madVR - high quality video renderer (GPU assisted)

[VBB]

Kind of like people who intentionally set their HDMI levels wrong on their Oleds.

Not sure if you meant this as a comment on what I said the other day, but just to clear things up: The reason I intentionally mismatch the black levels on my OLED is that I use RGB full for the output and limited in madVR. Since I don't use my HTPC for anything other than media playback, setting the TV to limited doesn't hurt anything. It just saves one conversion/expansion, whether that's beneficial or not.

[Asmodian]

And without a meter, neither the 709/p3 gamut are accurate So absent that dimension, I'm saying we may as well ensure we're getting that ferrari Red and lamborghini orange.

We do have colorimeter data for this projector model, why are you ignoring it? BT.709 is MUCH more accurate than DCI-P3. Displays are different unit to unit but not that different.

It doesn't even get significantly more saturated reds or oranges, look at the data, the saturation drops off as it nears 100% DCI-P3. The separation between saturation steps drops non-linearly, it is compressing the top of the range to allow the higher saturations in the lower range. Cyan looks especially bad, with the 80% and 100% saturation steps landing on top of each other. This means you lose even more detail in bright saturated colors.

Surly you do not prefer messing up skin tones and most other colors so you can very slightly increase the saturation of the brightest reds and oranges in an incorrect manner? :eek:

madVR will compress the DCI-P3 range into BT.709, with the knee roll off near 100% required by such a conversion. Sending this data to the projector in its DCI-P3 mode results in all the colors the projector could have accurately represented being oversaturated. The data near 100% BT.709 might have represented more saturated colors but the range has been compressed to max out at BT.709 so we don't know anymore. The projector further compresses this range, because its DCI-P3 mode has a similar non-linear range compression built into it, so it ends up oversaturating colors meant to be near 100% BT.709 while still under saturating near 100% DCI-P3. The result is all colors being wrong, and a severe loss of dynamic range at saturations nearing 100% BT.709.

That projector simply cannot display ferrari Red and lamborghini orange so don't sacrifice skin tones chasing the impossible.

[tp4tissue]

Not sure if you meant this as a comment on what I said the other day, but just to clear things up: The reason I intentionally mismatch the black levels on my OLED is that I use RGB full for the output and limited in madVR. Since I don't use my HTPC for anything other than media playback, setting the TV to limited doesn't hurt anything. It just saves one conversion/expansion, whether that's beneficial or not.

If you have a meter, and you measured it, then you know it's working, you know that the contrast isn't lower, you know that the blacks are not crushed, you know that the saturations track evenly.

WITHOUT a meter, it's hard for us on the other side of the internet to make that call for lx2, Those of us who have seen enough calibrated gamma ramps know what to look for, this isn't the case for everyone else just winging it.

[tp4tissue]

We do have colorimeter data for this projector model, why are you ignoring it? BT.709 is MUCH more accurate than DCI-P3. Displays are different unit to unit but not that different.

It doesn't even get significantly more saturated reds or oranges, look at the data, the saturation drops off as it nears 100% DCI-P3. The separation between saturation steps drops non-linearly, it is compressing the top of the range to allow the higher saturations in the lower range. Cyan looks especially bad, with the 80% and 100% saturation steps landing on top of each other. This means you lose even more detail in bright saturated colors.

Surly you do not prefer messing up skin tones and most other colors so you can very slightly increase the saturation of the brightest reds and oranges in an incorrect manner? :eek:

madVR will compress the DCI-P3 range into BT.709, with the knee roll off near 100% required by such a conversion. Sending this data to the projector in its DCI-P3 mode results in all the colors the projector could have accurately represented being oversaturated. The data near 100% BT.709 might have represented more saturated colors but the range has been compressed to max out at BT.709 so we don't know anymore. The projector further compresses this range, because its DCI-P3 mode has a similar non-linear range compression built into it, so it ends up oversaturating colors meant to be near 100% BT.709 while still under saturating near 100% DCI-P3. The result is all colors being wrong, and a severe loss of dynamic range at saturations nearing 100% BT.709.

That projector simply cannot display ferrari Red and lamborghini orange so don't sacrifice skin tones chasing the impossible.


