TR Forums

Waco wrote:

My HTPC is running HDMI to my plasma...and I've been having a hell of a time calibrating it properly with the nVidia card versus the AMD card I had in it before.

I'm going to be pissed if the solution involves .inf hacking.

EDIT: Yup, totally does this. Now I get to re-calibrate it after making the hack work...

Yeah, I can't get used to actually seeing black and actual colors that aren't washed out now.

Kind of odd that the Windows control panel doesn't have settings found in the Linux control panel.

It's unfortunate that there is such a bug... I am not affected because I only use DVI. More people should post about it in official Nvidia's forums - the more people will do it, the (hopefully) faster Nvidia will properly fix it.

If you Google it, you'll see that the Nvidia forum has plenty of posts dating back to 2009. I think this needs attention like CrossFire did, in order for them to pay attention.

Like CrossFire, most of the people who have the issue probably don't even realize it.

JohnC wrote:

It's unfortunate that there is such a bug...

It isn't however a "bug". It is expected and proper behavior based on the EDID it is presented. Should those monitors present a proper computer EDID modeline instead of a TV modeline it wouldn't happen.

Deanjo wrote:

It isn't however a "bug". It is expected and proper behavior based on the EDID it is presented. Should those monitors present a proper computer EDID modeline instead of a TV modeline it wouldn't happen.

Meanwhile, back in the real world, these monitors do exist, Nvidia's customers are using them, and Nvidia has omitted a setting from their control panel that they've known for years that it needed.

As you say, it's not a bug. It isn't even a missing feature. They just haven't bothered to add a check box for it.

Melvar wrote:

Deanjo wrote:

It isn't however a "bug". It is expected and proper behavior based on the EDID it is presented. Should those monitors present a proper computer EDID modeline instead of a TV modeline it wouldn't happen.

Meanwhile, back in the real world, these monitors do exist, Nvidia's customers are using them, and Nvidia has omitted a setting from their control panel that they've known for years that it needed.

As you say, it's not a bug. It isn't even a missing feature. They just haven't bothered to add a check box for it.

You mean like this one?

Deanjo wrote:

Melvar wrote:

Deanjo wrote:

It isn't however a "bug". It is expected and proper behavior based on the EDID it is presented. Should those monitors present a proper computer EDID modeline instead of a TV modeline it wouldn't happen.

Meanwhile, back in the real world, these monitors do exist, Nvidia's customers are using them, and Nvidia has omitted a setting from their control panel that they've known for years that it needed.

As you say, it's not a bug. It isn't even a missing feature. They just haven't bothered to add a check box for it.

You mean like this one?

That only applies to the video renderer, not to the actual color space being output by the GPU. That's why it's under the "video" tab. It basically changes nothing if you have it set to limited RGB in the Adjust Desktop Color Settings(which you would only know by doing the test to see if black is gray, like I stated earlier).

This is what everyone who doesn't understand the problem posts, but it actually isn't a fix, like YCrBr is not a fix either.

If you have the issue, either set it to a PC resolution or create a custom one with a 59Hz refresh rate, and check to see if something that's black gets blacker.

Savyg wrote:

auxy wrote:

Huh. I've been kind of disappointed in the contrast of my monitors lately, but all of mine are connected via dual-link DVI cables or VGA, so I guess I'm not affected.

I'm actually disappointed because I really wanted a free and easy improvement in contrast! T_T

That'll need OLED methinks.

If that ever happens.

No, sigh... you... ugh.
Both plasma displays and LCDs with VA-type panels have vastly superior contrast to IPS and TN LCDs. Nobody makes a desktop-sized Plasma, and in any case I am not willing to deal with the other limitations thereof, so I am looking real hard at that EIZO NANAO Foris FG2421 monitor. So expensive, tho...

auxy wrote:

Savyg wrote:

auxy wrote:

Huh. I've been kind of disappointed in the contrast of my monitors lately, but all of mine are connected via dual-link DVI cables or VGA, so I guess I'm not affected.

I'm actually disappointed because I really wanted a free and easy improvement in contrast! T_T

That'll need OLED methinks.

If that ever happens.

No, sigh... you... ugh.
Both plasma displays and LCDs with VA-type panels have vastly superior contrast to IPS and TN LCDs. Nobody makes a desktop-sized Plasma, and in any case I am not willing to deal with the other limitations thereof, so I am looking real hard at that EIZO NANAO Foris FG2421 monitor. So expensive, tho...

