AVS Forum banner

9561 - 9580 of 10380 Posts

·
Registered
JVC RS4500 | ST130 G4 135" | MRX 720 | MC303 MC152 | 6.1.4: B&W 802D3, 805D3, 702S2 | 4x15 IB Subs
Joined
·
9,342 Posts
Computer Specs:
Core i7-3770 @ 3.4Ghz (all the time, C states disabled)
16GB ram
GTX 1650 Super - High performance power state selected (never downclocks)
1TB Samsung SSD (evo 860)

I am watching a movie now using Mirror (silversomething's mirror) for the contrast setting. Any opinions on this curve? So far it seems nice and bright.

I noticed that if I uncheck compromise for hdr luminescence that my rendering times JUMP by 10. From 30 to 40ms - and frames begin to be lost. So NO compromise may not be possible for me.

I already had to drop chroma from AA High to AA Medium with beta 113 in order to keep rendering times around 30.
Yea, a 1650 super just isn't up to the task for this. Surprised you got out of it what you did.
 

·
Registered
Joined
·
8,983 Posts
Hey all - just making the dive into MadVR 113 up from the last 'stable' release.

Looks like I have some settings to play with now!

Quick thing I noticed - when I tried to use NeoXP High 1.0 my rendering times SHOT up - started losing frames. All other settings identical, if I switch to mirror, log/hill/loghill strong they all play fine.

Having a tough time with so many options. I am not a professional calibrator, so don't know what things are 'supposed' to look like.

Where do I select the NO Compromise option discussed above? Is that in the trade quality for performance tab?
You actually want Compromise ON. For Gamut, this will give you best performance and no saturation issues for the moment.

Interested to know your render time when you check the below option. Might be worth knowing how it performs with a card like yours, some people cant afford higher end GPU's nor want them if perhaps they might build a cheap PC for a room which is not their main rig... I know I am considering the option.


3050055


Put chroma up-scaling on Low.

3050058
 

·
Registered
Joined
·
139 Posts
JAVS - So switch the compromise from luminescence to tone/gamut? Right now I have it the opposite - and it is playing fine. Also, Chroma at AA Medium is playing fine too. Is 30ms render time ok? Optimal? High?

I will take a screen shot in a bit and post my exact settings right now (which are playing perfectly fine).
 

·
Registered
Joined
·
8,983 Posts
JAVS - So switch the compromise from luminescence to tone/gamut? Right now I have it the opposite - and it is playing fine. Also, Chroma at AA Medium is playing fine too. Is 30ms render time ok? Optimal? High?

I will take a screen shot in a bit and post my exact settings right now (which are playing perfectly fine).
We dont want to clog this thread any more, but the luma compromise can have very bad artifacts, whereas the gamut just has hue shift (which is not really bad at all IMO), ive been using it that way all year. You will gain more performance with the gamut comp ticked vs luma.

Right so you are getting 30ms just as is, thats not bad, so that GPU is fine. Just as I thought. I dont subscribe to the idea that you need a super computer to run MadVR. 30ms is just fgine, I typically would not want to see higher than about 35 for never every dropping frames. Sometimes it can spike for a moment so a little headroom is good. 41ms is the sync time at 24p, so you need to be under that.
 

·
Premium Member
Joined
·
9,951 Posts
@Javs two things:

1) The gamut saturation issue gets worse as peak brightness goes down, so it might be worth reminding people of that important caveat. At 120nits (my peak brightness) there is much less of an issue than at 75nits (yours, off memory).

2) Are you using NGU Sharp for Chroma, and please could you check if using NGU AA for chroma makes any difference? I can't test this here as I don't see anything as offensive as what you observe, but apparently some notice a saturation shift with different chroma upscaling settings, which sounds weird but might be worth checking when you have a minute?

Despite the peak brightness difference, I'm still wondering why you have such a big issue with gamut saturation when I don't.

I really wouldn't want madVR to become undersaturated so that it looks good with your settings, if there is any kind of variability in the results depending on settings.

Not saying that there is no issue, saturation definitely needs another pass, it's been planned for ages, but I'm wondering why it's so offensive at yours and not here.

I have no taste for oversaturation or burnt faces, I assure you :)

In any case, it's great that you plan to spend time with Mathias to get to the bottom of this, I'm looking forward to the results. :)
 

·
Registered
Joined
·
139 Posts
@Javs

Here are pics of my settings - I will switch the luma/gamut ones and see if it is still around 30ms.

