madVR - high quality video renderer (GPU assisted)

[THX-UltraII]

No. With RGB Full, the data rendered by madVR is sent to the display 1:1 without any changes. Only this way madVR can guarantee highest quality. With RGB Limited, the GPU driver stretches the data behind madVR's back, which will damage image quality.


ok, so RGB Full MUST be fixed asap by ATI! Do you recommend to switch to another renderer for the time beeing? We all know ATI and this could take at least 4-10 months before this is fixed (IF it is fixable anyway)

[madshi]

Do you recommend to switch to another renderer for the time beeing?
Just test for yourself, which renderer gives you the best image quality with the current drivers.

IIRC the problem only occurs with 10.4 and newer. So maybe downdating to 10.3 might help. Not 100% sure, though.

[THX-UltraII]

Just test for yourself, which renderer gives you the best image quality with the current drivers.

IIRC the problem only occurs with 10.4 and newer. So maybe downdating to 10.3 might help. Not 100% sure, though.

I m very sure that the problem is also with 10.2, 10.1 and 9.12. What s your source of this information?

[THX-UltraII]

Come on guys, there must be lots of 5xxx/madVR users out there. Please post in the ATI topic as much as possible so Spyre can take action.

[noee]

Just test for yourself, which renderer gives you the best image quality with the current drivers.

IIRC the problem only occurs with 10.4 and newer. So maybe downdating to 10.3 might help. Not 100% sure, though.

Fwiw guys, on my HD2600XT with 10.3 and madVR, I do see the difference between Full and Limited.

[LloydA]

Can everybody with an ATI 5xxx card please test whether switching between RGB Full and RGB Limited changes the image, when using madVR? Please report your GPU, OS and your test result.

GPU: ATI 5450
OS: Windows 7 x64

No difference between full and limited. Tested with 10.3, 10.6, 10.7. No difference with any Dynamic Range settings, either.

I verified with the "TV Levels" mp4s posted in the forum a few days ago. At any setting I get blacks down to 1-2, and whites up to 252-253.

In that AMD forum someone claimed that he gets "gray blacks" due to the bug. I can't confirm it, on my TV blacks are black at any setting, and since it's a plasma, non-pure blacks would show up quickly as dithered gray/green.

[mark0077]

When you say there is no difference in blacks. My question is what is your tv set to accept full or limited. If limited then blacks and whites won't change between windows settings because anything between 0 and 16 will show up as black right?

[LloydA]

My tv has no setting for TV/PC levels, it's all automagical. I'm running over HDMI, by the way (HTPC -> Denon A/V (untouched video signal) -> TV).

[mark0077]

Wow pretty nice feature. Ie auto detection. If it has an auto detect feature then of course blacks and whites will look similar no matter what you set your ati cp to. I don't know how auto works though. Don't know how it can decide if its full or limited input. All manual on my sammy series 9.

[leeperry]

My tv has no setting for TV/PC levels, it's all automagical. I'm running over HDMI, by the way (HTPC -> Denon A/V (untouched video signal) -> TV).
considering many encoders seem to use BTB/WTW for padding(and some movies would actually appear to use them), I don't see the auto-detection working very accurately. I personally prefer to ignore BTB/WTW, but Seb.26 added an automatic filter profile in ffdshow that will constantly look for valuable data outside 16-235...but if can easily pump the black levels.

[DigitalLF]

MadShi

my description is up to date... there is no change with MadVR and "EVR Sync" when i change "Dynamic Range" and force change it to 0-255 during playback... (or do i need to stop the playback?)

my projector is broken right now so i have only tested it on a computer screen but i can't see the normal change...

[LloydA]

Wow pretty nice feature. Ie auto detection. If it has an auto detect feature then of course blacks and whites will look similar no matter what you set your ati cp to. I don't know how auto works though. Don't know how it can decide if its full or limited input. All manual on my sammy series 9.

Seems so. I never thought about it. The downside is that I can never get smooth gradients with default settings on that tv, everything has bands in it (that is, when displaying synthetic 0->255 gradients) unless I dial down contrast to 0 (from the default 50). But then everything looks like a bleached mess.

