madVR - high quality video renderer (GPU assisted)

[chros]

chros, thanks for linking your post. I'll definitely give it a read.
One more time, which display (type) do you have with which black level? :)

[aufkrawall]

One more time, which display (type) do you have with which black level? :)
Typical IPS with ~1000:1. Though old Spyder 4 is said to be bad when it comes to measuring exact brightness differences and only measures ~770:1. The black level that is registered in the ICC profile should be around 0.11cd/m².

[cosmitz]

There shouldn't be any conflicts, no triple buffering, vsync app-based, no adaptive sync, maximum performance, and the issue has been there for relatively forever, not even sure i can trace it to anything changed in particular.

And here i am back again a year later, i'm determined to either solve this or try my luck with a non madVR solution.

Long story short, W7, no Aero, MadVR/MPCHC in exclusive fullscreen mode has a few frames of audio desync. Only happens in exclusive mode, only with madVR. The audio comes after the video, roughly about 3-4 frames in motion (at 23.98, so 120-170ms or so). Can anyone seriously help me solve it this time? Otherwise i think i'll just give up on madVR and just try my luck with smoothing motion somewhere else.

[chros]

Typical IPS with ~1000:1. Though old Spyder 4 is said to be bad when it comes to measuring exact brightness differences and only measures ~770:1. The black level that is registered in the ICC profile should be around 0.11cd/m².
Ok, then it's pretty bad, I bet it's even worse, mine was 0.14 (CCFL IPS). Read my link, there is an example for that.
In short, forget Rec1886, use relative gamma 2.4 with 98% black output offset for the evening.

[aufkrawall]

Thanks four your input! I'll give it a try on my notebook display first, which has way worse contrast.

Now I wonder if I have to re-run calibration from scratch for corresponding 1D LUT, or if there is some argyllcms shortcut (as DisplayCal doesn't seem to offer 1 click solutions for this).

[chros]

Thanks four your input! I'll give it a try on my notebook display first, which has way worse contrast.
It's not about contrast (balck level/highest brightness) but purely about black level, nothing else.

[LordX2]

Some updates:

I put my RTX 3080 into my HTPC - which allowed me to go back to driver version 456.71. This instantly brought my runtime until frame drop at 23.976 up from 50ish minutes to 1.15hours. So about an extra 15-20 mins. (The newer Nvidia drivers are worse for some reason in this regard!)

The RTX 3080 runs with all settings maxed on 4K HDR Tone Mapping @ 23.976 - as I would expect. 30-32ms processing times - NGU @ Very High, nothing checked on compromise, highest dithering option.

I downloaded some test videos of 4K HDR 60 just for fun. I was NOT able to run it with settings maxed. I had to check the Compromise on Tone and Gamut, and I think I lowered NGU Chroma from Very high to High. That kind of surprised me. I guess I just thought such a beast card would shred anything I threw at it. Getting to that 16.67ms cut off point is harder than it seems with such demanding content!

On the frame drop/repeat issue with MPC-HC - I spent some time, and found that the issue went away with version: 1.9.15 (Both 1.9.17 and 1.9.16 had the issue) This was true EVEN with having my tried and true 456.71 Nvidia driver installed - so not a driver issue there, something in the program or Lav Filter that changed that my system does not like (Make that both of my HTPC systems)!!

[flossy_cake]

And here i am back again a year later, i'm determined to either solve this or try my luck with a non madVR solution.

Long story short, W7, no Aero, MadVR/MPCHC in exclusive fullscreen mode has a few frames of audio desync. Only happens in exclusive mode, only with madVR. The audio comes after the video, roughly about 3-4 frames in motion (at 23.98, so 120-170ms or so). Can anyone seriously help me solve it this time? Otherwise i think i'll just give up on madVR and just try my luck with smoothing motion somewhere else.

I'm using a similar setup to you: W7 x64 SP1, latest K-lite "full" (https://codecguide.com/download_kl.htm) release which contains the latest MPC-HC x64 and MadVR.

