm.
It "seems" that they finally added support for custom res:
https://communities.intel.com/thread/78158

Unfortunately only from 4th gen, so no HD4000 support, f***ck :)

[Manni]

That is *VERY* surprising to me because the basic code to create a custom mode hasn't changed at all. Are you sure that the standalone tool still works? FWIW, the standalone tool doesn't even look at any madVR settings, so it doesn't care if you have the AVR set as AVR or as a display.


I did more investigation and realized that my HDMI Zone 2 in the AVR was on. I switched it off and the EDID discrepency went away. With it, the ability to use the custom modes tab with all my 4K modes came back. It worked fine and it gave me one frame drop every 1 hour.


I've found that sometimes I was able to create 23p overrides, but then at some point it stopped working for me completely. So I wonder if now maybe your driver got into a state where you can't define 23p overrides, anymore, either? At all, I mean? So please double check if the standalone tool really still works. Also interesting would be to check if the Nvidia custom res control panel works. In my situation, whenever the "custom modes" tab isn't able to create a custom mode, the Nvidia custom res control panel fails, too. Well, it doesn't report any errors, but it also doesn't create any custom modes.


I think this discrepency was due to the Zone2, now that the native res is correctly detected there doesn't seem to be any difference between the standalone and the custom mode tab tool.


That means that madVR is somehow interpreting a different EDID block. Does that make any sense to you? Are there more than just one "identification" cards for your AVR and/or PJ?


There is only one identification card for my AVR or PJ, unless I put something in between like an Integral/Linker/Vertex, in which case it can create a new one. There is only one active one at any time of course. I usually delete the old ones once my chain is stable. It is now, so I only have one ID card per device.


Could this maybe be the EDID block your AVR is sending out when no display is connected? Maybe the AVR then reports 720p as the native resolution? If you click on the "identification" card, you should see detailed information about what information the EDID contains.


Solved with Zone 2 switched off, but yes I was talking about the only difference in the ID cards between the two devices being 720p vs 4K for native resolution, probably because the PJ had been read when Zone 2 was switched off (as it should be when I'm testing, otherwise it adds another unnecessary degree of complexity).


P.S: Now that you declared your AVR to be a display, you might actually create custom modes in the wrong place, but I'm not completely sure. Please make sure that you edit the display which is "active" (should be shown in black in the settings dialog, while non-active displays are shown in gray). Also please double check that the EDID mentions JVC instead of the AVR manufacturer. You should get an JVC EDID block if you connect the JVC directly to the HTPC. I know, it's all a bit confusing. Hopefully it will get better when I fix the problem with the custom mode tab disappearing when you use an AVR.


I always edit the active display, and all this was solved with switching off zone 2 in the AVR.


My Nvidia contact is only replying once in a while, and I'm afraid of asking too much of him. He seems to be mostly responsible for NVAPI kind of stuff, so 3D might not be his area of interest. I'd rather not bother him with any new problems, at least until the custom mode problems are solved.

I totally understand, no worries.

Now that the custom modes tab works with 4K23, I've tried to optimize it, unfortunately MadVr doesn't register the playback of the video, and keeps saying timings not measured, so optimize always produces the same results (it tells you what to do and to come back there after playing).

I only have one display connected/detected, so everything always takes place on the same display. I tried leaving the optimize window open or closing it before playing the file, it doesn't seem to make a difference. Of course I play a file with audio in the same resolution/refresh rate as the mode I'm optimizing.

It would be great if the OSD in MadVR could display an indicator when it's "recording" the timings data, that way we would know if it's working right away before we spend 10/30mn playing the file.

Any suggestions regarding what may prevent MadVR from recording the data during playback? I tried playing an upscaled 1080p23 bluray and a 2160p23 file, same results. The resolution I'm trying to optimize is 2160p23, and I kind of know it's the same as my dropped frames go from 3-5mn (no custom res) to 1 hour (non-optimized custom res with EDID timings).

Note: all this testing was done after clean re-install of the latest nVidia drivers, as the older ones make no difference re the non-MadVR related 3D Wizard crashing bug.

[mitchmalibu]

Hmmm... So you're saying the custom mode seems to be effective for D3D11 FSE, only, but not for fullscreen windowed mode? How's non-fullscreen windowed mode? Would be nice if you could test all the variants and report which modes are effected by the custom mode and which are not. Please reboot the OS, just to be safe, before you do the tests. And it might make sense to make the custom mode extra weird, to get totally clear results. E.g. make it 23.700Hz instead of 23.976Hz, or something like that (if your display syncs to 23.700). Thanks!

I will try to get as exhaustive as possible in my next batch of tests. On a side note, it already was the case while trying the standalone tool.

[Sunset1982]

But the "display modes" is not listed twice, anymore?

It seems the custom modes tab currently doesn't like AVRs. Will have to fix that for the next build.


No it was shown twice only one time. Now the second is gone and only one is there. But the custom res tab is still not there... seems to be a problem with avr's.


Thank you! :)