Actually, IPS and OLED are both better than VA for contrast and color. Canon and Sony both have new 10 bit 4K IPS displays that are 30" and $30-40,000, and even Eizo's own color series of monitors is IPS.

However, all of this is irrelevant to this discussion, because sending the wrong color space to a monitor that's capable of deep color with a 16 bit LUT will make all of the colors totally distorted, and black will be gray, which is what's happening here.

I'm pressed for time this morning so I haven't had a chance to review this.

http://blog.metaclassofnil.com/?p=83

That should be an automated fix. Someone give a good once over.

Edit:

http://www.avsforum.com/t/1387832/htpc- ... t_22484920

Another option.

This was very instructive, nevertheless.

Everyone watch out for their video color settings in the Nvidia control panel. I found that while my regular stuff had the full dynamic range, the videos themselves did not. And that explains why I always thought that the black level of videos looked a little too bad on my (for now) TN main screen.

Having said that, there's something to be said about dark colors - the slightly crushed dynamic range helps here when your screen isn't very bright to begin with. So even though it's not correct, Nvidia was probably doing most users a favor by making this the default. I personally think that the full range is obviously the correct choice, but given 10 Regular Joes with their usually crappy low-end TN screens, 9 of them would prefer the limited range.

I have old TN panels with only vga inputs and I can tell the difference from the alternating colors used on the forums posts. I don't think it is the panel even if the color changes with the viewing angles severly you should still see the "white" and "blue" backgrounds as you scroll through the posts.

Looks like Intel is in the same boat as NVIDIA.

I've also read some claims that the HDMI specification calls for 16-235, the vast majority of TVs do 16-235, and that nearly all video sources are mastered in 16-235 (I recall the PS3 defaults to limited). You can create artifacts by forcing a content outside the native limited range.

This classes as irritating in certain scenarios, but I'm a bit unconvinced that NVIDIA is necessarily doing it wrong. If the claims are true, they're following the standard.

The Samsung HDTV I have connected to my PC over HDMI has a "HDMI black level" setting which when set to "Low" produces correct colors. Another option is to use Custom Resolution Utility and uncheck "Include extension block" which will present the display as a DVI monitor.

BlackDove wrote:

auxy wrote:

Both plasma displays and LCDs with VA-type panels have vastly superior contrast to IPS and TN LCDs. Nobody makes a desktop-sized Plasma, and in any case I am not willing to deal with the other limitations thereof, so I am looking real hard at that EIZO NANAO Foris FG2421 monitor. So expensive, tho...

Actually, IPS and OLED are both better than VA for contrast and color. Canon and Sony both have new 10 bit 4K IPS displays that are 30" and $30-40,000, and even Eizo's own color series of monitors is IPS.

However, all of this is irrelevant to this discussion, because sending the wrong color space to a monitor that's capable of deep color with a 16 bit LUT will make all of the colors totally distorted, and black will be gray, which is what's happening here.

You're wrong, but since it's offtopic, check your PM box!

morphine wrote:

This was very instructive, nevertheless.

Everyone watch out for their video color settings in the Nvidia control panel. I found that while my regular stuff had the full dynamic range, the videos themselves did not. And that explains why I always thought that the black level of videos looked a little too bad on my (for now) TN main screen.

Having said that, there's something to be said about dark colors - the slightly crushed dynamic range helps here when your screen isn't very bright to begin with. So even though it's not correct, Nvidia was probably doing most users a favor by making this the default. I personally think that the full range is obviously the correct choice, but given 10 Regular Joes with their usually crappy low-end TN screens, 9 of them would prefer the limited range.

Even if you have a bad TN screen, having the monitor set to the incorrect color space looks terrible. The numerical values for the colors have to match in order to get remotely decent color, even on a cheap monitor.

Using the wrong RGB, or using YCbCr instead of RGB on a monitor that's expecting full RGB is not doing ANYONE a favor. It doesn't matter if you have a complete piece of **** monitor. Black being gray just makes people adjust their monitors wrong, and think they have a worse monitor than they actually do.

Ryu Connor wrote:

I'm pressed for time this morning so I haven't had a chance to review this.

http://blog.metaclassofnil.com/?p=83

That should be an automated fix. Someone give a good once over.