If those measurements are accurate, then it's basically the saturation rendering intent as the default projector's mapping. I wouldn't call that more inaccurate. As for how much more saturated, it's not just about how much further, it's also how / where the curves begin, Assuming Sony is doing their job correctly, we can trust that their secret sauce is the better default, because WE have no other reference point for the projector's other settings. maybe the review got it wrong, maybe they have a setting enabled which makes that yellow wonky, we don't know. They could've gotten a bum projector unit. Having calibrated so many panels, I'm sure you know as well Every single one is different, no 2 alike, and they're almost all off in some way.

I don't think he will be sacrificing skin tones. Only very saturated skin tones will take on a small excess of color, it's nothing like the Orange people we had in 2006 tvs.

As for loss of dynamic range, I'm assuming Sony, the people who invented half this stuff DID what they did for a reason. pushing saturation over perceptual intent is probably an overall better presentation as tested with their focus groups.

[VBB]

I know you're a major advocate of using meters and LUTs, but there are definitely things you can just "wing" without either. Things like contrast, brightness, and sharpness are easily set correctly with the right test patterns. A level mismatch is therefore easy to spot, if you know what to look for. The average enthusiast doesn't absolutely need to know if the saturation tracks evenly, of if the gamma is spot on. In fact, even with a meter, you can't know for sure with HDR and Dolby Vision, especially with the incredibly unstable OLEDs. Like Asmodian suggested, if you don't have accurate numbers for your particular hardware, you go by the data that's out there, and you help madVR make the right decisions by giving it as much info as you can. LUTs are not working properly in the latest build, for example, so it's even more important to use the correct gamut and gamma.

[huhn]

well a 3D LUT can clearly not increass the brightness or the CR either fix the sharpness.

there are some sharpening patterns that only work with a good gamma calibration but that's it.

levels... a meter is one of the last things i would use to fix these.

[el Filou]

I understand why you say that....but, there is difference. Replaced the cable between the pc and the amp, and black levels are now as they should be - it's not a big difference, but its noticeable on the test clips.There is no way a cable itself is changing the levels. Disconnecting the cable may have reset something in the software config of the devices it was connecting. Or there's something hidden in your cable like a converting chip or something that either device detects to change its config (like how early Radeon HD cards detected when a genuine ATI DVI to HDMI adapter was present and only then allowed HDMI signaling). But it was not the cable itself changing the levels.
Did you try switching back to the old cable now that the levels are correct to be sure? Switching several times between them?
If you're happy now it's fine, but erroneously thinking it was the cable means there may be a hidden firmware/software issue in your devices that hasn't been resolved.The goal of colorimetry isn't accurate color persay, it's good LOOKING colors.How can you say something like this when you're one of the most vocal supporters of meters? The only goal is accuracy, good looking colors is just a result of accuracy.

[oldpainlesskodi]

Yes, have retested by swapping cables again and and same results - like it said, it's not changing levels per se, just seems to be clipping on the blacks ever so slightly. Anyway, could be just something weird with my setup, so as long as I am happy, as they say, it's gooood.

[chros]

How can you say something like this when you're one of the most vocal supporters of meters? The only goal is accuracy, good looking colors is just a result of accuracy.
:) Maybe he's wrong with other things but he's right with this one.

[el Filou]

How do you scientifically define 'looking good' when it comes to image reproduction fidelity?

[VBB]

I'm pretty sure chros read that wrong LOL. Of course the only goal is accuracy.

[chros]

:) No, I didn't, I meant it. It's a long story why, I think it's all BS :) For example if you try to run the same measurements 2 times you won't get back the very same results. How can we call this accurate? :) And I could go on: meters, software settings, etc. Not to mention you can easily make things worse than it is by default if you don't know what needs to be done (e.g. using bt1886 with an IPS panel, etc). It was never about accuracy, but pleasing colors :)
Let alone hdr10/dovi calibration where they measure 3 colors... in 2020... :) And if you watch 60% HDR content then you are screwed anyway.
Honestly, if you buy an LG or Panasonic Oled TV today then you don't have to deal with calibration. Fortunately a lot has changed in the last 10 years.
I also use 3dluts, don't get me wrong, but I don't state that is necessary. It's a hobby for me :)

And back to hdr, I just wrote this on the other forum: while we fool around with sdr 3dluts, the average Joe turns on his telly and just enjoys DoVi content on netflix/atv+/etc on his uncalibrated tv :)

[VBB]

I understand what you're saying, but that doesn't dispute that calibration = accuracy. You're right that it's just not worth it right now to chase perfection when it comes to HDR. An inferior meter, and the lack of a spectro to do proper profiling, is a major reason why I stopped creating new LUTs altogether. I agree it's not necessary. It's something you can aim for as a hobby, if you have the time and proper equipment.