I guess it's just inherent to the dithering technology in plasmas in general. Great PQ at the expense of "scientific accuracy".

[leeperry]

I guess it's just inherent to the dithering technology in plasmas in general. Great PQ at the expense of "scientific accuracy".
sounds like you need to calibrate your display using a commercial sensor. these test patterns could help too: http://www.google.com/search?hl=en&q=Milori+Colorfacts+Test+Patterns+1.0

[yesgrey]

True, but there is a "right" (and a wrong) Pixel Format setting. And it seems that this feature is broken in current ATI drivers.
From a previous discussion with another user it seems the problem only affects HDMI output. It seems that when using DVI output the Limited/Full setting is working. So, it might seem that ATI is considering that HDMI output should always be the limited output. Yes, I know this is wrong, but they might think differently...
Have you tried using a DVI connection instead of the HDMI one?

[janos666]

I will check it. I am using DP right now but I am planning to use HDMI because I want to use the xvYCC mode (and check if my VGA would output 12 bit because - according to the HDMI standard - it is mandatory for DeepColor and both my VGA and my display support DeepColor on paper but I have "only" 10 bit/color connection right now - through 4-line DP.).
But I can't hope any improvements if there is bug like this...

[Mark_A_W]

sounds like you need to calibrate your display using a commercial sensor. these test patterns could help too: http://www.google.com/search?hl=en&q=Milori+Colorfacts+Test+Patterns+1.0


My second system is a Plasma (main is CRT based running VGA).

And on the plasma I have been unable to get a band-free greyscale, no matter what HDMI settings I try on the PC (Nvidia).

HDMI is evil.

On my CRT I set the levels to 0-255 but unmolested (black at 16), then set the CRT brightness lower to compensate. Then I get a PERFECT greyscale with no banding whatsoever. It is SIMPLE.

I have a sensor and HCFR, and I don't think it's going to help with banding AT ALL.


From my testing, any expansion of levels (16 to 0 and 235 to 255) causes banding. The best path on a digital display would be to set it to crop WTW and BTB completely, and just use 16-235 unmolested.

But I can't seem to find the settings which do this on my Plasma system (and I run XBMC on that system, for the kids, and it seems to always expand).

[cyberlolo]

There is nothing to discuss here. madVR behaves like the desktop, like a game. The Pixel Format setting should show an effect, but does not. This is a clear bug in the ATI drivers. I don't know what the AVIVO Dynamic Range setting does, but I guess it does not impact games or the desktop, so it also will not affect madVR. There's no use in testing with EVR/VMR. If you want to test with something other than madVR, try desktop and games.

I've got an ATI HD4850, and I can confirm that madVR is affected by Catalyst desktop controls (you can change brightness, contrast, saturation, gamma...), but not by the AVIVO ones.

Theorically, the AVIVO controls are there to configure the video playback, but madVR overrides them (that's why I requested the madVR's brightness and contrast controls some time ago). If you use any other renderer, then the AVIVO controls will work.

In the other hand, with the 4xxx series, RGB Full and RGB Limited give me the same result (RGB Full) using HDMI.

I don't recommend to use the ffdshow's levels filter (any of them), as they work in 8bit mode, and introduce a lot of banding. The better options for me are (having RGB Full):

1) Use yCMS to generate a 3dlut to make the expansion 16->0 and 235->240. This is the way I work, and I have a perfect grayscale with no banding using HDMI. (if anybody needs help to achieve this, please let me know)

2) Set madVR to PC levels and let madVR do it's great work. This way, you'll be losing some info (235-240) that it's recommended to keep (although you can live without that) because of the possible excursions above reference white in some videos, and also to "fix" defective gamma at high levels (a lot of displays suffer this problem).

[fps]

Can you check if it makes a difference when you switch to Limited RGB as ati Pixel Format?

Exactly.