Although I'm leaving Aero enabled and letting MadVR auto disable it for fullscreen exclusive mode, which I've set to d3d9 instead of 11.

I have noticed on certain files the audio will go way out of sync if I skip around to other parts of the video by clicking the progress bar at bottom of screen. However I can immediately fix it by issuing a jump forward/back (left/right arrows on keyboard, which I've got bound to mouse wheel up/down for convenience).

Prior to that I tried unticking "auto a/v sync detection" in LAV audio decoder, without success.

Of course there is also the option in LAV to set a negative audio delay, but I'm guessing that was the first thing you tried.

Failing that, is it possible to try using a different audio decoder than LAV? Or maybe rolling back to an older version of LAV?

[huhn]

Typical IPS with ~1000:1. Though old Spyder 4 is said to be bad when it comes to measuring exact brightness differences and only measures ~770:1. The black level that is registered in the ICC profile should be around 0.11cd/m².

when an bad IPS panel is calibrated it's not uncommon to end up with such a low CR. let me guess the white point is off?

the meter could be the issue to not sure why a spider 4 and older is even allowed to exist.

[cosmitz]

Failing that, is it possible to try using a different audio decoder than LAV? Or maybe rolling back to an older version of LAV?

I have resorted to setting about -150ms of delay, but if the issue is present only in exclusive mode madVR, kiiiinda would expect to solve it madVR side. I don't want to do audio sync guesswork.

[aufkrawall]

It's not about contrast (balck level/highest brightness) but purely about black level, nothing else.
Well, I'd say it's the other side of the same coin with IPS.

when an bad IPS panel is calibrated it's not uncommon to end up with such a low CR. let me guess the white point is off?

the meter could be the issue to not sure why a spider 4 and older is even allowed to exist.
No, native white point is ~perfect. There's nothing wrong with old Spyder 4 according to DisplayCal dev, except that it can't measure low brightness well (Spyder 5 can't either). It doesn't affect gamut mapping in any way.

[flossy_cake]

I have resorted to setting about -150ms of delay, but if the issue is present only in exclusive mode madVR, kiiiinda would expect to solve it madVR side. I don't want to do audio sync guesswork.

Just a random idea, if you are sending audio over HDMI, maybe it's an issue with the HDMI audio driver? Perhaps you could try rolling it back to the generic Win7 driver by uninstalling it from device manager. Or if you are already using the generic driver try the Nvidia/AMD one.

Also when in full screen mode does the display mode hz change and are you using custom timings for those modes which might be interfering with the audio clock somehow?

[huhn]

Well, I'd say it's the other side of the same coin with IPS.


No, native white point is ~perfect. There's nothing wrong with old Spyder 4 according to DisplayCal dev, except that it can't measure low brightness well (Spyder 5 can't either). It doesn't affect gamut mapping in any way.

why are you now talking about gamut when the topic was CR.
while also saying you can't read low brightness well one of the two numbers to calculated CR.

[aufkrawall]

What is your point exactly? BT.1886 gamma curve would also look bad with a newer colorimeter. My calibration reults with gamma 2.2/2.4 are fine (I never indicated they weren't) and again a more recent colorimeter might not change situation at all.

For instance just explain what you think would happen if the colorimeter could detect black level more precisely.

[huhn]

you talk about a low measured CR using a spyder.
so if the spyder is the issue here that's the sole reason the CR is that low because it isn't.

if the black floor is read to high bt 1886 will result in even more elevated black gamma which makes it worse then it already is. you can simply not use it because normal gamma calibration doesn't care as much.

not even sure the spyder is the issue here ANSI IPS black is not deep at all.

[Siso]

displaycal can create these 3DLUT they are static tonemapping LUTs so don't except anything like dynamic tone mapping from them.

you take bt 2020 PQ are source apply a tone curve of smpte 2084 to it and use the trade of setting and type in the true peak nits just like in madVR. define a target either measured or something like DCI P3 D65 bt709
after that you can load it in madVR under HDR.

Should I apply black output offset too? One more thing, what should I put in target peak nits? I have calibrated my monitor to 100 nits, at factory reset the nits were 189. So what should I put?

[huhn]

depends on your device i'm not going to tell you what gamma/offset settings is correct.

if you have 189 nits i guess should type that in for HDR->sdr 3D LUT.
i don't use that feature.

[Siso]

depends on your device i'm not going to tell you what gamma/offset settings is correct.

if you have 189 nits i guess should type that in for HDR->sdr 3D LUT.
i don't use that feature.

In the "tone map hdr using pixel shaders".

[flossy_cake]

[deleted - misunderstood Siso's comment]

[huhn]

In the "tone map hdr using pixel shaders".

that's where you can apply a tone mapping 3D LUT.
and the nit number should not matter with a HDR -SDR 3D LUT on that page you may need the release version of madVR. i don't test this feature the page in the current beta release looks like it is for HDR to HDR. madVR will most liekely see that it targets SDR and will not send HDR meta data so the nit number there does nothing.

someone that uses that feature should answer that not me.

[djsolidsnake86]

Last update is from Oct 1, 2021

https://www.avsforum.com/threads/improving-madvr-hdr-to-sdr-mapping-for-projector.2954506/page-681#post-61067550

how install this?
by overwriting last version, it always show 0.92.17

[jmonier]

how install this?
by overwrting last version, it always show 0.92.17

All of the versions in that thread are Test Versions, not official releases, so they all (and there are many) are 0.92.17.

[tp4tissue]

Well, I'd say it's the other side of the same coin with IPS.


No, native white point is ~perfect. There's nothing wrong with old Spyder 4 according to DisplayCal dev, except that it can't measure low brightness well (Spyder 5 can't either). It doesn't affect gamut mapping in any way.


Spyder 4 's organic filters are just like the spyder 5's

They're only good for ~3 years from the date of manufacturing. Then they're out.

At this point in time, all spyder4s go in the trash, they shouldn't be counted as calibration tools for ANY purpose. NOT even whitepoint.

Get yourself a i1d3 or i1studio, these are good for 5-10 years, glass encased filters.

Spyder4and 5 also can not read the darker zones, with displaycal (argyll) cms it creates clipping and posterization when creating 3dlut. Even the 1d lut (cal) is problematic.

I owned a spyder 5 for weeks and I couldn't figure out why the colors were so garbage. Then I got i1d3, just don't waste your time with datacolor products, they're just bad.

Even the new spyderx is Crap. Again with the dark read problems even with its new lens.

[Manni]

Spyder 4 's organic filters are just like the spyder 5's

They're only good for ~3 years from the date of manufacturing. Then they're out.

At this point in time, all spyder4s go in the trash, they shouldn't be counted as calibration tools for ANY purpose. NOT even whitepoint.

Get yourself a i1d3 or i1studio, these are good for 5-10 years, glass encased filters.

Spyder4and 5 also can not read the darker zones, with displaycal (argyll) cms it creates clipping and posterization when creating 3dlut. Even the 1d lut (cal) is problematic.

I owned a spyder 5 for weeks and I couldn't figure out why the colors were so garbage. Then I got i1d3, just don't waste your time with datacolor products, they're just bad.

Even the new spyderx is Crap. Again with the dark read problems even with its new lens.

Wrong again. Old Spyders are fine to be used for gamma auto cal with the JVC Autocal software. How correct they are for color isn’t relevant in that case.

[aufkrawall]

Spyder 4 's organic filters are just like the spyder 5's

They're only good for ~3 years from the date of manufacturing. Then they're out.

At this point in time, all spyder4s go in the trash, they shouldn't be counted as calibration tools for ANY purpose. NOT even whitepoint.

Maybe you just had bad luck with your display(s) and correction profiles. I couldn't get a really decent result either with the Spyder 4 and Qnix Korea import display (white balance off). Though it had no OSD for adjusting white point and had cold white + nasty red/cyan tint at the same time.

However, calibrations of all other displays are fine for me.

When the filters fade or otherwise degrade, you’ll most likely notice that the measured color gamut of a display seems to shrink, there may be side-effects like an overall color cast (which happens when the filters don’t degrade uniformly), which is also why whitepoint and overall color accuracy suffer.

https://hub.displaycal.net/forums/topic/colormunki-display-vs-i1-display-pro/#post-11998
I don't have any of such issues, it detects native 6500k white fine, gamuts look proper and I also achieve 99.4% DCI-P3. It's definitely good enough for my purposes. If I'm not mistaken, the Spyder 4 lifespan was extremely improved over that of 3 due to changing to a glass lense or similar.

[arrgh]

...I know... old stuff and I have already done and read some things...
in any case I wanted to reuse an old machine for a simple HTPC (you know: CO2-Footprint and so forth);
nominally the processor should be able to do it... but I get for 3DMVC an extremely jerky picture;
system: Win10 LTSC 1809; latest Intel driver for the graphics; madVR .17 and LAV 0.75.1; D3D11 (native) as decoder;
the pictures are taken (trough a 3D glass) from the same movie; one can see that the statistics are extremely fluctuating; the average rendering times seem to be ok; but the queues seem to be quite empty...
so, my question is is the CPU too week or do I have options to improve (a 4K movie seems to work ok with the settings...)
thanks

https://abload.de/img/20qjgx.jpg (https://abload.de/image.php?img=20qjgx.jpg)

https://abload.de/img/1i1jnk.jpg (https://abload.de/image.php?img=1i1jnk.jpg)

[huhn]

looks like a decoding issue.

[clsid]

H.264 MVC is decoded using Intel decoder (libmfxsw64.dll). There is a separate hardware decoding checkbox for that in LAV settings. It is disabled by default because of stability reasons.

[arrgh]

yes, I'm using version 3 (LAV download page) of the dll; it doesn't make a fundamental impact whether I checkmark or not "h264 MVC" decoding in D3D11: jerky playback, ...except that unchecked the average rendering times go up above 40 ms...

[berkkocaturk]

Hello madshi I have a question

MadVR's HDR to SDR tone mapping is great but my question is will SDR content look better than same thing in HDR converted to SDR or will they look the same?

[ryrynz]

Hello madshi I have a question

MadVR's HDR to SDR tone mapping is great but my question is will SDR content look better than same thing in HDR converted to SDR or will they look the same?Madshi won't answer you here in any short timeframe unfortunately. Short answer.. They won't look the same. Generally speaking HDR to SDR tone mapping isn't always the best option. You have to evaluate this on a case by case basis as the results are highly dependent on your peak brightness output and personal preference among a few other things.

[Charky]

Hello madshi I have a question

MadVR's HDR to SDR tone mapping is great but my question is will SDR content look better than same thing in HDR converted to SDR or will they look the same?

Madshi has already answered that question IIRC.

See the end of that post : https://www.avsforum.com/threads/improving-madvr-hdr-to-sdr-mapping-for-projector.2954506/post-58893766

Short answer : he believes madvr HDR tone mapping results in superior image quality than native SDR bluray.

[mclingo]

surely MADVR doesnt do any tone mapping for SDR, only for HDR, so SDR will just look the same?

[brazen1]

2160p SDR has more resolution than 1080p SDR so it doesn't look the same - it looks better on a 4k display. (hint - Open Matte)

[SamuriHL]

That's not tone mapping, though. That's upscaling and the two are quite different.

[Asmodian]

surely MADVR doesnt do any tone mapping for SDR, only for HDR, so SDR will just look the same?

The question is "Does HDR tone mapped to SDR look like native SDR?"

No, it does not. Tone mapping HDR to SDR does not look like native SDR, even ignoring any resolution differences. HDR is mastered differently and madVR's automatic tone mapping is not going to change that. After tone mapping some scenes mike look very similar to the SDR master, but some might look very different.

It also depends on the SDR and HDR masters, there are no standards or anything connecting the two. It all depends on the artistic intent of the director and colorist/mastering engineer(s).

That's not tone mapping, though. That's upscaling and the two are quite different.

Hopefully it isn't upscaling. The HDR tone mapped to SDR should still have a higher native resolution then the native SDR bluray. :)

[berkkocaturk]

Madshi has already answered that question IIRC.

See the end of that post : https://www.avsforum.com/threads/improving-madvr-hdr-to-sdr-mapping-for-projector.2954506/post-58893766

Short answer : he believes madvr HDR tone mapping results in superior image quality than native SDR bluray.

I have checked your link. I always thought that HDR to SDR conversion was done to reflect best image quality to people that doesn't have HDR. Turns out not.

Reason I asked this question was I noticed HDR conversion looked great with madvr. But I still thought studio will do a better job.

I will stick with madvr HDR to SDR conversion rather than full SDR.

Thanks.

[SamuriHL]

Bahahaha. Now that was the funniest thing I've read all day. The studios doing a good job at grading. Lol! It depends. If the director is directly involved in the grading maybe. Otherwise you end up with fury road or the Meg. [emoji16]

Sent from my SM-G998U1 using Tapatalk

[Alexkral]

Another thing to consider regarding this, is that that "rather poorly done" tone mapping that madshi talks about, is what we are most used to seeing. A clear example of this is fire and explosions, which with a bad tone mapping algorithm that damages the hue are usually seen in orange and yellow tones, while with a more correct tone mapping algorithm are seen as a mixture of red and white that doesn't look natural.

In fact, one way to classify tone mapping methods is based on their intent, and even those trying to preserve the original scene appearance have to make some compromises.

For example, the increase in contrast that occurs as a result of tone mapping can be easily imitated with an SDR source by using local contrast enhancement, which has always been advised to be used with care due to its potential to create images with an overly artificial appearance.

[SamuriHL]

We really need screens with higher nit levels so we can make "tone mapping" go away.

[ryrynz]

We really need screens with higher nit levels so we can make "tone mapping" go away.See you in about ten years. Most people here would've upgraded to HDR displays by then. I've started using the Netflix app on my TV it actually appears to provide smoother output than Windows/MadVR... the beginning of the end.

[SirMaster]

Lol, projectors are never going to be 1,000 nits let alone 10,000 nits.

Tone-mapping will never go away.

[SamuriHL]

Maybe not but what's going to happen is as the display get brighter, the need for insane tone mapping like taking 10000 nits down to 50 will become less and less.

[Siso]

Why is there a huge difference in rendering times for tonemaping between the official old build and for example beta 113? Because of the dynamic tonemaping?

[SamuriHL]

Oh yea. There's a ton more processing being done in the newer builds.

[Siso]

Oh yea. There's a ton more processing being done in the newer builds.

That's good but it is eating my card :D I'm new to the HDR game, and my goal is static HDR to SDR tonemaping.

[SamuriHL]

I don't remember all the settings off the top of my head, but there are ways to mitigate it and turn some stuff down.

[Siso]

I don't remember all the settings off the top of my head, but there are ways to mitigate it and turn some stuff down.

I tried that but still got high rendering times.

[Megalith]

What's the deal with HDR10+ and MKVs? I just tried playing back a file and madVR appears to support it, but is it actually just playing back regular, non-dynamic HDR? Or does HDR10+ simply work because it's not a proprietary mess like Dolby Vision?

EDIT: It must be the latter, since LG OLED displays apparently don't support HDR10+.

[LordX2]

So, after experimenting with my RTX 3080, and enjoying the old driver version etc, I have decided to try out a GTX 1080 Ti card (I want my 3080 back for gaming!!!).

I have enjoyed cranking up all the settings with my 3080 though, and don't want to downgrade those if possible.

Anyone here have a 1080 Ti? Can I still keep Chroma @ NGU AA VERY High for 4k (23,24,25fps) content? WITH HDR to SDR tone mapping going (NO compromises)?

On the 3080 I am getting between 28-33ms rendering times for those settings. Can anyone compare with 1080 Ti?