I am using: D3D11 (native), Chroma @ AA Medium, Luma @ Sharp High

**ONE important thing I noticed is that my video ram usage is VERY high. 3924MB (out of 4GB) so I am pretty much peaked - which could result in some playback issues on some scenes.

Is there a way to reduce video ram WITHOUT reducing quality in any way? Something like reducing the amount of frames presented in advance to 4 instead of 8?? Does that help video card ram at all? Any other ideas would be appreciated!**
3050112
3050113
 

·
Registered
Joined
·
8,983 Posts
@Javs two things:

1) The gamut saturation issue gets worse as peak brightness goes down, so it might be worth reminding people of that important caveat. At 120nits (my peak brightness) there is much less of an issue than at 75nits (yours, off memory).

2) Are you using NGU Sharp for Chroma, and please could you check if using NGU AA for chroma makes any difference? I can't test this here as I don't see anything as offensive as what you observe, but apparently some notice a saturation shift with different chroma upscaling settings, which sounds weird but might be worth checking when you have a minute?

Despite the peak brightness difference, I'm still wondering why you have such a big issue with gamut saturation when I don't.

I really wouldn't want madVR to become undersaturated so that it looks good with your settings, if there is any kind of variability in the results depending on settings.

Not saying that there is no issue, saturation definitely needs another pass, it's been planned for ages, but I'm wondering why it's so offensive at yours and not here.

I have no taste for oversaturation or burnt faces, I assure you :)

In any case, it's great that you plan to spend time with Mathias to get to the bottom of this, I'm looking forward to the results. :)
1. Thanks, you just reminded them yourself I guess. But remember, I dont test just at 75 nits, I test at over 100 on the UHDTV I use as my monitor and everything I have shared is at 100 nits peak here. Please lets not discuss the potentials of this until later on, we are not looking at this yet. The issue is less pronounced but still there with things like a P3 LUT and BT2020, but its amplified if you convert to 709. Just a data point.

2. NGU Sharp is onnly a recent thing, all the images before I shared are actually AA does not solve it, or even visbily affect it at all to be honest.

The saturation thing is definitely not always there. I think you are expecting burnt faces, but I only see it to that extreme very rarely. I would not say its burnt faces. The Greatest Showman shot is a great test because there are no VFX and it seems to be highly affected in all colour gamuts. But also that kind of shot is not in every scene in every film.

Dont worry I am making some tests on this end which will pretty surely show us what things are meant to look like.

Likewise though, I think its fair this becomes a proper option 'Saturation Recovery' and allow us to set the intensity of it, some of us like the slightly more relaxed saturation of beta 10 (I dont know if you have looked at that recently, but that seriously looks good to me and is hue perfect).

:)
 

·
Registered
Joined
·
8,983 Posts
@Javs

Here are pics of my settings - I will switch the luma/gamut ones and see if it is still around 30ms.

I am using: D3D11 (native), Chroma @ AA Medium, Luma @ Sharp High

**ONE important thing I noticed is that my video ram usage is VERY high. 3924MB (out of 4GB) so I am pretty much peaked - which could result in some playback issues on some scenes.

Is there a way to reduce video ram WITHOUT reducing quality in any way? Something like reducing the amount of frames presented in advance to 4 instead of 8?? Does that help video card ram at all? Any other ideas would be appreciated!**
Your ram usage is not really ram used as shown in task manager, its actually more like allocated. Dont sweat it at all. Fill up like that with most things you do with your GPU.
 

·
Registered
Joined
·
139 Posts
@Javs

Ok - the video ram was measured using GPU-Z (setting the stat to show MAX usage). If it is just an allocation, no worries - but I thought I saw some slow down in a scene or two (no frame drops or glitches during that scene). GPU processor is only hitting 71% MAX - so its not the GPU maxing out. Only thing I could think of was the RAM being so close.

The 'slow down' was so minor that it was almost not noticeable, I am just very sensitive to dropped frames etc - especially when viewing on 130" screen!

I haven't made the 'sacrifice' switch yet - so I will see if that yields any difference later tonight.
 

·
Registered
Joined
·
8,983 Posts
@Javs

Ok - the video ram was measured using GPU-Z (setting the stat to show MAX usage). If it is just an allocation, no worries - but I thought I saw some slow down in a scene or two (no frame drops or glitches during that scene). GPU processor is only hitting 71% MAX - so its not the GPU maxing out. Only thing I could think of was the RAM being so close.

The 'slow down' was so minor that it was almost not noticeable, I am just very sensitive to dropped frames etc - especially when viewing on 130" screen!