Can everybody with an ATI 5xxx card please test whether switching between RGB Full and RGB Limited changes the image, when using madVR? Please report your GPU, OS and your test result.
I tried various settings now and switching between Limited RGB and Full RGB as Pixel format changes the image considerably on my computer.
The Dynamic Range option has no effect as far as I can tell, though.
To get correct levels I need to enable TV levels in madVR for Full RGB Pixel format, but PC levels in madVR for Limited RGB Pixel format.
Otherwise I get some kind of "greyish black" in the credits for example.
This test picture gives me funny results.
http://img155.imageshack.us/img155/5223/btbtest.th.jpg (http://img155.imageshack.us/i/btbtest.jpg/)
With Limited RGB I can see most of the numbers down to 3. Switching to Full RGB results in unreadable numbers till half the picture.
The funny thing is, if I open the image with the windows internal image viewer, for a split second I can see all the numbers and then it switches to the described image above.

Win 7/32
Catalyst 10.6
ATI 5670
A/V Receiver (hdmi passthrough): Onky TX-SR608
Projector: Sony VPL-HW10

[janos666]

I've got an ATI HD4850, and I can confirm that madVR is affected by Catalyst desktop controls (you can change brightness, contrast, saturation, gamma...), but not by the AVIVO ones.

There are two ways to do it from CCC:
- The Desktop_Properties/Color will mess with the 16 bit LUT (it will clear your calibration LUT when you use it).
- The Desktop_and_Display/Configure.../Color will do some real-time (high precision) calculations.
Normally, I don't want to use any of them but you should play with the second one. (A little brightness boost and contrast reducing (lika +2 ; -2) or the inverse of it may help a bit with a little near-black or near-white banding issues.)

Theoretically, the AVIVO controls are there to configure the video playback, but madVR overrides them
I think it simply ignores them because they are supposed to be applied for WMR9 and EVR. MadVR is a standalone renderer, so it isn't supposed to be affected by EVR settings.
Moreover, madVR was designed to be unaffected. This is why we like it...

1) Set madVR to video levels (unmolested) and use yCMS to generate a 3dlut to make the expansion 16->0 and 235->240. This is the way I work, and I have a perfect grayscale with no banding using HDMI. (if anybody needs help to achieve this, please let me know)

I think when you use 3DLUTs then madVR's TV and PC level setup does absolutely nothing expect it choose the corresponding *.3dlut file. So, TV and PC level settings will produce exactly the same result with the same (duplicated) 3dlut files.
Am I wrong?

Edit:
I made two 3DLUTs: one with Output_Range 0 255 and one with Output_Range 0 240
There was no difference between TV and PC modes when I used the same 3dlut file for both settings (I duplicated the one with 0;240 settings).
The 0;240 setting gives me a little darker image but I can't see more details.
I can't see any difference between madRV's TV setting and a 3dlut processing with 0;255 levels.
May be I should check it with test patterns instead of Blu-Ray movies. :D

[cyberlolo]

There are two ways to do it from CCC:
- The Desktop_Properties/Color will mess with the 16 bit LUT (it will clear your calibration LUT when you use it).
- The Desktop_and_Display/Configure.../Color will do some real-time (high precision) calculations.
Normally, I don't want to use any of them but you should play with the second one. (A little brightness boost and contrast reducing (lika +2 ; -2) or the inverse of it may help a bit with a little near-black or near-white banding issues.)

I don't understand this. What's the difference between "The Desktop_Properties/Color" and "The Desktop_and_Display/Configure.../Color" in CCC?

I think it simply ignores them because they are supposed to be applied for WMR9 and EVR. MadVR is a standalone renderer, so it isn't supposed to be affected by EVR settings.
Moreover, madVR was designed to be unaffected. This is why we like it...

Agree.

I think when you use 3DLUTs then madVR's TV and PC level setup does absolutely nothing expect it choose the corresponding *.3dlut file. So, TV and PC level settings will produce exactly the same result with the same (duplicated) 3dlut files.
Am I wrong?

Yes, that's right. I'll edit my post.

Edit:
I made two 3DLUTs: one with Output_Range 0 255 and one with Output_Range 0 240
There was no difference between TV and PC modes when I used the same 3dlut file for both settings (I duplicated the one with 0;240 settings).
The 0;240 setting gives me a little darker image but I can't see more details.
I can't see any difference between madRV's TV setting and a 3dlut processing with 0;255 levels.
May be I should check it with test patterns instead of Blu-Ray movies. :D

What I use is a 3dlut file with Output_Range 16 240 and PF=Full RGB, and it gives me perfect results.

