madVR - high quality video renderer (GPU assisted) [Archive] - Page 569

TheRyuu

6th March 2015, 21:42

All I've said was meant to be a joke, because your bug report happened to have the ID 0x100. But I was serious about not "fixing" your bug report 256. You had reported this earlier in this thread, and I had explained why it's the way it is, and why I don't plan on doing anything about it. My opinion hasn't changed about that.

It is such an insanely trivial fix I really don't follow as to what the problem is with your refusal to address it. It is literally adding two linker options do not affect performance or compatibility in any way. You have to remember that your modules are being loaded into a media player which:

May be playing untrusted content.
May be playing content over an internet connection in which it is inherently untrusted.

Binaries not linked with dynamicbase and which do not make use of ASLR are one of the single biggest problems that other Microsoft programs had to deal with over the years and it used to be one of the easiest ways to defeat ASLR. Internet Explorer is an example of this as are explorer extensions (the 7-zip extension is still an example of this).

Whatever argument you're trying to make against it isn't valid so I'm really struggling to understand your refusal here since it is such an insanely trivial and EASY fix. How long do you think it's going to take you to turn on these options in Visual Studio? 10 seconds? If that?

madshi

7th March 2015, 03:03

It is such an insanely trivial fix I really don't follow as to what the problem is with your refusal to address it. It is literally adding two linker options do not affect performance or compatibility in any way. You have to remember that your modules are being loaded into a media player which:

May be playing untrusted content.
May be playing content over an internet connection in which it is inherently untrusted.

Binaries not linked with dynamicbase and which do not make use of ASLR are one of the single biggest problems that other Microsoft programs had to deal with over the years and it used to be one of the easiest ways to defeat ASLR. Internet Explorer is an example of this as are explorer extensions (the 7-zip extension is still an example of this).

Whatever argument you're trying to make against it isn't valid so I'm really struggling to understand your refusal here since it is such an insanely trivial and EASY fix. How long do you think it's going to take you to turn on these options in Visual Studio? 10 seconds? If that?
We've already discussed this. Do we really have to do it again?

My reasons are:

1) If there was no potential for madVR to break with those flags enabled, then why do these flags even exist? Why doesn't the OS simply pretend the flags were there although they aren't? The flags are there because not every DLL is able to handle the consequences of what the OS will do if the flags are there. I've no idea whether madVR will have problems if those flags are enabled or not. Probably not. But why should I risk problems without getting any practical benefit? madVR won't get faster or run more stable if I enable those flags. In the best case scenario users won't notice any difference. In the worst case scenario there will be new bugs, crashes, whatever.

2) Only madVR.ax is written in MSVC++. All other executable files are written with Delphi, using different versions, not all of them supporting these flags.

3) madVR is still not even feature complete. There are much more important things I need to worry about. Feel free to ask again when I've reached v1.0, maybe I'll reconsider then, but before v1.0 I refuse to risk any problems (even if they're unlikely to occur) for anything that doesn't bring any noticeable benefits.

James Freeman

7th March 2015, 07:47

By the time madVR 1.0 is here, it'll need to support the new UHD standard too.

vivan

7th March 2015, 08:40

What is so special about UHD? madVR supports BT.2020 just fine.

ikarad

7th March 2015, 10:10

When you say "true 24p playback" is not nearly as smooth than madVR's smooth motion playback, then your "true 24p playback" chain is broken. That is probably caused by your display running at 60Hz, so you get the 3:2 pulldown judder. Please understand that the 3:2 pulldown judder is *not* the way the movie was meant to be seen. It's also *not* what you see in commercial movie theaters. Commercial movie theaters do not have 3:2 pulldown judder. The smoothness achieved by madVR's smooth motion algorithm exactly matches what you get with a perfect 24fps playback chain without any motion interpolation. Basically if you turn madVR's smooth motion algo on, you get exactly the same smoothness you have in a commercial movie theater. However, you do get some blurring added. But the blurring just makes the image softer, or introduces some slight ghost images, but it doesn't make motion smoother than it should be. Making motion smoother than it would be in a commercial movie theater is not within the technical capabilities of my algorithm.

.
With a 120hz monitor, it's a "true 24p playback" or there is 3:2 pulldown judder?

-----------------------------------------
Ah, I see. So the imperfect 3:2 pulldown used by the current madVR build doesn't really bother you at all. You just noticed it, correct? I think since the introduction of smooth motion FRC most users don't use madVR's pulldown, anymore. Instead most users either have a matching framerate / refresh rate setup, or they use smooth motion FRC. So although I'm aware of that the pulldown needs an improvement, I don't think it's very urgent...
What is 3:2 pulldown? Is it inverse telecine (IVTC)? Ivtc in madvr doesn't function? How make with a 1080i30 movie? Must I use ivtc in madvr or another ivtc filter like decomb with avisynth/ffdshow?

-----------------------------
In madvr 0.86.11, in windowed mode, fraps indicates 24fps and in fullscreen mode, madvr indicates 120fps. In mad 0.87.14 fraps indcates 120fPs and in fullscreen mode, madvr indicates 120fps.
Why in windowed mode, there is a change? How back to 25fps in windowed mode?
http://img11.hostingpics.net/thumbs/mini_303057Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=303057Sanstitre.jpg)
http://img11.hostingpics.net/thumbs/mini_959400Sanstitre4.jpg (http://www.hostingpics.net/viewer.php?id=959400Sanstitre4.jpg)
-------------------------------------------------
In madvr 0.87.14, there are these new options. Must I turn off or not for the best quality?
http://img11.hostingpics.net/thumbs/mini_716448Sanstitre3.jpg (http://www.hostingpics.net/viewer.php?id=716448Sanstitre3.jpg)
---------------------------------------------------
In madvr 0.87.14, there are theses new options. What are the best options for the best quality?
http://img11.hostingpics.net/thumbs/mini_165366Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=165366Sanstitre.jpg)

Must I use error diffusion- option 1 or 2? Must I check off use colored noise and change dither for every frame or not?

I have TN 6bits 1080p, nvidia gpu gtx780, intel i7 4.1ghz quad core and win7

If I have a ips 8bits, must I use dithering options?
-------------------------------------------
Before 0.87.14, I used enable automatic fullscreen exclusive mode. Is always usefull now or is it outdated?
http://img11.hostingpics.net/thumbs/mini_565748Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=565748Sanstitre.jpg)

Sorry for my noob questions and my poor english.
-----------------------------------------------
In chroma upscaling, between 0.86.13 and 0.87.14, a new option appears: nnedi3
What are the differences between jinc3 and nnedi3?
http://img11.hostingpics.net/thumbs/mini_578005Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=578005Sanstitre.jpg)

On bluray 1080p, nnedi3 is smooth until 64neurons. With 128neurons and 256 neurons, it's not smooth. With 4k movie like sintel, even with neurons at 16 it's not smooth. With jinc3 it's smooth even with 4k movie.