I did read your post over at AVS ;) Always been like that, and lately that's exactly what I've been doing, too (the just enjoying content part).

[chros]

Then I don't understand what you guys call accurate :) For me it means repeatable results: if you can replicate your test case 100% with the very same outcome. If you can't that's not accurate, that's something completely different :) And unfortunately the tons of guys who do this by living spread this nonsense to the uneducated public :) But that's only what I think... :) So yes, I agree with @tp4tissue in this.

[VBB]

Accurate in the sense of calibrated to a standard. Of course, you'll always get slightly different results due to meter inaccuracy and fluctuation, panel drift, etc. Believe me, I don't blindly follow the likes of D-Nice, et al, but I will have my next TV calibrated professionally. For my own peace of mind ;)

And BTW, I don't think tp4tissue would agree with you on what you've been saying here today. After all, he touts calibration and LUT use above all else non-stop :P

[SamuriHL]

These OLEDs drift WAY too much to accurately calibrate HDR and DV. Trust me, I tried chasing that and learned a lot in the process. However, that being said, you CAN improve them by calibrating a proper white point balance. And of course, SDR can and should be calibrated properly.

[chros]

:) That's not an issue, that's why I said "in this", meaning accuracy vs pleasing colors :)
Of course there's valid reason for calibration/profiling as well:
- when your panel/mainboard is changed and the unmatched white point hurts your brain :D
- or when dear LG FAQs the gamma with a sudden firmware update resulting brutal black crush :D

[SamuriHL]

THAT never happens! :D LOL

[VBB]

Alright, let's leave it at that. I'm fairly certain that we agree on most things anyway. It just comes down to semantics :)

[chros]

:D Yes, you can put it in this way :D

[tp4tissue]

I'm pretty sure chros read that wrong LOL. Of course the only goal is accuracy.

@ VBB/ el filou,

I am literally Paraphrasing out of a colorimetry textbook. The goal of colorimetry is Good Looking colors.

There is a separation between what colorimetry IS, and the GOAL of colorimetry.

[tp4tissue]

Then I don't understand what you guys call accurate :) For me it means repeatable results: if you can replicate your test case 100% with the very same outcome. If you can't that's not accurate, that's something completely different :) And unfortunately the tons of guys who do this by living spread this nonsense to the uneducated public :) But that's only what I think... :) So yes, I agree with @tp4tissue in this.

Accurate means the Exact color as stipulated by the standards.

Precision is the operation on the repeatability aspect.

[tp4tissue]

Also, if you guys really wanna blow out dat saturation, go with a 2.2g rec709 saturation lut. You go from Red to BLOOD LAKE. Yellow Trees, Irradiated Blue skies.

I've watched entire movies like this, And you know what, It's fine most of the time because colorists use the end of the saturation scale very sparingly, and the lut still preserves proper gamma/whitepoint/greybalance.

You will lose some texture detail on highly saturated tiles, and it clips at places, but meh.. it's mostly-fine.

[SamuriHL]

It's not fine. The latest madvr betas have been trying to fix saturation problems when heavily compressing with DTM. I get that some people love that oversaturated "look and feel" but that doesn't make it correct. madshi is fighting for a scientific approach. I tend to agree.

[70MM]

Hi Guys I hope someone can help me please....
I have just two profiles only, one for SD and the other for HD.
If (srcWidth <= 1050) and (srcHeight <= 768) "SD"
else "HD"

However I want to refine it more to the following:
SD 24fps
SD 25fps
SD 60fps

HD 24fps
HD 25fps
HD 60fps

HDR 24fps
HDR 25fps
HDR 60fps

I have tried and tried to write the script but I just cant make anything work, can someone kindly help me please?

[Asmodian]

if (deintfps < 24.5) AND (srcWidth <= 1050) and (srcHeight <= 768) "SD 24fps"
if (deintfps < 25.5) AND (srcWidth <= 1050) and (srcHeight <= 768) "SD 25fps"
if (deintfps < 61) AND (srcWidth <= 1050) and (srcHeight <= 768) "SD 60fps"
if (deintfps < 24.5) AND (HDR) "HDR 24fps"
if (deintfps < 25.5) AND (HDR) "HDR 25fps"
if (HDR) "HDR 60fps"
if (deintfps < 24.5) "HD 24fps"
if (deintfps < 25.5) "HD 25fps"
else "HD 60fps"

[chros]

Or this (https://www.avsforum.com/threads/guide-building-a-4k-htpc-for-madvr.2364113/post-60096583).