[DigitalLF]

yesgray & madshi: i use a dvi-hdmi (high-quality) but with a dvi adapter at the hdmi side of that cable because it is the only 8m cable i got.. so its something like a dvi-dvi :P but i see no difference with 0-255 or 16-235 on my computer screen..

[THX-UltraII]

Pffff, things are getting too complicated and technical for me now since the last posts.

So is the conclusion still that it s not recommended to use madVR as renderer when you have a 5xxx series card because Madshi said that madVR was made to work in combination with RGB Full and RGB Full is broken for this moment on the
5xxx serie?

[fps]

I don't know why limited and RGB full displays the same on your machine, but as posted above I get a considerably different output when switching between these two settings.
It might be depending on the display wheter there is a bug in the ATI driver or not.

I don't think it's wise just to not use madVR because Full RGB "might be broken".
If it is indeed broken every other renderer will have the same problem in RGB, too.
Just trust your eyes what looks better I'd say :cool: (I bet it's still madVR ;)).

[mark0077]

Note: NVidia drivers don't even have a setting to choose the output range of the entire desktop. For my Samsung TV, it forces output to limited range until I create a custom resolution, which then magically sends output in full range mode.

The 0-255 / 16-235 color option they have only applies to video output (and only non rgb I think), and even that isn't smart enough to output correctly depending on the output range of the desktop I speak about above... I have asked on the nvidia forums to add this option, similar to what ATI have.

[Andy o]

The ATI 5000 cards seem to work with Full RGB and Limited RGB pixel formats correctly only with some displays (or maybe receivers). I have a Pioneer KRP-500 through a Pioneer VSX-01 receiver, and Full and Limited RGB Pixel formats behaved as they should. On the other hand, Nvidia's GTX460 is giving me perfect levels everywhere (including MC which was giving me trouble with ATI) on YCC mode. On RGB mode, it outputs Full RGB on my setup (no need for custom resolutions) and also I don't have any levels problems.

I use YCC cause it doesn't apply unnecessary expansion though, thus I get perfect gradients in overlay mixer and TMT3 and PowerDVD.

[cyberlolo]

The ATI 5000 cards seem to work with Full RGB and Limited RGB pixel formats correctly only with some displays (or maybe receivers). I have a Pioneer KRP-500 through a Pioneer VSX-01 receiver, and Full and Limited RGB Pixel formats behaved as they should. On the other hand, Nvidia's GTX460 is giving me perfect levels everywhere (including MC which was giving me trouble with ATI) on YCC mode. On RGB mode, it outputs Full RGB on my setup (no need for custom resolutions) and also I don't have any levels problems.

I use YCC cause it doesn't apply unnecessary expansion though, thus I get perfect gradients in overlay mixer and TMT3 and PowerDVD.

If you use madVR, you need to set RGB to get max quality. With YCC, you're making double conversion and losing PQ.

[THX-UltraII]

Pffff, things are getting too complicated and technical for me now since the last posts.

So is the conclusion still that it s not recommended to use madVR as renderer when you have a 5xxx series card because Madshi said that madVR was made to work in combination with RGB Full and RGB Full is broken for this moment on the
5xxx serie?


Don t want to be a pain in the ass here but I still haven t got a clear answer to this.....

[madshi]

Don t want to be a pain in the ass here but I still haven t got a clear answer to this.....
You have been told by 2 different people that you should try for yourself which renderer gives you the best image quality with the current broken drivers.

[makakam]

To me it's still madvr that is the best in PQ. And btw I've recently played with scaling settings and set them to lanczos8, lanczos8, spline64 and smoothnes is so much better than on default settings. Are these settings less demanding or what?

[fps]

Actually it's the opposite :cool:. lanczos8 ist the most demanding resizer followed by the spline one's.
In the OSD you should see the rendering time going up a bit.