I haven't made the 'sacrifice' switch yet - so I will see if that yields any difference later tonight.
You should be using Ctr-J to look at things like render times and dropped frames, it should show you the frame queue, that will tell you if its healthy or not, rather than GPUZ or the like.

Enough of that here, there is a MadVR support thread elsewhere.
 

·
Registered
JVC NX5 at 140", Denon X4200W (5.1.2) with Axiom Audio speakers + Bass Shakers
Joined
·
2,194 Posts
If people want more or less saturation, why can't they just use the color saturation control in the color & gamma madVR tab?

Hue shift is the main problem that should be addressed. I agree there is too much saturation at low nits too, but this can be fixed by going to like -5 color saturation or so in my experience. Not sure we need another color saturation control.

Though you do need a separate profile for HDR and SDR then so you aren't under-saturating SDR.
 

·
Registered
Joined
·
8,983 Posts
If people want more or less saturation, why can't they just use the color saturation control in the color & gamma madVR tab?

Hue shift is the main problem that should be addressed. I agree there is too much saturation at low nits too, but this can be fixed by going to like -5 color saturation or so in my experience. Not sure we need another color saturation control.

Though you do need a separate profile for HDR and SDR then so you aren't under-saturating SDR.
Because that's global saturation, I think what is happening here is actually targeted saturation and it only affects certain scenes.

It's not as simple as a global setting
 

·
Premium Member
Joined
·
9,951 Posts
1. Thanks, you just reminded them yourself I guess. But remember, I dont test just at 75 nits, I test at over 100 on the UHDTV I use as my monitor and everything I have shared is at 100 nits peak here. Please lets not discuss the potentials of this until later on, we are not looking at this yet. The issue is less pronounced but still there with things like a P3 LUT and BT2020, but its amplified if you convert to 709. Just a data point.

2. NGU Sharp is onnly a recent thing, all the images before I shared are actually AA does not solve it, or even visbily affect it at all to be honest.

The saturation thing is definitely not always there. I think you are expecting burnt faces, but I only see it to that extreme very rarely. I would not say its burnt faces. The Greatest Showman shot is a great test because there are no VFX and it seems to be highly affected in all colour gamuts. But also that kind of shot is not in every scene in every film.

Dont worry I am making some tests on this end which will pretty surely show us what things are meant to look like.

Likewise though, I think its fair this becomes a proper option 'Saturation Recovery' and allow us to set the intensity of it, some of us like the slightly more relaxed saturation of beta 10 (I dont know if you have looked at that recently, but that seriously looks good to me and is hue perfect).

:)
Thanks for this. I’m not surprised by the lack of impact of the chroma upscaling setting, but it was worth checking. I use a P3 LUT and never tonemap HDR content to ref-709 so that might explain why, along with my 120nits peak, the issue is less visible, if at all, here.

I haven’t looked at any previous build because there is nothing significantly wrong here with 113, at least more than other minor issues that await improvement. I’m also puzzled by your decision to not use some other options and to prefer the performance option that disables a lot of good stuff. We usually have fairly similar taste re picture accuracy, so I have no doubt that what you see and leads you to make these corrections is different from what I see here.

I have some issues with some of the contrast curves, including mine, so I use a very conservative one, otherwise it leads to brightness jumps etc that I find a lot more jarring than any minor oversaturation. I disable the shadow restore option because I never liked what it can do to contrast in order to restore shadow detail that I never felt was meant to be seen. I use a fairly moderate setting for restore highlights too. But I don’t experience the saturation issues you report.

I think it’s great to improve things for users with a low peak brightness, as they are probably in the majority, but please don’t call the option “accurate”. It might be improved accuracy for low light or undersaturated gamut, but it might be misleading for users with higher peak brightness (above 100nits) and full P3 cover, who can already get fairly accurate results with the current build and a P3 LUT, at least as far as I can see. Using that new option might lead them to under saturated results If it tunes down what they are getting currently with P3.

There is no point in having full P3 cover and Dolby Cinema level of peak brightness (around 106 nits) if it is to end up with rec-709 like results.

Needless to say that if your tests show that this option is what things are meant to look like with 120nits and full P3 cover, I’ll be using it :)

Thanks again for making the time to look into this, I’m sure we’ll all benefit from this.
 

·
Premium Member
Joined
·
9,951 Posts
One thing I forgot, in order to be able to compare LUT on/off with HDR pixel shader in madVR and get correct results, you need three things:

  • the baseline calibration on the PJ has to be a gamma 2.2 and a full JVC Autocal has to be performed and known to provide good results without a 3D LUT.
  • the LUT target has to be a gamma 2.2
  • you need to make sure that the gamma target in madVR when not using the LUT calibration but the “This display is already calibrated for” is set to 2.2 as well (and the correct gamut obviously). Otherwise, when you toggle the LUT on/off, madVR applies the settings that aren’t visible when a LUT calibration isn’t selected.
If you don’t do this, in my experience you get inconsistent results when comparing LUT on/off.

On the JVCs, unless you use profile off with the filter and a very large LUT with a software known to not deliver poor large LUTs, I wouldn’t use a 3D LUT for these tests. The JVCs compress the edges of the gamut as soon as you use a color profile, as it enables the CMS. That can cause very weird things with the LUT applied, especially if your baseline doesn’t cover at least 100% of the target gamut.

For true color accuracy, you need to be using profile off with the filter and a large LUT, with the JVC CMS entirely disabled (no factory color profile preset used, hence no Autocal). That’s what I would be using if I was testing this, in order to rule out entirely the buggy JVC CMS.

Currently, you need to enable the filter manually (or with a HD Fury Maestro/Diva) as profile off is only available without the filter, but the next JVC f/w for the 4K models will finally offer profile off / filter on as an option (After years of lobbying!)

If you haven’t tried it yet, it would be a good thing to do before you do your tests, in order to rule out a calibration issue. Of course if you use another display only the first part of this post applies (the gamma targets / settings).

Which JVC model will you be using for these tests?
 

·
Registered
Joined
·
8,983 Posts
I will be renting an HDR reference monitor for these tests. It takes all this JVC stuff out of the equation as they can be properly calibrated to tight specs.

These monitors have pro 100 nit and 1000 nit modes for this reason.

I don't have time to address the rest but thanks.

If I use a JVC it will be my 9500. A Panasonic FZ950 OLED.

p.s I never called compromise on accurate, at the same time we don't really know that it's even accurate at the moment with comp off, hence the motivation behind renting pro cinema gear on my dime to find out.

Please also understand I'm seeing this in every gamut and mode, lut, no lut, whatever. We are not going to get to the bottom of it with a set of settings, way past that point. It happened right when Mathias started adding surface mode etc after beta 10, traced it all the way back to then.
 

·
Premium Member
Joined
·
9,951 Posts
Sounds great!

Just to clarify, I didn't mean you said that compromise on was accurate (or more accurate than compromise off), I meant the NEW setting / option, whatever it is.

It does sound, given your methodology, that it might deserve to be called accurate if it's tested in all these situations with reference monitors, so looking forward to it :).

The first part of my post still applies if you want to be able to compare LUT on/LUT off, reference monitor or not.
 

·
Registered
Joined
·
8,983 Posts
Sounds great!

Just to clarify, I didn't mean you said that compromise on was accurate (or more accurate than compromise off), I meant the NEW setting / option, whatever it is.

It does sound, given your methodology, that it might deserve to be called accurate if it's tested in all these situations with reference monitors, so looking forward to it :).

The first part of my post still applies if you want to be able to compare LUT on/LUT off, reference monitor or not.
Thanks, it's all good to know!

I did just get a spectro (efs2000) also thanks to your advice a few weeks back, and it's already been really useful. Good to know things are accurate. My iD3 was out a bit in green.
 

·
Registered
Joined
·
560 Posts
I posted this in the Doom9 forum, but thought this might be the more appropriate forum.

Is anyone else using a 3080 or 3090 GPU yet with HDR - SDR? I upgraded my HTPC with an EVGA 3080 FTW3 Ultra and I can only get MPC-BE (1.5.5.5433) with its internal filters to convert HDR properly.

If I use anything with LAV filters such as MPC-HC or kodi DiSplayer I get extreme blown highlights when in Fullscreen windowed and FSE. However, it’s fine when windowed. This is with the older internal LAV and newer 0.74.1 via external.

The attached pictures show the (1] excess highlights (not sure if this is the correct term) when fullscreen and (2) correction to normal when windowed (and what it looks like with correct conversion using MPC-BE’s internal filters when in fullscreen windowed or FSE).

This happens with MPC-BE as well when I use LAV External with 0.74.1. However, its fine with the included internal filters.

This occurs with 456.38 and 456.71 drivers. I’ve also tried 8 and 10 bit, and RGB &YCbCr.

I also just reinstalled Windows 10 this past weekend and tested MadVR before installing any other program. Same results as when I noticed this a couple of weeks ago.

3050240

3050241
 
9561 - 9580 of 10380 Posts
Top