Edit:

http://www.avsforum.com/t/1387832/htpc- ... t_22484920

Another option.

I think an actual fix from Nvidia is better than running things that random people on forums wrote on your computer. Modifying the .inf is bad enough, and these may lose compatibility as new drivers get released, and there are of course security concerns using these kind of "fixes".

You're not going to get a fix.

there are of course security concerns using these kind of "fixes".

One of them is open source, so no, not really.

Ryu Connor wrote:

You're not going to get a fix.

there are of course security concerns using these kind of "fixes".

One of them is open source, so no, not really.

Why wouldn't we get a fix? The option is literally already there, you just can't access it and their default setting is wrong.

All they have to do is change the setting from RGB/YCbCr to RGB full/RGB limited/YCbCr like AMD has in their driver control panel.

People noticing that CrossFire was broken prompted AMD to fix their stuff(or attempt to). If enough people with this issue notice and send feedback, or someone like Scott or Ryan Shrout mentioned it to Tom in their next discussion about their latest GPU, it might make a difference.

I don't care if it's open source or not, I still don't want to run something from some random website. That's a legitimate security concern, especially since I don't think most people are going to analyze the code.

And even if they do analyze the code, and it's safe, there's no guarantee it won't cause some other compatibility issue. There's no guarantee it will continue to work with every driver update either.

BlackDove wrote:

Why wouldn't we get a fix?

Like Intel they apparently have decided the HDMI standard and EDID is the way to carry this out. This is years old. The age and silence of this issue is clear in intent.

This is not broken performance. It's not even a broken standard on NVIDIAs part. This is nothing like AMDs latency issues and their resultant consequences.

If your display violates the standard the community has created fixes. Use them or don't.

As an aside my Sony XBR TV does not exhibit this issue. The EDID works correctly (the TV is also a full RGB set.)

Ryu Connor wrote:

BlackDove wrote:

Why wouldn't we get a fix?

Like Intel they apparently have decided the HDMI standard and EDID is the way to carry this out. This is years old. The age and silence of this issue is clear.

This is not broken performance. It's not even a broken standard.

As an aside my Sony XBR TV does not exhibit this issue. The EDID works correctly (the TV is also a full RGB set.)

How is setting a default color space output by a computer graphics card, that is the wrong color space(the computer monitor standard is RGB 0-255), NOT broken?

Let me get this straight. Your TV is listed in the PC resolutions, despite being connected via HDMI and its resolution being 1080p?

And are you sure your TV is set to full RGB? TV's are designed for Rec. 709 color, since that's the color space that things like HDTV and Blu-Ray, DVD etc. all use. That would look normal, if it were connected to a GPU outputting limited RGB.

What's your TV's model number?

Ryu Connor wrote:

This is nothing like AMDs latency issues and their resultant consequences.

I agree with that part. It has 0 impact on majority of users who do not use HDMI connection for their primary monitor. Unlike the issues with latency which affected almost every gamer (until AMD started to finally fix it), regardless of number of monitors or the type of connection used.

JohnC wrote:

Ryu Connor wrote:

This is nothing like AMDs latency issues and their resultant consequences.

I agree with that part. It has 0 impact on majority of users who do not use HDMI connection for their primary monitor. Unlike the issues with latency which affected almost every gamer (until AMD started to finally fix it), regardless of number of monitors or the type of connection used.

As I've said before, I'm a big Nvidia fanboy, but this issue affects a lot more people than the CrossFire thing did, and has been known for longer. I'd be willing to guess that since the majority of monitors in use are 1080p, and a lot of people use HDMI, and there aren't that many people actually using CrossFire, that this affects more people.

The CrossFire thing is a much harder to fix issue, and they can't fix it with the old GPU's, and they didn't even fix it that well with XDMA.

This fix apparently exists in the Linux control panel, and all Nvidia needs to do is add control over a setting that already exists, and they won't do it!

People don't seem to understand how big of a deal this is. Using the wrong color space means that you will never have accurate color on your monitor, no matter how you calibrate it. I plan on getting a new monitor, but for the last two years, I've thought I just had a bad monitor. I was wrong, and the issue is a really stupid one with an easy fix.

BlackDove wrote:

[b]Using the wrong color space means that you will never have accurate color on your monitor,

Except for the mass plethora of media all mastered for limited range?

I've thought I just had a bad monitor.

We could argue it is a bad monitor, after a fashion. Bad EDID data.



The camera originally took the image as a JPG. So the image has been degraded a bit, but as you can see there is no banding.

Ryu Connor wrote:

BlackDove wrote:

Using the wrong color space means that you will never have accurate color on your monitor,

Except for the mass plethora of media all mastered for limited range?

I've thought I just had a bad monitor.

We could argue it is a bad monitor, after a fashion. Bad EDID data.



The camera originally took the image as a JPG. So the image has been degraded a bit, but as you can see there is no banding.

Ok, you don't understand what's going on here.

The HDTV standard is Rec. 709 color, where 16=black and 235=white. The PC monitor RGB standard is 0=black 255=white. http://en.wikipedia.org/wiki/Rec._709

When you give the 16=black, 235=white color space to a monitor set to 0=black, 255=white monitor, it distorts the output and black becomes gray. That's what's happening here.

If you were to give a full RGB signal to a TV that couldn't accept it, you would have the opposite, where grays would become black, and you'd lose detail.

You can calibrate a monitor improperly so that your gray black will be really dark, and black when you look at it, [b]but everything else is distorted.

If your TV is not set to be a PC monitor(which you'd have to do manually usually), your TV is set to 16-235, not 0-255, since that's the HDTV color standard. That's why you're not seeing banding, or you have your TV calibrated improperly. If your TV's resolution is listed under the TV section, and you haven't done the fix with the .inf, then your GPU is outputting the 16-235 color, which won't show banding on a TV that's set to 16-235 color.

What is happening here, is that the GPU is sending the wrong signal when it detects an HDMI and a TV resolution, even if the monitor is set to PC monitor color space. You can't argue that any monitor that's using HDMI is a "bad" monitor, simply because someone plugged it in using that cable. It's not bad EDID either. There are so many monitors being produced that they don't need to add the EDID info of every one. They need to make an option so people can set their color space manually, like the Linux drivers and AMD drivers allow.

If you read the posts going back for years, you'll see that it doesn't matter what monitor you plug in via HDMI, it will do this if it has a "TV" resolution.

then your GPU is outputting the 16-235 color, which won't show banding on a TV that's set to 16-235 color.

The output is roughly identical with and without the fix.

The display uses xvYCC (it's a 10bit panel) and is set to wide instead of limited.

Ryu Connor wrote:

then your GPU is outputting the 16-235 color, which won't show banding on a TV that's set to 16-235 color.

The output is identical with or without the fix.

The display uses xvYCC (it's a 10bit panel) and is set to wide instead of limited.

xvYCC is a compatible with RGB 16-235 NOT 0-255! http://www.sony.net/SonyInfo/technology ... tml#block1

xvYCC is just an extension of Rec. 709. It's a TV color standard, and you won't get banding with RGB 16-235!

If your TV is using xvYCC right now, and you have it plugged in via HDMI, and you are using RGB 16-235, then you have the same issue I'm talking about here.

The only difference is that you're using a TV instead of a computer monitor, and since the default is right for a TV but wrong for a monitor, it looks fine for your TV!

This issue applies to people using computer monitors, not TV's!

If you did the .inf fix and you can't get limited RGB, you need to do a clean install, and get back the 16-235, because your detail will be destroyed if you use RGB 0-255 on a TV that's supposed to be using xvYCC or normal Rec. 709 color.

The xvYCC color space permits YCC values that, while within the encoding range of YCC, have chroma values outside the range 16–240, or that correspond to negative RGB values, and hence would not have previously been valid.

My confusion centers around the fact that the PS3 makes a distinct difference when switched between limited and full on said TV. Meanwhile limited or full from the graphics adapter have only a very tiny difference (the difference is subtle, which is why I said roughly equivalent) when a PC is connected. My presumption is the latter is negotiating xvYCC from the start where as the PS3 really is flipping down to limited when told so. Unfortunately neither the TV nor the NVIDIA control panel provide the sort of diagnostics I'd need to tell what's going on (NVCP doesn't show me the TV section item you keep mentioning). I'm not entirely sure you're right, but the conversation is done in my mind.

I appreciate you walking through this. It helps to have a better grasp of some of the standards, but my opinion is largely unmoved. I still don't see the point of fixing it, it is the HDMI standard after all, but I definitely see the wisdom in not buying an HDMI "monitor."

The details imply it would be easy for the monitor vendors to work around this issue with EDID (a common hack I've read today involves refresh rate), but then again HDMI has been the abused child for years in the PC industry. Many large resolution monitors do not support HDMI and even those that do - like my 3008wfp for example - can only do 1920x1200 via HDMI (pointless). Technologies like EyeFinity pushed DisplayPort (and required at least one in a triple monitor setup) for years. Meanwhile classic 24" 1920x1200 16:10 monitors using HDMI would have no issues either, albeit those displays typically come with other inputs. I find the tweak tools sufficient for the niche scenarios I need them (which at the moment would appear to tally up to zero for me). HDMI has always been a TV technology and apparently that statement is rather literal.

None of this even centers on if it is a negative issue. How many laymen calibrate their screens or are negatively impacted by limited? There is a whole slew of Twitch TV users outputting in 16-235, I suspect YouTube is little different. This displeases you, and I can appreciate that, but becareful of putting the rest of the world in your shoes. There's a good chance you stand in a very tiny minority. This feels like a conversation on the cusp of a IPS vs 120Hz TN style battle.

Getting excited here isn't going to do much good. This isn't the venue to petition NVIDIA or for that matter Intel. Both NVIDIA and Intel are adhering to the HDMI standard. This is a community forum and this thread has community fixes. Our obligation has been met.

This is going to be my last post. This thread has a fix in it. I'm going to see if the thread dies without my participation. At this point I actually feel kind of guilty of keeping this alive.

Ryu Connor wrote:

The xvYCC color space permits YCC values that, while within the encoding range of YCC, have chroma values outside the range 16–240, or that correspond to negative RGB values, and hence would not have previously been valid.

My confusion centers around the fact that the PS3 makes a distinct difference when switched between limited and full on said TV. Meanwhile limited or full from the graphics adapter have only a very tiny difference (the difference is subtle, which is why I said roughly equivalent) when a PC is connected. My presumption is the latter is negotiating xvYCC from the start where as the PS3 really is flipping down to limited when told so. Unfortunately neither the TV nor the NVIDIA control panel provide the sort of diagnostics I'd need to tell what's going on (NVCP doesn't show me the TV section item you keep mentioning). I'm not entirely sure you're right, but the conversation is done in my mind.

I appreciate you walking through this. It helps to have a better grasp of some of the standards, but my opinion is largely unmoved. I still don't see the point of fixing it, it is the HDMI standard after all, but I definitely see the wisdom in not buying an HDMI "monitor."

The details imply it would be easy for the monitor vendors to work around this issue with EDID (a common hack I've read today involves refresh rate), but then again HDMI has been the abused child for years in the PC industry. Many large resolution monitors do not support HDMI and even those that do - like my 3008wfp for example - can only do 1920x1200 via HDMI (pointless). Technologies like EyeFinity pushed DisplayPort (and required at least one in a triple monitor setup) for years. Meanwhile classic 24" 1920x1200 16:10 monitors using HDMI would have no issues either, albeit those displays typically come with other inputs. I find the tweak tools sufficient for the niche scenarios I need them (which at the moment would appear to tally up to zero for me). HDMI has always been a TV technology and apparently that statement is rather literal.

Getting excited here isn't going to do much good. This isn't the venue to petition NVIDIA or for that matter Intel. Both NVIDIA and Intel are adhering to the HDMI standard.

This is going to be my last post. This thread has a fix in it. I'm going to see if the thread dies without my participation. At this point I actually feel kind of guilty of keeping this alive.

The point of fixing it is so that people can get the correct colors displayed on their monitors!

The Linux control panel does have this feature built in, as does AMD's driver control panel.

In the Nvidia control panel, there is a "change resolution" section. The resolutions are listed under "Ultra HD, HD, SD" or "PC". If your resolution is listed under "Ultra HD, HD, SD" and your monitor is RGB 0-255, your color is distorted.

Since you have a TV hooked up to it, you're fine(as long as you didn't do the fix).

This is the important part of the article I linked to:

"This standard encompasses xvYCC709, which is upwardly compatible with the B7.709 standard, and xvYCC601, which is upwardly compatible with BT.601. The four characteristics of these standards are shown in Figure 2.

1.The chromaticity coordinates of RGB primary colors and the reference white (D65) are the same as for the existing BT.709 standard."