[pankov]

makakam, fps,
there is a very simple explanation why even though lanczos8 is more demanding it gives more smooth results ... or at least there is one for me - an ATI user.
It's possible that the default settings are at the threshold at which ATI PowerPlay kicks into "higher gear" (increases the GPU/Memory speeds) for your video card/PC. This switching usually gives spikes in rendering times and you should make your best to avoid them. Personally I've created a CCC profile which forces high frequencies (using ATI Overdrive) and this way I'm sure there is no switching of frequencies without my control.

[makakam]

pankov, I'm not an expert but this explanation sounds really probable to me :)

[Andy o]

If you use madVR, you need to set RGB to get max quality. With YCC, you're making double conversion and losing PQ.

I was just commenting on the 5000 cards' ability to do RGB Full Pixel format. I'm using overlay actually.. I don't think that converts to RGB if you're using YCC.

[cyberlolo]

I was just commenting on the 5000 cards' ability to do RGB Full Pixel format. I'm using overlay actually.. I don't think that converts to RGB if you're using YCC.

Ok then. But, just curiosity, why do you use overlay instead of madVR?

[Andy o]

If I set my system to RGB Full (the only option with Nvidia, and the only option that would work well with ATI for me) to use madVR, then I'll have the banding with TMT and PowerDVD. Using Overlay and YCC I don't have to expand anywhere so I don't get banding. I get around the resizing limitations and other processing by using ffdshow for scaling and such.

[cyberlolo]

If I set my system to RGB Full (the only option with Nvidia, and the only option that would work well with ATI for me) to use madVR, then I'll have the banding with TMT and PowerDVD. Using Overlay and YCC I don't have to expand anywhere so I don't get banding. I get around the resizing limitations and other processing by using ffdshow for scaling and such.

Ok, thanks for explaining, Andy.

[Razoola]

Sorry I havent been giving much info recently, the thing is everything has been working perfectly for me over the last few releases so don't hold back from starting full screen exclusive mode on my behalf :)

[pankov]

gusy,
I've found a temporary solution for my Pixel Format issues with my ATI 5xxx card (5750 to be precise).
Since my card has two DVI connectors I decided to try and use the second one with a DVI-HDMI converter/dongle and guess what - it works perfectly in Full RGB Mode ... well to be hones I lost the Pixel Format tab in CCC but as far as I can see the current output format according to my projector is RGB-Video (0-255) - I think so because I can now use HDMI Video range - Expanded and still get perfect deep black levels - the same ones I had when I used my ATI 4670 and Full RGB.
If you are lucky enough to have two DVI connectors you can use this until AMD/ATI fixes the drivers.

[makakam]

gusy,
I've found a temporary solution for my Pixel Format issues with my ATI 5xxx card (5750 to be precise).
Since my card has two DVI connectors I decided to try and use the second one with a DVI-HDMI converter/dongle and guess what - it works perfectly in Full RGB Mode ... well to be hones I lost the Pixel Format tab in CCC but as far as I can see the current output format according to my projector is RGB-Video (0-255) - I think so because I can now use HDMI Video range - Expanded and still get perfect deep black levels - the same ones I had when I used my ATI 4670 and Full RGB.
If you are lucky enough to have two DVI connectors you can use this until AMD/ATI fixes the drivers.
pankov, I also have 2 dvi connectors, the question is how do I know if the pixel format is full RGB after connecting to the second one? Right now I am already connected to one of the dvis with a dvi<->hdmi cable and I'm missing pixel format tab. I haven't tried the other one though.

[pankov]

makakam,
I'm not sure if ATI cards are at all capable of outputting anything else but RGB over DVI, but in my case I'm 99% sure I'm getting RGB Full because my projector (EPSON TW2000) says that the input is "RGB-Video" and I get the correct levels when I use HDMI Video Range "Extended". Just give it a try and see if it's different from the dedicated HDMI output.

[janos666]

Can you implement a little option (which is disabled by default, and I don't need a check-box on the GUI, an ini variable would be good enough) which clears the VGA LUT but reloads it after you close the video player?

I think it is possible because some PC games do this. (I wish they would not do it... but the "smarter" games won't mess with the VGA LUT, so this is not a D3D Fullscreen limitatin or the "smarter" games implemented some tricks to bypass this limitation which I doubt.)