[QBhd]

I always find simple is best:

if (srcHeight < 721) and (deintFps < 25) "720p23"
elseif (srcHeight < 1081) and (bitdepth > 8) "1080p23 HDR"
elseif (srcHeight < 1081) and (deintFps < 25) "1080p23"
elseif (srcHeight < 2161) and (deintFps < 25) "2160p23"
elseif (srcHeight < 721) and (deintFps < 26) "720p25"
elseif (srcHeight < 1081) and (deintFps < 26) "1080p25"
elseif (srcHeight < 721) and (deintFps < 31) "720p29"
elseif (srcHeight < 1081) and (deintFps < 31) "1080p29"
elseif (srcHeight < 2161) and (deintFps < 31) "2160p29"
elseif (srcHeight < 721) and (deintFps < 51) "720p50"
elseif (srcHeight < 1081) and (deintFps < 51) "1080p50"
elseif (srcHeight < 721) and (deintFps < 61) "720p59"
elseif (srcHeight < 1081) and (deintFps < 61) "1080p59"
elseif (srcHeight < 2161) and (deintFps < 61) "2160p59"

QB

[el Filou]

I am literally Paraphrasing out of a colorimetry textbook. The goal of colorimetry is Good Looking colors. There is a separation between what colorimetry IS, and the GOAL of colorimetry.OK I get this, I also get that many meters don't give 100% repeatable results and when you're not sure between two not 100% accurate results it's better to go with the most pleasing, and also that displays drift and have other weaknesses (I still have a plasma, so trust me I know about that), what I meant was that 'good looking' is kinda subjective, and you guys' subjective idea of that notion is simply closer due to experience to a studio mastering technician's when they grade a consumer release, than to someone's who just buys a TV and puts it into torch mode because it 'looks great'.
The goal is to be good looking, but the means by which we try to achieve that goal is to try to approach accuracy as much as possible. With some modern TVs, putting them in the right factory mode may be the only thing that's needed, and meters may not be 100% accurate but I've yet to see a display that looks worse after tuning it with any meter than before it, and if you just keep the factory "best mode" that will also drift with time.
HDR is of course a whole different subject as no consumer display is even currently capable of fully matching the spec.

[tp4tissue]

Hi Guys I hope someone can help me please....
I have just two profiles only, one for SD and the other for HD.
If (srcWidth <= 1050) and (srcHeight <= 768) "SD"
else "HD"


I have tried and tried to write the script but I just cant make anything work, can someone kindly help me please?

The other option is to buy a 1080ti, or even a 3080/3090.

And you won't have to do any of this stuff, just max setting the whole lot. :D

[70MM]

The other option is to buy a 1080ti, or even a 3080/3090.

And you won't have to do any of this stuff, just max setting the whole lot. :D

I already have the 1080ti and I certainly can not max the settings out on SD, HD or HDR.

[tp4tissue]

I already have the 1080ti and I certainly can not max the settings out on SD, HD or HDR.

3090, It's time. :D

[70MM]

Guys thank you so so much for the help on the script after years of not having the changes I need for 24,25 & 60 in SD, HD and HDR I finally have it all working and switching just like magic!!!
Wow Im over thew moon..
Thank you Asmondian, chros and QBhd for going to the time to help me, Im so happy!
My 1080ti does it all and I don’t have tyo go and buy the 3090 now.

[chros]

:D No worries, enjoy!

[HillieSan]

If it's of any help, try win10-radeon-pro-software-enterprise-20.q3.1-sep11.

How does the RX 5700 XT perform with madvr? Can you use NGU very high settings?

I run a RX 480 at NGU high and I am considering the RX 5700 XT. Any advice?

[70MM]

Can some of the experts tell my what NGU settings do they prefer for DVD rips on a screen say 120", with upscaling to 4K with the 1080ti card?

[Asmodian]

120" DVD rips?! :eek:

NGU AA as high as you can set it, and do not expect too much.

[70MM]

120" DVD rips?! :eek:

NGU AA as high as you can set it, and do not expect too much.

I dont expect much from them and hardly ever use them. But using NGU AA doesnt allow x4 upscaling so is that still better with NGU AA without x4?

[chros]

I only use NGU sharp (for chroma as well) with everything, I don't see the point of AA. Compare the 2, see the link in my signature

[VBB]

I find that when upscaling DVDs or lower res stuff in general, it can help quite a bit to add some grain. Otherwise some stuff can look highly artificial. I also prefer AA to sharp in this case, for the same reason.

[70MM]

And you guys that use NGU AA for DVD dont mind not being able to use x4 upscaling as you can with the other NGU settings? Do you think x4 upscaling is work it to help impove things or just sounds like its better?

[huhn]

4x is supposed to be faster not really better.

not sure why NGU sharp is even considered for chroma or a DVD. for NGU to work best you need an relative error free and sharp image both are not really given on a DVD or chroma.

How does the RX 5700 XT perform with madvr? Can you use NGU very high settings?

I run a RX 480 at NGU high and I am considering the RX 5700 XT. Any advice?

why would you consider a 5700 xt now. they are so over priced compared to a 3060 ti or the upcomming 6700.

[oldpainlesskodi]

@HillieSan - given the current landscape, Huhn makes a valid point.

As far as the 5700xt is concerned, it does what I want it to do.

[QBhd]

Yeah I love my 5700 XT... but it is still almost the same price as I paid a year ago... so unless you are getting a great deal on the used market, it makes no sense to buy one now.

QB

[Asmodian]

Do you think x4 upscaling is work it to help impove things or just sounds like its better?

I agree with huhn. x4 is faster but x2 twice is higher quality. I always use x2 twice for quadrupling even for NGU Sharp, it is noticeably better than x4.

It can get complicated trying to find the optimal options if you are performance limited, comparing x4 at a higher quality to x2x2 at lower qualities.

[HillieSan]

why would you consider a 5700 xt now. they are so over priced compared to a 3060 ti or the upcomming 6700.

I am asking because I expect that the prices of the 5700 xt will be dropping. A 3060 is at least €100 more then the 5700 xt. The 5700 xt already drops under the €400. For me around €450 for a graphics card is more than enough for HT.

[HillieSan]

I am looking at the 5700 xt and the 3060 ti. Is the 3060 ti a much better choice then the 5700 xt for madvr? Is it worth it to pay €100 more? I am not playing games.

5700 xt MSI MECH OC 8G is €399
5700 xt ASRock Challenger 8G is €379.90
3060 ti GIGABYTE Gaming OC 8G is €499.

EDIT: I have found a few more 3060 ti cards within my budget.

It looks like the price difference between 5700 xt and 3060 ti is less €50-€100. Interesting. Of course the GPU clock determines the price at well.

What I like about AMD cards is the right refresh rate for HT out of the box. With NVIDIA I had to tweak the refresh rate in the past. Is this still the case?

[aufkrawall]

Ampere has lots of FLOPS, this could/should be helpful for pure shader workloads like madVR. Navi 10 die size also is also not much bigger than Polaris 10, but it consumes more power under high load. Thus I expect that you need a bigger heatsink to get a comparable level of temperatures and noise vs. the 3060 Ti.
While my 5700 XT MSI Gaming is hardly audible after some tweaking, its typical price is rather unattractive vs. the 3060 Ti (let's pretend shops were stocked and the whole market wasn't broken).

[huhn]

the msrp is 400 for the 3060 ti and i'm even seeing them from time to time for that price and it massively faster than a 5700 xt this is not a comparison is a slaughter.

and yes 5700 XT could fall in price in a couple of months or get extreme expensive because they will soon not be produced anymore there is simple no point.

AMD has to follow up with a card in the same price range too.
the msrp for AMD card is at best a lie which is very disappointing.
a even cheaper 3060 will be available "soon" too which should still be faster than a 5700 XT.

[HillieSan]

The RX 480 gets really hot. The fans turn on most of the time and sound like a turbo engine. It's annoying and it has become a problem since I have moved the HTPC closer.

The other problem with the RX 480 is that if the TV and projector are switched on in an incorrect sequence then it takes the card lost of time to switch hdmi and display port. Sometimes it fails and I need to reboot the system.

The RX 480 and madvr with NGU sharp high looks fine. But I think the image on the projector could be better in very high mode and with additional tweaks settings. The RX 480 has no room to experiment. For example I want to experiment with smooth motion on a higher frame rate on the projector. I have used Bluesky Frame Rate Converter in the past and it was amazing. The panning at 60 Hz is so much better than on 24Hz.

The 3060 ti are out of stock. I guess I have to wait until next year.