---------------------------------------------
What are these options? They don't exist with 0.86.13. I suspect it's your deband filter.
http://img11.hostingpics.net/thumbs/mini_411363Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=411363Sanstitre.jpg)
Must I turn on ?

If yes, what are the good options for 720p anime mkv? 1080p blueray?

huhn

7th March 2015, 12:32

With a 120hz monitor, it's a "true 24p playback" or there is 3:2 pulldown judder?

it should be 5:5 pulldown with a 120 hz screen and a normal 23/24p source. don't forget nearly all source are 23p (24000/1001) not 24p so using smooth motion on a 120 hz display is a very good idea.

What is 3:2 pulldown? Is it inverse telecine (IVTC)? Ivtc in madvr doesn't function? How make with a 1080i30 movie? Must I use ivtc in madvr or another ivtc filter like decomb with avisynth/ffdshow?

madVR IVTC works fine with 3:2 source and some other cadence. 1080i30 doesn't mean it is telecine could be interlaced too or things like 2:2 or even worse hybrid.

to use madVR IVTC press control+alt+shift+t until it shows source type:film or set deinterlacing to force film mode in the options.

In madvr 0.86.11, in windowed mode, fraps indicates 24fps and in fullscreen mode, madvr indicates 120fps. In mad 0.87.14 fraps indcates 120fPs and in fullscreen mode, madvr indicates 120fps.
Why in windowed mode, there is a change? How back to 25fps in windowed mode?
http://img11.hostingpics.net/thumbs/mini_303057Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=303057Sanstitre.jpg)
http://img11.hostingpics.net/thumbs/mini_959400Sanstitre4.jpg (http://www.hostingpics.net/viewer.php?id=959400Sanstitre4.jpg)

what does fraps has to do with this? 25 hz sounds strange to me on a 120 hz display.

In madvr 0.87.14, there are theses new options. What are the best options for the best quality?
http://img11.hostingpics.net/thumbs/mini_165366Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=165366Sanstitre.jpg)

Must I use error diffusion- option 1 or 2? Must I check off use colored noise and change dither for every frame or not?

I have TN 6bits 1080p, nvidia gpu gtx780, intel i7 4.1ghz quad core and win7
use what looks best to your eyes. it is really that simple.

because you use a 6 bit tn panel (most likely an benq or asus) set madVR to 6 bit for this display you find this option under devices -> "name of your display" -> properties. the choice of dither has a way bigger impact on low bit deep displays compared to high bit deep displays.

Before 0.87.14, I used enable automatic fullscreen exclusive mode. Is always usefull now or is it outdated?
http://img11.hostingpics.net/thumbs/mini_565748Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=565748Sanstitre.jpg)
because you have a windows 7 i highly recommend to use window 7 overlay mode or fullscreen exclusive. the windows 7 desktop composition is so broken with a two display setup or even with just one display. fullscreen exclusive mode and windows 7 overlay mode avoid these issue.

In chroma upscaling, between 0.86.13 and 0.87.14, a new option appears: nnedi3
What are the differences between jinc3 and nnedi3?
http://img11.hostingpics.net/thumbs/mini_578005Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=578005Sanstitre.jpg)

On bluray 1080p, nnedi3 is smooth until 64neurons. With 128neurons and 256 neurons, it's not smooth. With 4k movie like sintel, even with neurons at 16 it's not smooth. With jinc3 it's smooth even with 4k movie.

nnedi3 is a very demanding. it is an completely new resize algorithm introduced in madVR. it should have an edge over all other resizer. unlike the TAB based filter higher neuron just make the decision making better.

choice an option that looks best to your eyes and still works.

What are these options? They don't exist with 0.86.13. I suspect it's your deband filter.
http://img11.hostingpics.net/thumbs/mini_411363Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=411363Sanstitre.jpg)
Must I turn on ?

If yes, what are the good options for 720p anime mkv? 1080p blueray?
it's an debanding filter there to remove banding.

depends on the source quality a low quality source can look a lot better with high settings while high settings on a high quality source can harm picture quality. you can set it to low it can harm picture quality but that's rare. and again choice an option that looks good to your eyes.

ikarad

7th March 2015, 13:15

it should be 5:5 pulldown with a 120 hz screen and a normal 23/24p source. don't forget nearly all source are 23p (24000/1001) not 24p so using smooth motion on a 120 hz display is a very good idea.
it's not 23p but 23,976p. 23,976 * 5 = 119.88. 23*5 = 115
I ask if the difference of 0.10 harm the smoothness or not.

smooth smotion introduce interpolated image. It's bad.

madVR IVTC works fine with 3:2 source and some other cadence. 1080i30 doesn't mean it is telecine could be interlaced too or things like 2:2 or even worse hybrid.
I speak about 1080p24fps movie interlaced in 30 fps

what does fraps has to do with this? 25 hz sounds strange to me on a 120 hz display.
It's a mistake. The good number is 24 not 25.
I want to know why there is a difference in madvr behavior between 0.86.13 and 0.87 in window mode.

because you use a 6 bit tn panel (most likely an benq or asus) set madVR to 6 bit for this display you find this option under devices -> "name of your display" -> properties. the choice of dither has a way bigger impact on low bit deep displays compared to high bit deep displays.
With 6bits on my 6bits display, image is very ulgy.
With 8bits in my 6bits display, image is good.

With dithering option in dithering menu
6bits
http://img11.hostingpics.net/thumbs/mini_775345Sanstitre2.jpg (http://www.hostingpics.net/viewer.php?id=775345Sanstitre2.jpg)

8bits
http://img11.hostingpics.net/thumbs/mini_799058Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=799058Sanstitre.jpg)

With dithering option (error diffusion) and 6bits in option, it's like 8bits without error diffusion.
Dithering menu is not redundancy with native display bitdepth in device menu?

If I make 8bits + error diffusion, I can't harm picture quality?

it's an debanding filter there to remove banding.

depends on the source quality a low quality source can look a lot better with high settings while high settings on a high quality source can harm picture quality. you can set it to low it can harm picture quality but that's rare. and again choice an option that looks good to your eyes.
what means low settings because there are two option in deband filter menu?

huhn

7th March 2015, 14:33

it's not 23p but 23,976p. 23,976 * 5 = 119.88. 23*5 = 115
I ask if the difference of 0.10 harm the smoothness or not.[quote]
23p is ~23.976 it just the name for it under windows. and correctly it is 24000/1001 like a said before.

you will get repeated or dropped frame from time to time without SM or something like reclock. about every 42 sec.

[QUOTE]smooth smotion introduce interpolated image. It's bad.
no it doesn't because it is not an frame creation interpolate it just blends frames nothing else. don't compare it to SVP or TV interpolations it totally different.

if you see huge issue than there is something wrong in madVR or your display.

using madVR SM on a 120 HZ screen should be super save.
I speak about 1080p24fps movie interlaced in 30 fps

if it is 3:2 it should work perfectly with forced film mode.

It's a mistake. The good number is 24 not 25.
I want to know why there is a difference in madvr behavior between 0.86.13 and 0.87 in window mode.

i guess it's the new window mode that is presenting frames in advanced does it matter? is there a really problem with it?

With 6bits on my 6bits display, image is very ulgy.
With 8bits in my 6bits display, image is good.

With dithering option in dithering menu
6bits
http://img11.hostingpics.net/thumbs/mini_775345Sanstitre2.jpg (http://www.hostingpics.net/viewer.php?id=775345Sanstitre2.jpg)

8bits
http://img11.hostingpics.net/thumbs/mini_799058Sanstitre.jpg (http://www.hostingpics.net/viewer.php?id=799058Sanstitre.jpg)

With dithering option (error diffusion) and 6bits in option, it's like 8bits without error diffusion.
Dithering menu is not redundancy with native display bitdepth in device menu?

If I make 8bits + error diffusion, I can't harm picture quality?

it is double dithered first to 8 bit and than to 6 bit from your display. not an ideal solution but if it looks ok to you you are fine. dithering shouldn't be used 2 time this just adds noise.

it's not the same frame it's not possible to compare. and don't disable dithering when you set it to 6 bit the result will be terrible.

i can't look at frames at the moment if have terrible lightning conditions at the moment.

maybe your display has a problem with it what is it i have a asus ve248qe right here a very common 120/144 HZ display.

what means low settings because there are two option in deband filter menu?

the second option has to do with a detection of a "fade out/in" or something like this. these scene have usually a lot more banding than the rest so a different debanding strength can be used for this.

ikarad

7th March 2015, 15:35

no it doesn't because it is not an frame creation interpolate it just blends frames nothing else. don't compare it to SVP or TV interpolations it totally different.

if you see huge issue than there is something wrong in madVR or your display.
It's not a native framerate and it can allow artifact.
If you advance frame by frame, you can see that it's not good.

I have asked the question to madshi some months ago, he said that smooth motion on 120hz was useless.

If I activate in madvr and I select only if there would be motion judder without it, madvr indicates it's not turn on.

i guess it's the new window mode that is presenting frames in advanced does it matter? is there a really problem with it?
I just want to know and to ask mashi.

it is double dithered first to 8 bit and than to 6 bit from your display. not an ideal solution but if it looks ok to you you are fine. dithering shouldn't be used 2 time this just adds noise.
:confused:

You said that I can't turn on 6bits in devices menu and turn on random, ordered or error diffusion in dithering menu?

maybe your display has a problem with it what is it i have a asus ve248qe right here a very common 120/144 HZ display.
My display doens't have any problem. The bad quality at 6 bits is the same on two displays (vg236h asus tn 120hz and benq2420T 120Hz TN).

With 0.86.13, no problem in 6 bits.
Maybe before madvr 0.87 dithering was applied automatically even withtout this menu. I don't know but If yes, was it ordered or random?

ikarad

7th March 2015, 15:48

because you have a windows 7 i highly recommend to use window 7 overlay mode or fullscreen exclusive. the windows 7 desktop composition is so broken with a two display setup or even with just one display. fullscreen exclusive mode and windows 7 overlay mode avoid these issue.
.

Whate are the differences between 7overlay mode and fullscreen exclusive?
What is the best?

I notice in overlay mode fraps indcates 24FPs in windowed or fullscreen. In fullscreen exclusive, frap indicates 120 fps.

snowpiercer

7th March 2015, 15:55

E.g. the NVidia 750 might be a good affordable choice. It might even allow you to use JincAR. Not sure about NNEDI3, which is the best of the scaling algorithms, but also requires the most GPU power. For that the 750 might not be fast enough. Well, maybe it *is* fast enough for 576p25, I'm not sure.

The nvidia Maxwell 750 Ti can do all that just fine.
Usually rendering times with JincAR@3Taps on both luma and croma is about 10-16ms, depending on content (480p to 1080p @24-30fps).
Error Diffusion works fine too. Even on latest nvidia drivers both ED1 and ED2 work fine. No artifacts as far as i can tell.

NNEDI3 does work with the 750 Ti on 32 neurons for luma only (not with chroma together) on everything up to 720p@24fps with rendering times about 32ms.

1080p@24fps does not work with nnedi3 regardless of neuron count on the 750 Ti. Rendering times are ~50/60ms on 16/32 neurons.

Overall the gpu is very capable of everything madvr has to offer, except nnedi3 on higher resolution/refresh rate content over 720p@~30. Also 32 neuron count is probably the highest for this gpu regardless of content.

huhn

7th March 2015, 17:32

It's not a native framerate and it can allow artifact.
If you advance frame by frame, you can see that it's not good.
you can't find these frames from madVR if you advance from frame to frame it's not possible.

I have asked the question to madshi some months ago, he said that smooth motion on 120hz was useless.

If I activate in madvr and I select only if there would be motion judder without it, madvr indicates it's not turn on.

you need to use "... or if the display refreshrate is a an exact..."

i would like to read that post where he said it is "useless" for 120 hz X-) it is not as important for a 120 hz screen as it is for a 60 hz playing a 23p file.

if you have issue with it don't use it but it should work really really well with a 120 HZ screen.
You said that I can't turn on 6bits in devices menu and turn on random, ordered or error diffusion in dithering menu?
just never disable dithering. of cause you can use dithering for 6 bit and you should do that!! that the hole point of it!

i said it is not ideal to do it 2 times.
My display doens't have any problem. The bad quality at 6 bits is the same on two displays (vg236h asus tn 120hz and benq2420T 120Hz TN).

With 0.86.13, no problem in 6 bits.
Maybe before madvr 0.87 dithering was applied automatically even withtout this menu. I don't know but If yes, was it ordered or random?

not sure which version it was but older madVR version used random dithering and it was only possible to disable it under trade for quality.

it shouldn't be disabled anyway.

Whate are the differences between 7overlay mode and fullscreen exclusive?
What is the best?

I notice in overlay mode fraps indcates 24FPs in windowed or fullscreen. In fullscreen exclusive, frap indicates 120 fps.

they are totally different. with fullscreen exclusive madVR uses a directx 9 fullscreen surface to display frames and this way madVR is in full control presentation no desktop composition no nothing. it can black some games from starting and a lot of people don't like the transition as far as i know this is still the recommended mode.

the windows 7 overlay mode doesn't work with AMD because they don't support it but this mode is my personal favorite mode it avoids the desktop composition and does an outstanding job on dual monitor setup

ryrynz

8th March 2015, 03:18

The nvidia Maxwell 750 Ti can do all that just fine.
Also 32 neuron count is probably the highest for this gpu regardless of content.

I use 64 neurons just fine with SD content on it. 128 and 256 don't yield quite the same IQ improvement that 64 and 32 have over 16 in my experience.

Dion

8th March 2015, 07:50

Is madVR not able to display Dxtory Codec? I cannot get it to display. Works fine with EVR and Haali.

RamGuy

8th March 2015, 10:10

huhn

8th March 2015, 10:50

Is madVR not able to display Dxtory Codec? I cannot get it to display. Works fine with EVR and Haali.

do you have a sample madVR shouldn't really care about codecs?
please post your filter list. to do this please make a screen/copy a list to clip board and post it here like this:

Filters currently loaded:
- Default DirectSound Device
- madVR
- Audio Switcher
- DirectVobSub (auto-loading version)
- LAV Audio Decoder (internal)
- LAV Video Decoder (internal)
- LAV Splitter Source (internal)

my guess is that it is filter that transformes colorspaces.

Skwelcha

8th March 2015, 11:01

I have this weird problem since I reinstalled Windows, I can't activate the exclusive mode on my Sony TV. This worked fine before and now works on my two 23" 1080p monitors, but not on the Sony. I tried to change the resolution because I read somewhere that this helped someone and if I try a 720p resolution it actually works. But back on 1080p, the problem remains... I tried using a shortcut to force the exclusive mode as well, but it always shows "windowed mode" with ctrl + J without any error messages.

I attached a very short madvr log, since the attachment file size is so low... if it has to be longer, I can upload it somewhere else.

EDIT: fixed it myself, the windows 8.1 desktop and font scaling was the culprit

Patrik G

8th March 2015, 20:18

How much difference would you expect there to be between a GTX 770 2GB and a R9 290X 4GB? I'm debating whether to take the GTX 770 2GB from my server which don't need the performance any more and strap a NZXT Kraken G10 + Corsair H110 on it in order to cool it silently, or perhaps grabbing a Sapphire Tri-X R9 290X 4GB for cheap and toss a Corsair HG10 + Corsair H110 on it.

Why so upset with graphics cards ?
As i said
3D LUT calibration is what makes the real visible impact on picture quality not swapping graphics cards.

Your picture on the kuro will still be off even if you get triple gtx 980s and max out evrerything in madvr.

As 3D LUT calibration requires almost no knowledge in calibration as its adjusting 90℅ of the settings automatically im surprised that more people hasnt done it.
Set up the programs and place the colormeter and wait a few hours
no adjustments or calibration knowledge is needed.
After that you will have a Studio reference picture that no hired pro calibrator can achieve from just doing a regular calibration of the tv.

As i see it 3D LUT support in madvr is the biggest improvement over all other video renderers Not the scaling algorithms.

digitech

8th March 2015, 20:53

Why so upset with graphics cards ?
As i said
3D LUT calibration is what makes the real visible impact on picture quality not swapping graphics cards.

Your picture on the kuro will still be off even if you get triple gtx 980s and max out evrerything in madvr.

As 3D LUT calibration requires almost no knowledge in calibration as its adjusting 90℅ of the settings automatically im surprised that more people hasnt done it.
Set up the programs and place the colormeter and wait a few hours
no adjustments or calibration knowledge is needed.
After that you will have a Studio reference picture that no hired pro calibrator can achieve from just doing a regular calibration of the tv.

As i see it 3D LUT support in madvr is the biggest improvement over all other video renderers Not the scaling algorithms.

In my country colorimeters are really expensive, any tip where to get one online with the right price?

Asmodian

8th March 2015, 21:35

As 3D LUT calibration requires almost no knowledge in calibration as its adjusting 90℅ of the settings automatically im surprised that more people hasnt done it.
Set up the programs and place the colormeter and wait a few hours
no adjustments or calibration knowledge is needed.
After that you will have a Studio reference picture that no hired pro calibrator can achieve from just doing a regular calibration of the tv.

As i see it 3D LUT support in madvr is the biggest improvement over all other video renderers Not the scaling algorithms.

3DLUT calibration is a more complex topic than you make it sound.

If you have a display with a good factory calibration, at best creating a 3DLUT will have almost no effect at all. In that case I find Jinc resizing a much more significant improvement.

Also a 3DLUT does have downsides. The only way to change the color displayed is by reducing the intensity of primary colors. e.g. What was 0-255 for Green will become 0-238 or something. This reduces the intensity of 100% green which reduces contrast.

On one of my displays creating a 3DLUT improves the picture a lot while on another every 3DLUT I have ever created resulted in a worse picture. That one has a decent factory calibration so the colors are already good enough that the downsides from using a 3DLUT are worse than the benefits.

I am using Argyllcms with an i1d3. Maybe Calman or other expensive option could do a better job but the reduction in contrast is unavoidable and with a good factory calibration there simply isn't much for a 3DLUT to improve.

Dion

8th March 2015, 23:00

do you have a sample madVR shouldn't really care about codecs?

It wasn't able to render it. However I discovered MagicYUV which works with Dxtory and works better in general and madVR is able to render it.

baii

9th March 2015, 00:25

3dlut and calibration is more than just color; gamma and color temperature that actually fit your viewing environment. Decent factory calibration can not do bt1886, gamut clipping for wide gamut displays etc.

Point is, spending $100 on a calibrator probably improve IQ more than a extra $100 on GPU. If it is igpu+ 3dlut vs a dgpu then it probably is a valid argument.
How much native contrast ratio you lost?
In my experience, I lost about 10% if i use a calibration embed 3dlut . dispcalgui florech have a workflow that does not require calibration which may better preserve contrast ratio.

Sent from my 306SH

huhn

9th March 2015, 00:57

It wasn't able to render it. However I discovered MagicYUV which works with Dxtory and works better in general and madVR is able to render it.

would be nice if you have a look into it to make sure madVR is working as intended.

of cuase nice that you found a "workaround".

huhn

9th March 2015, 01:16

3DLUT calibration is a more complex topic than you make it sound.[quote]
it is way more simple than setting up a grey scale calibration but the results are still a lot better.

i mean you are doing a hole gamut mapping and you don't have to deal with a CMS with up to 21 points of calibration...

[QUOTE]If you have a display with a good factory calibration, at best creating a 3DLUT will have almost no effect at all. In that case I find Jinc resizing a much more significant improvement.

this display is not born yet.

Also a 3DLUT does have downsides. The only way to change the color displayed is by reducing the intensity of primary colors. e.g. What was 0-255 for Green will become 0-238 or something. This reduces the intensity of 100% green which reduces contrast.

that simply show how bad the green was before calibration it is clipping, to bright compare to the rest or just out of bt 709.

On one of my displays creating a 3DLUT improves the picture a lot while on another every 3DLUT I have ever created resulted in a worse picture. That one has a decent factory calibration so the colors are already good enough that the downsides from using a 3DLUT are worse than the benefits.

so you got higher delta E after 3d LUT calibration or is this your personal opinion and not the data from your meter?

I am using Argyllcms with an i1d3. Maybe Calman or other expensive option could do a better job but the reduction in contrast is unavoidable and with a good factory calibration there simply isn't much for a 3DLUT to improve.

let the delta E values decide this not the price of a software.

argyllCMS has an option to make sure the black level doesn't rise for the cost of precision.

don't forget 3 D LUT is doing a complete gamut mapping unlike a grey scale calibration.

Bulldogger

9th March 2015, 02:14

Warner306

9th March 2015, 03:38

How GPU-intensive is a 3D LUT? My display is calibrated close to a perfect grayscale, but I don't have a colorimeter. It would be the natural evolution to add a 3D LUT before I build a new computer.

How does using a 3D LUT compare to adding smooth motion, artifact removal or error diffusion dithering in terms of the GPU resources it consumes?

Bandits

9th March 2015, 10:16

Has anyone had good results running madvr on a surface pro 3? I have the I7 with the Intel 5000 and cannot keep the queues full. I only play direct Blu-Ray rips 1080/24. I use MPC-HC with LAV filters and madvr wirh default settings. I can change to ever presenter and everything runs fine. I do not use LAV in hardware mode so madvr has the video processor to itself. The movie's will play about 4 to 5 minutes then queues drop to 0.

I was hoping someone has figured out the best settings for this device and can help with what I need to get the best results.

6233638

9th March 2015, 11:39

Has anyone had good results running madvr on a surface pro 3? I have the I7 with the Intel 5000 and cannot keep the queues full. I only play direct Blu-Ray rips 1080/24. I use MPC-HC with LAV filters and madvr wirh default settings. I can change to ever presenter and everything runs fine. I do not use LAV in hardware mode so madvr has the video processor to itself. The movie's will play about 4 to 5 minutes then queues drop to 0.Sounds like thermal throttling if it happens after 4-5 minutes of playback. Try DXVA decoding and scaling.

e-t172

9th March 2015, 11:45

If you have a display with a good factory calibration, at best creating a 3DLUT will have almost no effect at all. In that case I find Jinc resizing a much more significant improvement.
this display is not born yet.

There are plenty of displays on the market (Dell UltraSharp displays such as U3014 or U2714 are good examples) that have a factory-calibrated sRGB preset. Since sRGB and BT.709/BT.1886 are almost identical to each other for typical 1000:1 IPS displays, DeltaEs below 1 (with BT.709/BT.1886 as reference) are common. In these cases, applying a 3DLUT is more trouble than it's worth - a before/after comparison will barely show any difference. And I speak from personal experience.

Also a 3DLUT does have downsides. The only way to change the color displayed is by reducing the intensity of primary colors. e.g. What was 0-255 for Green will become 0-238 or something. This reduces the intensity of 100% green which reduces contrast.
that simply show how bad the green was before calibration it is clipping, to bright compare to the rest or just out of bt 709.

That's not necessarily as bad as you seem to think. This just means the native white point of the display is different from the BT.709 white point (D65). Non-standard white points can be acceptable in some cases, thanks to the human eye's chromatic adaptation (http://en.wikipedia.org/wiki/Chromatic_adaptation) mechanism. Since adjusting the white point reduces contrast, tradeoffs have to be made between perfect white point and maximum contrast, and it's not always a good idea to blindly go for "perfect white point" in all cases.

That said, it is possible to tell Argyll to generate a 3DLUT while preserving the native white point of the display, so it is possible to have the best of both worlds.

so you got higher delta E after 3d LUT calibration or is this your personal opinion and not the data from your meter?

let the delta E values decide this not the price of a software.

DeltaEs are not the answer to everything. Because of the way DeltaEs are typically calculated (i.e. black compensation), it's possible to have a display that has a contrast ratio of 100:1 despite DeltaEs < 1. Needless to say, such a display would look terrible.

argyllCMS has an option to make sure the black level doesn't rise for the cost of precision.

So first you're saying that we need to look at DeltaE for the ultimate metric, but then you're pointing to an Argyll option that increase DeltaEs for the sake of more contrast? Which is it?

How GPU-intensive is a 3D LUT? My display is calibrated close to a perfect grayscale, but I don't have a colorimeter. It would be the natural evolution to add a 3D LUT before I build a new computer.

How does using a 3D LUT compare to adding smooth motion, artifact removal or error diffusion dithering in terms of the GPU resources it consumes?

My experience is that using a 3DLUT in madVR is quite cheap, and is usable even on low-end hardware. I usually don't see a noticeable decrease in performance after enabling a 3DLUT. They do use a somewhat large amount of video memory though, around 100M if you go for maximum accuracy.

huhn

9th March 2015, 13:28

There are plenty of displays on the market (Dell UltraSharp displays such as U3014 or U2714 are good examples) that have a factory-calibrated sRGB preset. Since sRGB and BT.709/BT.1886 are almost identical to each other for typical 1000:1 IPS displays, DeltaEs below 1 (with BT.709/BT.1886 as reference) are common. In these cases, applying a 3DLUT is more trouble than it's worth - a before/after comparison will barely show any difference. And I speak from personal experience.

the gamma curve is different and BT 1886 is not an simple gamma curve it changes from display to display and sRGB gamma is not a simple curve too but unlike bt 1886 it is always the same.

but how rare are these displays ? and for how long are they this good? and the most important part how common are they used with madVR?
That's not necessarily as bad as you seem to think. This just means the native white point of the display is different from the BT.709 white point (D65). Non-standard white points can be acceptable in some cases, thanks to the human eye's chromatic adaptation (http://en.wikipedia.org/wiki/Chromatic_adaptation) mechanism. Since adjusting the white point reduces contrast, tradeoffs have to be made between perfect white point and maximum contrast, and it's not always a good idea to blindly go for "perfect white point" in all cases.

That said, it is possible to tell Argyll to generate a 3DLUT while preserving the native white point of the display, so it is possible to have the best of both worlds.

this assumes the green is not clipping how do you make that sure? setting up a white point of D65 before calibration is something that should be done in my eyes and a display that loses CR in this process is pretty rare and if it loses notable CR it's a pretty bad one.

this hole CR over every has to stop the hole world is not about black level a very slightly risen black is not notable in normal contend

good LCD are usually crippled to 0.080 to 0.040 cm² and even displays like the Samsung UE65HU8500 with 0.079 black are treated like "high end displays"
if your display turn from 1000:1 to 100:1 after apply a 3D LUT you may should think about other way to calibrated your display(or a another display...) I can't disagree.

and there are always exceptions but what is the point of these because this doesn't cont for 100% of all displays just 98%?

DeltaEs are not the answer to everything. Because of the way DeltaEs are typically calculated (i.e. black compensation), it's possible to have a display that has a contrast ratio of 100:1 despite DeltaEs < 1. Needless to say, such a display would look terrible.

delta E values doesn't take CR into the calculations yes but 100 CR is more than totally unrealistic

So first you're saying that we need to look at DeltaE for the ultimate metric, but then you're pointing to an Argyll option that increase DeltaEs for the sake of more contrast? Which is it?
I just said it is possible to use not that it should be used. it's like the white point options you talked about before i guess.(I don't know this option and it doesn't make sense to me).

it's the users choice. I just said it to show that you can apply a 3D LUT without rising the black point. and if I understand this right this is not an argyllCMS option it is an part in dispcalGUI. so I'm wrong about argyllCMS.

e-t172

9th March 2015, 14:09

the gamma curve is different and BT 1886 is not an simple gamma curve it changes from display to display and sRGB gamma is not a simple curve too but unlike bt 1886 it is always the same.

I am well aware of that. And if you look at the sRGB curve and the BT.1886 curve for a typical 1000:1 IPS display, you will notice they are quite close to each other. Which is all I was saying.

this assumes the green is not clipping how do you make that sure?

If a channel is clipping, it means the display is set up wrong. It's easy to "make sure" simply by measuring it using something like ColorHCFR. It's easy to fix, too, using hardware controls that pretty much all displays (except laptops) have.

setting up a white point of D65 before calibration is something that should be done in my eyes and a display that loses CR in this process is pretty rare and if it loses notable CR it's a pretty bad one.

Well, that's, like, your opinion, man. Also, a display that loses CR during white point calibration is not rare at all; pretty much all displays I've came across do this to various degrees, unless they were accurately factory-calibrated to D65. In fact, it's mathematically impossible to change the white point of a display away from native without losing some CR.

this hole CR over every has to stop the hole world is not about black level a very slightly risen black is not notable in normal contend

Could you please pay some attention to your writing and produce sentences that make sense?

if your display turn from 1000:1 to 100:1 after apply a 3D LUT you may should think about other way to calibrated your display(or a another display...) I can't disagree.
delta E values doesn't take CR into the calculations yes but 100 CR is more than totally unrealistic

Of course. I've used an extreme example to better deliver my point, which is that display calibration often involves tradeoffs, and focusing on only one thing (i.e. DeltaE) is not a good idea.

and there are always exceptions but what is the point of these because this doesn't cont for 100% of all displays just 98%?

You and I clearly haven't been calibrating the same displays. I've calibrated a dozen or so displays of various types (TVs, PC monitors, laptops, projectors) over the years, and it is rare to find one where calibration doesn't involve making various tradeoffs between DeltaE, contrast ratio, white point chromacity, etc. in order to obtain the best performance. Again, it's not as simple as you think it is.

if I understand this right this is not an argyllCMS option it is an part in dispcalGUI. so I'm wrong about argyllCMS.

AFAIK dispcalGUI is a frontend to ArgyllCMS. So dispcalGUI cannot offer an option that Argyll doesn't have. Specifically, I believe you're referring to dispcal's -k option (http://www.argyllcms.com/doc/dispcal.html#k). Which, by the way, is again documented as a tradeoff.

Patrik G

9th March 2015, 16:19

In my country colorimeters are really expensive, any tip where to get one online with the right price?

i1 Display Pro is the most accurate colormeter if you dont want to rob a bank :)

155£
http://www.ebay.co.uk/itm/X-Rite-i1-Display-Pro-Monitor-Projector-Calibrator-Easy-Very-Accurate-NEW-/331252655324?pt=LH_DefaultDomain_3&hash=item4d203264dc

its cheap compared to a high end graphics card.

info
http://www.chromapure.com/newgear_display3.asp

The X-Rite Display 3 is a truly revolutionary product. It offers near professional performance at a price affordable to any enthusiast. In fact, its performance is so close to professional colorimeters, such as the Klein K-10 or X-Rite’s own Hubble, than it has become extremely difficult to justify purchasing such devices that come in at 10-20 times the cost of the Display 3.

If you have a display with a good factory calibration, at best creating a 3DLUT will have almost no effect at all.

it will trust me.
i already have a calibrated KRP-500A but the 3D LUT calibration made it like a new tv.

some displays are close to reference with their ISF or movie modes but they are still far off.
100% saturation points maybe accurate but thats just 6 points measured at the end of the gamut.
with 3D LUT you measure alot more points inside the gamut and the program you use also corrects them.

its a huge difference measuring 6 color points which is probably what they are doing at the factory compared to 9200 points.

you get colorgradings perfect which also ads more 3d depth to the picture.
the 3d lut not only corrects saturation it also corrects colorluminance and hue along with gamma and grayscale.

some info here on standard 1D LUTs and 3D LUTs
http://www.lightillusion.com/luts.html

I am using Argyllcms with an i1d3. Maybe Calman or other expensive option could do a better job but the reduction in contrast is unavoidable and with a good factory calibration there simply isn't much for a 3DLUT to improve.

same here
Argyllcms and i1d3
and no calman or lightspace will not do the job better than Argyllcms.
ther is a comparsion thread over at avs and the results from the different calibration programs is almost identical.

if you get reduced contrast you probably have some settings wrong from the beginning such as blacklevel or whitelevel (those must be calibrated first before making measurements for the 3D LUT) or videolevels that doesnt match the display.
there are also some other settings that can raise blacklevel or change gamma curve near black.
those are blackpoint compensation or black output offset.

i could probably make a guide for this but that will only be suitable for the Kuro Plasmas.

yok833

9th March 2015, 19:40

it will trust me.
i already have a calibrated KRP-500A but the 3D LUT calibration made it like a new tv.

Yes the results with the 3dlut is really amazing :) (thanks madshi) ... Unfortunately even with a colorimeter as the i1d pro you cannot be 100% sure of the results if you do not have a spectrophotometer .. A spectrophotometer is quite expensive but is essential with a colorimeter to generate a ccmx file and have a proper color correction...

e-t172

9th March 2015, 20:04

A spectrophotometer is quite expensive but is essential with a colorimeter to generate a ccmx file and have a proper color correction...

I wouldn't say it's "essential" unless you have really special requirements (e.g. color-focused professional setting). The Eye-One display pro is typically accurate to 1.1dE on white and 2.0dE on black (http://www.drycreekphoto.com/Learn/Calibration/MonitorCalibrationHardware.html#Accuracy), with 0.4dE of unit-to-unit variability (http://www.drycreekphoto.com/Learn/Calibration/MonitorCalibrationHardware.html#Variability). It also comes with calibration files so that the sensor can be adapted to work on a wide variety of light sources (e.g. CCFL backlights, LED backlights, wide gamut, etc.). In other words, it's good enough.

A spectrophotometer is overkill for the video playback use case; most likely you'll end up with a calibration that's even more accurate than the monitors the movie was produced on, which is pointless. Besides, when dealing with such small color differences, other factors come into play (such as viewing conditions) that are not (yet) standardized and thus unpredictable.

Asmodian

9th March 2015, 22:29

it is way more simple than setting up a grey scale calibration but the results are still a lot better.

i mean you are doing a hole gamut mapping and you don't have to deal with a CMS with up to 21 points of calibration...

this display is not born yet.

that simply show how bad the green was before calibration it is clipping, to bright compare to the rest or just out of bt 709.

so you got higher delta E after 3d LUT calibration or is this your personal opinion and not the data from your meter?

let the delta E values decide this not the price of a software.

argyllCMS has an option to make sure the black level doesn't rise for the cost of precision.

don't forget 3 D LUT is doing a complete gamut mapping unlike a grey scale calibration.

I have done 3DLUT creation on many displays, some are really hard to notice if the 3DLUT is active. My current decent factory calibration has an almost perfect BT.1886 shape gamma (with average of 2.21 :mad:). All gray-scale points are <1 dE, the worst color error measured is <5dE (primaries at ~5dE sideways :p). A 3DLUT moves it around, maybe improves max dE by up to 0.5 but afterwards nothing has noticeably improved, contrast is down a bit depending on the method (absolute, perception, luminance axis). I am not correcting white, of course, as it is already good. I see minor banding in ramp test patterns with the 3DLUT active but I don't without it. Using madVR's simple gamma adjustment from an "already calibrated BT.709 2.2" to a "pure power 2.4" looks better than any 3DLUT I have been able to make for that display.

Delta E isn't everything, at least not in my experience. How much contrast is going from 6dE to 5dE on an unusual orange worth? Any? What about visible banding? I have never been able to judge visible banding by looking at dE values.

Some displays need so much correction that wrong colors look better than the banding or loss of contrast from the 3DLUT. Some laptop displays have had primaries that are so bad you shouldn't even think of any type of calibration. A meter is not going to help these people.

All I am saying is that a 3DLUT is not magic, it is possible to have a display on which a 3DLUT will not help or even hurt. These displays are common. In my experience the displays that usually benefit the most from a 3DLUT are actually the better ones because the correction is not big enough to cause terrible side effects but manufactures don't usually do very good factory calibrations unless targeting graphic design.

3DLUTs can help a lot. They really improve the gamma and warm colors on my second monitor. It is amazing how much better it looks simply getting a good BT.1886 gamma curve. A 3DLUT made my wide gamut display usable for watching video. Even yCMS, before 3DLUTs were implemented, was a big improvement on that display. My 2010 LG plasma benefited a lot.

I was mostly wondering if Calman would generate 3DLUTs with less banding, that seems to be the bane of my existence with Argyll. I often get at least very slight banding in test patters.

That black level option was an excellent addition, I use it all the time as no one wants raised blacks for video. It can increase banding but it is (almost) always worth it.

AFAIK dispcalGUI is a frontend to ArgyllCMS. So dispcalGUI cannot offer an option that Argyll doesn't have. Specifically, I believe you're referring to dispcal's -k option (http://www.argyllcms.com/doc/dispcal.html#k). Which, by the way, is again documented as a tradeoff.

Actually dispcalGUI had (has?) a hacky option that modified files in between steps in the process before ArgyllCMS got -k. Using -k is recommended now.

it will trust me.
i already have a calibrated KRP-500A but the 3D LUT calibration made it like a new tv.

some displays are close to reference with their ISF or movie modes but they are still far off.
100% saturation points maybe accurate but thats just 6 points measured at the end of the gamut.

It isn't that simple on any and all displays, trust me. I am not basing this opinion off of 6 or 10 or whatever number of saturation points I might decide to measure.

Also 9200 patches is overkill in my experience, I notice diminishing returns when going over ~2500. A trick to reduce banding on some ill behaved displays is to use a fairly low patch count, you get higher dE's but it can be worth it.

I am not saying 3DLUTs are usually bad, usually they help. It is only that generating a good one for any and all displays is a tricky and trade-off ridden process that can fail such that you are better off not using a 3DLUT at all. You can get great results from a simple guide but sometimes you don't.

baii

9th March 2015, 23:05

I was referring to fhoech suggestion that
1. do not set target tone cuve.
2. make 3dlut w.o the use of the vgct tag

http://www.avsforum.com/forum/139-display-calibration/1471169-madvr-argyllcms-88.html#post27376514

Initially (almost two years:thanks: how time flies by), during calibration, a tone curve of srgb or rec 1866 is usually "recommended"

[Recommended] Calibrate the white point and neutral axis using ArgyllCMS dispcal. Since the Device Link will override the calibrated transfer curve characteristic of the display, there there may be no point in doing much more than a medium calibration, and choosing a standard that has a straight segment from black, such as L*a*b*, sRGB, Rec709 or SMPTE240 curve. The exact shape of the calibration curve is not critically important, as the profiling and 3dLut will set the final response. If this approach is taken, then the resulting calibration file should be provided to dispread as the -k parameter or -K parameter. See also below Choice of where to apply display per channel calibration curves.

Spectro is nice too have if you have some uncommon backlight system, old CCFL, RGB LED, maybe OLED . i1d3 factory spectral correction are pretty "good" for wled and GB-r.

e-t172

9th March 2015, 23:27

I have done 3DLUT creation on many displays, some are really hard to notice if the 3DLUT is active. My current decent factory calibration has an almost perfect BT.1886 shape gamma (with average of 2.21 :mad:). All gray-scale points are <1 dE, the worst color error measured is <5dE (primaries at ~5dE sideways :p). A 3DLUT moves it around, maybe improves max dE by up to 0.5 but afterwards nothing has noticeably improved, contrast is down a bit depending on the method (absolute, perception, luminance axis). I am not correcting white, of course, as it is already good. I see minor banding in ramp test patterns with the 3DLUT active but I don't without it. Using madVR's simple gamma adjustment from an "already calibrated BT.709 2.2" to a "pure power 2.4" looks better than any 3DLUT I have been able to make for that display.

When you say "BT.1886 with average of 2.21", do you mean 2.21 as the average (effective) gamma or the technical (exponent) gamma? It sounds like you're using BT.1886 with an average gamma of 2.21, which is wrong (at least in dark viewing conditions). You should use an average gamma of 2.4 as recommended in the Argyll docs (http://www.argyllcms.com/doc/Scenarios.html#TV1) (specifically, -Ib:2.4). I've even seen 2.5 being recommended by some for pitch black viewing conditions. This probably explains your observation that setting gamma to 2.4 in madVR looks better.

I was mostly wondering if Calman would generate 3DLUTs with less banding, that seems to be the bane of my existence with Argyll. I often get at least very slight banding in test patters.

That's an interesting point. Have you tried Argyll's -q option (http://www.argyllcms.com/doc/colprof.html#q)? It seems like reducing "quality" would reduce the number of points and result in a "smoother" LUT, thereby reducing banding.

I wonder if Graeme Gill (Argyll's author) really thought about banding issues though. I'm not familiar with Argyll's internal algorithms, but maybe colprof's 3DLUT generation could be improved through additional constraints that would prevent two LUT steps from getting too far from each other, at the expense of DeltaE. I suspect that banding is the direct result of the way interpolation is done between data points, and could maybe be alleviated by the use of a different interpolation function or different parameters.

That black level option was an excellent addition, I use it all the time as no one wants raised blacks for video.

Yeah. By default I find that Argyll is somewhat overzealous with blacks - I think people care a lot more about deep blacks then they care about blacks being slightly tinted, especially since it's quite hard to distinguish colors at the bottom of the luminance scale.

Asmodian

10th March 2015, 00:10

When you say "BT.1886 with average of 2.21", do you mean 2.21 as the average (effective) gamma or the technical (exponent) gamma? It sounds like you're using BT.1886 with an average gamma of 2.21, which is wrong (at least in dark viewing conditions).

That isn't what I am using, that is what it is before calibration. That was why :mad:. I would have preferred 2.4 of course. Everything else is good enough that madVR's simple gamma changer offers better quality than every 3DLUT I have tried. I did try really hard to get a good 3DLUT. I can spend weeks calibrating instead of watching. :o

That's an interesting point. Have you tried Argyll's -q option (http://www.argyllcms.com/doc/colprof.html#q)? It seems like reducing "quality" would reduce the number of points and result in a "smoother" LUT, thereby reducing banding.

I wonder if Graeme Gill (Argyll's author) really thought about banding issues though. I'm not familiar with Argyll's internal algorithms, but maybe colprof's 3DLUT generation could be improved through additional constraints that would prevent two LUT steps from getting too far from each other, at the expense of DeltaE. I suspect that banding is the direct result of the way interpolation is done between data points, and could maybe be alleviated by the use of a different interpolation function or different parameters.

A good idea, thanks. I have always been using -qh maybe -qm or even -ql would actually be better on those screens with too much banding after calibration. I will have to test it next time I find one. Maybe even this new monitor as it doesn't need much and banding has been the main downside with its 3DLUTs. :)

Yeah. By default I find that Argyll is somewhat overzealous with blacks - I think people care a lot more about deep blacks then they care about blacks being slightly tinted, especially since it's quite hard to distinguish colors at the bottom of the luminance scale.

A lot of color management is based off of experience calibrating a graphic design workflow to match printed content which is such a different use case that I think a lot of "the way it is done" is less than optimal for calibrating a video playback device. Graeme Gill does seem interested in developing new methods for video playback calibration given evidence another way is better. This is very tricky because all the big studies are so old, done with different goals, with different technologies, and it is all subjective. :scared:

ryrynz

10th March 2015, 06:17

Interesting discussion, but someone will mention the OT side of it, so might as well be me :P

Patrik G

11th March 2015, 00:08

I have done 3DLUT creation on many displays,

then you must have solved all of the problems you just described ;)

the 3D LUT for my KRP made a huge diffrence thats all i can say.
no banding
no raised blacklevels
no loss in contrast.

the KRP-500s in the pure mode generate one of the most accurate pictures out of the box for any tv.
still the picture is better with more 3d depth/better details and colors near black/better skintones than it was with just a regular calibration.

so the suggestion here is to get a real tv instead of those crappy LG plasmas you have :p
you will probably have a better result then.

Asmodian

11th March 2015, 00:49

then you must have solved all of the problems you just described ;)

the 3D LUT for my KRP made a huge diffrence thats all i can say.
no banding
no raised blacklevels
no loss in contrast.

the KRP-500s in the pure mode generate one of the most accurate pictures out of the box for any tv.
still the picture is better with more 3d depth/better details and colors near black/better skintones than it was with just a regular calibration.

so the suggestion here is to get a real tv instead of those crappy LG plasmas you have :p
you will probably have a better result then.

You seem to be misunderstanding me. 3DLUTs can be very helpful, I am sure it made your KRP-500 look better. The 3DLUT I created for my "crappy LG plasma" was excellent and the picture looked much better than without it.

On some displays this is not the case, either because they are already good enough a 3DLUT doesn't help (rare) or because they are so bad the trade-offs required are worse than without a 3DLUT (more common than I would like).

I was simply arguing against the blanket statement:
3D LUT calibration is what makes the real visible impact on picture quality not swapping graphics cards.

It needs to be qualified with a "usually", i.e. "Usually a 3D LUT calibration is what makes the real visible impact on picture quality not swapping graphics cards." :p

Interesting discussion, but someone will mention the OT side of it, so might as well be me :P

We are trying to stick to what madVR features offer more benifit for an optimal experience with madVR. :o