As we talked about it in the yCMS thread, it would be better to let the madVR+yCMS combo to handle the full display calibration for movies. But I want to keep my VGA LUT loaded anywhere else (web browsing, PC games, ect.).
There are some little and smart utilities for this but I don't want to keep in mind that I always have to clear and reload it manually every time I watch some movies. It would lead to mistakes or "anger" (so, I will simply ignore it after some days or week because I am bored about it.)

[janos666]

Sorry if this is not a new problem or not a problem at all but I want to clarify this question.
I used MPC-HC in the "Touch windows from Insise" full screen mode and I made some screenshots of the smallramp.ytp with madVR 0.23 (with disabled dithering -> but yes, I usually keep it enabled. This helps now to clarify the problem...)
The 3DLUT input commands were simple:
Input_Format HD YCbCr 8
Output_Format HD RGB_Video 16
and
Input_Format HD YCbCr 8
Output_Format HD RGB_PC 16
There were no any additional processing. Only the two mad* filters were loaded. And I made a reference image from the EIZO monitor test program (a full 8 bit gray scale with PC levels).

TV_off (http://img12.tar.hu/janos666/img/82128429.png) ; TV_on (http://img12.tar.hu/janos666/img/82128430.png) ; PC_off (http://img12.tar.hu/janos666/img/82128427.png) ; PC_on (http://img12.tar.hu/janos666/img/82128428.png) ; EIZO_Reference_PC_Pattern (http://img12.tar.hu/janos666/img/82129341.png) (ON=3DLUT enabled ; OFF=3DLUT disabled)

The first questionable thing is that there is no two equal images between the first four scrrenshots. The 3DLUT processed images are very close to their unprocessed brothers but they are not equal. Which one has more rounding errors?
(May be you should consider to make 32 bit 3DLUTs for the 32 bit processing pipeline?) Or is there some little bug which still remained in yCMS? But yes, this is far from an actual "problem". I just noticed it. (You shouldn't take it too seriously.)

Because none of the four madRV images matches with the reference image from the EIZO monitor test, I assume that the smallramp.ytp behaves like a Rec709 source. (It uses only the limited range of the available 8 bit integer values.)

The real question in my mind the very noticeable difference between the TV and PC modes. This shows a very strange behavior for a laymen at the first time.
It looks like a heavy banding issue but this is not a simple banding (or not an issue at all, but I can't judge alone).
You can see that the number of the available shades are equal between the TV and PC modes (otherwise, it is only a guess, I didn't really counted them manually :D). But there are big black and white bars on the edges in PC mode.
What are those bars? Are they the missing 0...16 and 240...255 values? Is it normal that I can see those bars? Is it the correct TV->PC level mapping?
I think it is normal because the "working" bars are thinner in the PC mode, so the big bars looks like they are only additional things there and every steps exist.
But I cannot be sure...

[leeperry]

Can you implement a little option [..] which clears the VGA LUT but reloads it after you close the video player?

I think it is possible because some PC games do this.
Honestly, I think it's the ATi video drivers that are doing it...nvidia don't mess w/ the CLUT in games IME. When D3D exclusive mode will be effective in mVR, the ATi drivers will flush your CLUT.

But indeed it's quite redundant to use a CLUT on top of a 16bit 3DLUT. Madshi spoke about PS gamut mapping a while ago(useful for ppl w/ 10/12bit corrected displays who don't need calibration from the PC), but my last posts about it didn't raise much attention :/

[Virtual_ManPL]

Using madVR 0.23 + MPC-HC 1.3.2179.0 causing frame freezing in fullscreen mode, it unfroze when I move mouse on seeking menu. Works fine in windows mode or using older madVR 0.9.

It's only me or some1 have this problem too ?



P.S. Using latest nVidia drivers 259.09

[namaiki]

What about when you are using 258.96, 257.21, or 257.15?

[yesgrey]

I made a reference image from the EIZO monitor test program (a full 8 bit gray scale with PC levels).
It's not a full 8 bit gray scale. It only goes from 0-239. madVR test pattern truly is a full 8 bit gray scale, it goes from 0-255, hence the difference you notice between the madVR and EIZO grayscales.

Because none of the four madRV images matches with the reference image from the EIZO monitor test, I assume that the smallramp.ytp behaves like a Rec709 source. (It uses only the limited range of the available 8 bit integer values.)

Wrong. See previous answer.

The first questionable thing is that there is no two equal images between the first four scrrenshots. The 3DLUT processed images are very close to their unprocessed brothers but they are not equal. Which one has more rounding errors?
The 3DLUT processed images have slightly higher rounding errors, because it's 16 bit integer output of the 3DLUTs vs 32 bit FP of shader math in madVR. So, if people only want to perform YCbCr->RGB conversion with madVR they should not use 3DLUTs, but the idea of the 3DLUTs is not for performing only YCbCr->RGB conversion. ;)

But there are big black and white bars on the edges in PC mode.
What are those bars? Are they the missing 0...16 and 240...255 values? Is it normal that I can see those bars? Is it the correct TV->PC level mapping?

Yes, it is correct, and yes, it is normal that you can see those bars. As I have told previously, the madVR test pattern is a full 8 bit YCbCr greyscale, so it has all values from 0 to 255. When you use PC levels YCbCr [16,16,16] is expanded to RGB [0,0,0] and YCbCr [235,235,235] is expanded to RGB [255,255,255]. That's why you are seeing those black and white bands, because all YCbCr values < 16 are expanded to RGB values < 0 (and then clipped to 0), and all YCbCr values > 235 are expanded to RGB values > 255 (and then clipped to 255).

[makakam]

don't hold back from starting full screen exclusive mode on my behalf :)

Totally agree :)

[Octo-puss]

I have one stupid question. Can madVR potentially replace Haali renderer? I do not technically understand these things, but given how poorly updated Haali is, I am curious.

[noee]

My advice is, if your GPU can handle madVR (and it appears anything newer can), don't look back.

[janos666]

Honestly, I think it's the ATi video drivers that are doing it...nvidia don't mess w/ the CLUT in games IME. When D3D exclusive mode will be effective in mVR, the ATi drivers will flush your CLUT.

I don't think so. The calibration LUT usually remains untouched by the D3D fullsreen programs.
I made some tests when I noticed that StarCraft 2 clears the LUT. I manually edited the LUT entries to make it obviously noticeable.
Mass Effect 2 and Assassin's Creed 2 worked well (Dx9).
StarCraft 2 (Dx9?) and Riddick - AotDA (OpenGL ?.?) cleared the LUT.
Well, the Riddick game has a gamma slider between the graphic options. So, may be it uses the LUT to perform the user specified correction. And these are old games (SC2 is old from technology sides as well).

But indeed it's quite redundant to use a CLUT on top of a 16bit 3DLUT.

I used to take the measurements with loaded VGA LUT to feed yCMS. So, it is not so squarely redundant.

The current yCMS version can't correct the white balance
across the full gray scale. So, it is better to keep the LUT loaded for now.

But may be it would be better to clear the VGA LUT as soon as yesgrey will implement the ability to insert the gray scale measurement values in XYZ format. (And it will take care about the white balance as well.)

Madshi spoke about PS gamut mapping a while ago(useful for ppl w/ 10/12bit corrected displays who don't need calibration from the PC), but my last posts about it didn't raise much attention :/

Well. I think it is not so simple. If you have a display with nearly perfect sRGB gamut and accessible internal LUT (or wide gamut + accessible internal 3DLUTs) you don't need the yCMS software.
And may be a very good factory calibration can be good enough with some finetuning through the VGA LUT with 10/12 bit connection.
But these are very rare situations.

I am at the borderline. The factory calibration is far from perfect and the 10+ bit connection is just enough to avoid the banding after the white balance and gamma calibration. The gamut emulation is close but not perfect and I can't finetune it. So I want to compare the results between the sRGB mode + VGA LUT and Custom mode (unfortunately, this is not a "fully native" mode either, but this is the closest) + yCMS


@Octo-puss
Not just potentially, madVR is already superior. :rolleyes: