madVR - high quality video renderer (GPU assisted)

[Phoenixia]

1. the curve you want. if you tone map you need 2 3D LUT cause the one for HDR needs to be 2.2.
2. what do you want? with a perfect black device it does not matter else it matters a lot.
3. who knows is it on by default?
4. no. meter or bust.

ICC profiles are useless without a colorimeter.

Every TV is different. Often by an extreme amount.

You also have to know the EXACT setting/condition that ICC profile was created, because different TV modes, settings will drastically alter behavior.

But it ALL starts with buying a colorimeter, avoid the Spyder brand, they're useless for OLED, get a calibrite. Doesn't really matter which one, get the one rated for lower brightnesses, the 3000nit one has worse near black stability.

If you can find a new old stock, the OLD x-rite i1d3 (Pre rebrand Calibrite) are the Best for OLED, or CCDIS3 PRO

141%, I'm assuming you have a qdOLED.

ALSO, you don't want to use 3DLUT to watch most movies. For most movies, you just want the 2D corrections, Gamma, RGB curve, Whitepoint.

If you use 3DLUT, it will clamp to either DCIp3 or Rec709(srgb), which rarely represent the author's intent, only the edges of the container spec..

On OLED, a 3DLUT for SDR will actually function properly (Though, you SHOULD NOT use one for 99.99% of movies).. But HDR 3DLUT will not work at all, due to how ABL algorithms behave, completely unpredictably, and MADVR doesn't know what's happening. The test pattern output does not have high correlation to real-content, so you can't really characterize HDR on consumer-use OLED.


In 99.99% of cases, HDR or SDR, you want Gamma/RGB/Whitepoint, and a reasonable HDR EOTF roll off setting.

DO NOT use an ICC from the internet, it's 100% wrong, 100% of the time.

The only thing you can download are Spectral Corrections to use with your Colorimeter (once you get one), which are typically close enough for general use.

Spectral correction, done correctly reflects the optical technology stack of the panel, whereas ICC includes all the firmware and software who knows what in between.

I have the OSD settings for each profile and since it's an IPS panel I believe there are less variations between units so ΔE should be tolerable, contrast is between ~700 to ~1000, the reason why I'm using a 3dlut in the first place because everything looks oversaturated like neon and the monitor sRGB mode locks many settings and still not as good as a 3dlut in my opinion, most contents I watch are either YouTube or SDR movies so source should be Rec.709, I don't know if this is useful but madVR is identifying the monitor as BT.2020, my questions remain the same (link) (https://forum.doom9.org/showthread.php?p=2030092#post2030092), below are profiles information using DisplayCAL-profile-info.exe:

First Profile :
Size 968420 Bytes (945.72 KiB)
Preferred CMM 0x6172676C 'argl' ArgyllCMS
ICC version 2.2
Profile class Display device profile
Color model RGB
Profile connection space (PCS) XYZ
Created 2022-06-13 16:14:07
Platform Microsoft
Is embedded No
Can be used independently Yes
Device
Manufacturer 0x0000541C GIGA-BYTE TECHNOLOGY CO., LTD.
Model 0x00002718
Media attributes Reflective, Glossy, Positive, Color
Default rendering intent Perceptual
PCS illuminant XYZ 96.42 100.00 82.49 (xy 0.3457 0.3585, CCT 5000K)
Creator 0x4443414C 'DCAL' DisplayCAL
Checksum 0x0C2CE5CE4899FCA3E6A0B8C6B831B397
Checksum OK Yes
Description (ASCII) M27Q P #3 2022-06-13 16-04 0.3127x 0.329y sRGB F-S
XYZLUT+MTX
Copyright No copyright. Created with DisplayCAL 3.8.9.3 and
ArgyllCMS 2.1.2
Device manufacturer name (ASCII) GIGA-BYTE TECHNOLOGY CO., LTD.
Device model name (ASCII) M27Q P
Luminance 200.21 cd/m²
Media white point
Is illuminant Yes
Illuminant-relative XYZ 94.92 100.00 108.83 (xy 0.3125 0.3292)
Illuminant-relative CCT 6512K
ΔE 2000 to daylight locus 0.25
ΔE 2000 to blackbody locus 4.84
Media black point
Illuminant-relative XYZ 0.1114 0.1175 0.1266 (xy 0.3133 0.3305)
Illuminant-relative CCT 6460K
ΔE 2000 to daylight locus 0.62
ΔE 2000 to blackbody locus 5.14
Absolute to media relative transform Bradford (ICC recommendation)
Matrix 0.8951 0.2664 -0.1614
-0.7502 1.7135 0.0367
0.0389 -0.0685 1.0296
Video card gamma table
Bitdepth 16
Channels 3
Number of entries per channel 256
Channel 1 gamma at 50% input 1.01
Channel 1 minimum 0.0000%
Channel 1 maximum 98.77%
Channel 1 unique values 256 @ 8 Bit
Channel 1 is linear No
Channel 2 gamma at 50% input 1.01
Channel 2 minimum 0.0000%
Channel 2 maximum 99.86%
Channel 2 unique values 256 @ 8 Bit
Channel 2 is linear No
Channel 3 gamma at 50% input 1.00
Channel 3 minimum 0.0000%
Channel 3 maximum 99.21%
Channel 3 unique values 256 @ 8 Bit
Channel 3 is linear No
Device to PCS: Intent 0
Matrix 1.0000 0.0000 0.0000
0.0000 1.0000 0.0000
0.0000 0.0000 1.0000
Input Table
Channels 3
Number of entries per channel 2049
Color Look Up Table
Grid Steps 33
Entries 35937
Output Table
Channels 3
Number of entries per channel 2
Red matrix column
Illuminant-relative XYZ 47.35 21.16 0.29 (xy 0.6882 0.3076)
PCS-relative XYZ 50.14 22.38 0.10 (xy 0.6904 0.3082)
Green matrix column
Illuminant-relative XYZ 28.39 71.24 6.85 (xy 0.2666 0.6691)
PCS-relative XYZ 31.09 71.27 5.96 (xy 0.2871 0.6579)
Blue matrix column
Illuminant-relative XYZ 19.19 7.59 101.69 (xy 0.1494 0.0591)
PCS-relative XYZ 15.19 6.36 76.43 (xy 0.1550 0.0649)
Red tone response curve
Number of entries 256
Transfer function sRGB
Minimum Y 0.0000
Maximum Y 100.00
Green tone response curve
Number of entries 256
Transfer function sRGB
Minimum Y 0.0000
Maximum Y 100.00
Blue tone response curve
Number of entries 256
Transfer function sRGB
Minimum Y 0.0000
Maximum Y 100.00
Device to PCS: Intent 1
Matrix 1.0000 0.0000 0.0000
0.0000 1.0000 0.0000
0.0000 0.0000 1.0000
Input Table
Channels 3
Number of entries per channel 2049
Color Look Up Table
Grid Steps 33
Entries 35937
Output Table
Channels 3
Number of entries per channel 2
PCS to device: Intent 1
Matrix 4.7569 -1.9712 -0.7461
-1.4673 3.3766 0.0462
0.1380 -0.2310 2.5432
Input Table
Channels 3
Number of entries per channel 4096
Color Look Up Table
Grid Steps 33
Entries 35937
Output Table
Channels 3
Number of entries per channel 256
PCS to device: Intent 0
Matrix 4.7695 -1.9794 -0.7508
-1.4763 3.3870 0.0442
0.1345 -0.2331 2.5499
Input Table
Channels 3
Number of entries per channel 4096
Color Look Up Table
Grid Steps 33
Entries 35937
Output Table
Channels 3
Number of entries per channel 256
Characterization target [23257 Bytes]
Characterization device values [23257 Bytes]
Characterization measurement values [23257 Bytes]
Chromaticity (illuminant-relative)
Channel 1 (R) xy 0.6882 0.3076
Channel 2 (G) xy 0.2666 0.6691
Channel 3 (B) xy 0.1494 0.0591
Make and model
Manufacturer 0x1C54 GIGA-BYTE TECHNOLOGY CO., LTD.
Model 0x2718
Metadata
prefix EDID_
CMF_
DATA_
MEASUREMENT_
OPENICC_
MAPPING_
ACCURACY_
GAMUT_
EDID_mnft GBT
EDID_mnft_id 7252
EDID_model_id 10008
EDID_date 2021-T48
EDID_red_x 0.685546875
EDID_red_y 0.30859375
EDID_green_x 0.2646484375
EDID_green_y 0.66796875
EDID_blue_x 0.150390625
EDID_blue_y 0.0576171875
EDID_white_x 0.3134765625
EDID_white_y 0.3291015625
EDID_manufacturer GIGA-BYTE TECHNOLOGY CO., LTD.
EDID_gamma 2.2
EDID_model M27Q P
EDID_serial 21483B000127
EDID_md5 b996730acf7800a315731ec786c28ff7
CMF_binary DisplayCAL
CMF_version 3.8.9.3
CMF_product DisplayCAL
License Public Domain
Quality high
OPENICC_automatic_generated 0
DATA_source calib
MEASUREMENT_device i1 displaypro, colormunki display
MAPPING_device_id xrandr-GIGA-BYTE TECHNOLOGY CO., LTD.-M27Q
P-21483B000127
ACCURACY_dE76_avg 0.064494
ACCURACY_dE76_max 1.103100
ACCURACY_dE76_rms 0.153430
GAMUT_volume 1.41511018917
GAMUT_coverage(dci-p3) 0.9449
GAMUT_coverage(srgb) 0.9981
GAMUT_coverage(adobe-rgb) 0.8573




Second Profile :
Size 6960 Bytes (6.80 KiB)
Preferred CMM 0x41444245 'ADBE' Adobe
ICC version 2.0
Profile class Display device profile
Color model RGB
Profile connection space (PCS) XYZ
Created 2023-08-08 14:55:21
Platform Microsoft
Is embedded No
Can be used independently Yes
Device
Manufacturer 0x00000000
Model 0x00000000
Media attributes Reflective, Glossy, Positive, Color
Default rendering intent ICC-absolute colorimetric
PCS illuminant XYZ 96.42 100.00 82.49 (xy 0.3457 0.3585, CCT 5000K)
Creator 0x5350434C 'SPCL'
Checksum 0x6450F3B6EB916A3DA6098BB0008F9405
Checksum OK Yes
Copyright SpectraCal Inc, 2023
Device manufacturer name (ASCII) SpectraCal LLC
Device model name (ASCII) M27Q P
Media white point
Is illuminant Yes
Illuminant-relative XYZ 96.50 100.00 108.68 (xy 0.3162 0.3277)
Illuminant-relative CCT 6323K
ΔE 2000 to daylight locus 3.73
ΔE 2000 to blackbody locus 1.14
PCS-relative XYZ 96.42 100.00 82.49 (xy 0.3457 0.3585)
PCS-relative CCT 5000K
Media black point
Illuminant-relative XYZ 0.1153 0.1136 0.2706 (xy 0.2307 0.2274)
PCS-relative XYZ 0.1083 0.1114 0.2045 (xy 0.2554 0.2626)
PCS-relative CCT 18588K
Red tone response curve
Number of entries 1024
Transfer function ≈ Gamma 2.22 100% (Δ 2.16%)
Minimum Y 0.0000
Maximum Y 100.00
Green tone response curve
Number of entries 1024
Transfer function ≈ Gamma 2.19 100% (Δ 3.95%)
Minimum Y 0.0275
Maximum Y 100.00
Blue tone response curve
Number of entries 1024
Transfer function Unknown
Minimum Y 0.1114
Maximum Y 100.00
Red matrix column
Illuminant-relative XYZ 51.04 22.73 0.24 (xy 0.6897 0.3071)
PCS-relative XYZ 53.24 23.69 0.06 (xy 0.6915 0.3077)
Green matrix column
Illuminant-relative XYZ 26.85 70.14 7.92 (xy 0.2559 0.6685)
PCS-relative XYZ 28.59 70.41 6.82 (xy 0.2702 0.6654)
Blue matrix column
Illuminant-relative XYZ 18.60 7.13 100.52 (xy 0.1473 0.0565)
PCS-relative XYZ 14.59 5.90 75.61 (xy 0.1518 0.0614)
Chromatic adaptation transform Bradford (ICC recommendation)
Matrix 1.0359 0.0164 -0.0477
0.0198 0.9980 -0.0158
-0.0094 0.0157 0.7529
Description (ASCII) M27Q P external




Third Profile (provided by manufacturer) :
Size 6940 Bytes (6.78 KiB)
Preferred CMM 0x41444245 'ADBE' Adobe
ICC version 2.1
Profile class Display device profile
Color model RGB
Profile connection space (PCS) XYZ
Created 2022-05-18 10:06:08
Platform Microsoft
Is embedded No
Can be used independently Yes
Device
Manufacturer 0x47425400
Model 0x73524742 'sRGB'
Media attributes Reflective, Glossy, Positive, Color
Default rendering intent Perceptual
PCS illuminant XYZ 96.42 100.00 82.49 (xy 0.3457 0.3585, CCT 5000K)
Creator 0x47425420 'GBT '
Checksum 0x2DB57FAF7D816A44B00E138F97B624FE
Checksum OK Yes
Copyright Copyright (c) 2022 GIGABYTE
Description (ASCII) GIGABYTE M27Q P
Media white point
Is illuminant Yes
Illuminant-relative XYZ 96.42 100.00 82.49 (xy 0.3457 0.3585)
Illuminant-relative CCT 5000K
ΔE 2000 to daylight locus 0.08
ΔE 2000 to blackbody locus 4.73
Media black point
Illuminant-relative XYZ 0.0000 0.0000 0.0000
Red matrix column
Illuminant-relative XYZ 54.44 24.52 0.40 (xy 0.6860 0.3090)
PCS-relative XYZ 54.44 24.52 0.40 (xy 0.6860 0.3090)
Green matrix column
Illuminant-relative XYZ 27.76 69.98 7.02 (xy 0.2650 0.6680)
PCS-relative XYZ 27.76 69.98 7.02 (xy 0.2650 0.6680)
Blue matrix column
Illuminant-relative XYZ 14.22 5.50 75.07 (xy 0.1500 0.0580)
PCS-relative XYZ 14.22 5.50 75.07 (xy 0.1500 0.0580)
Device manufacturer name (ASCII) https://www.gigabyte.com
Device model name (ASCII) GIGABYTE M27Q P
Technology Active matrix display
Red tone response curve
Number of entries 1024
Transfer function sRGB
Minimum Y 0.0000
Maximum Y 100.00
Green tone response curve
Number of entries 1024
Transfer function sRGB
Minimum Y 0.0000
Maximum Y 100.00
Blue tone response curve
Number of entries 1024
Transfer function sRGB

[tp4tissue]

You're asking the wrong questions.

Your work flow does not make sense for what you're trying to accomplish.

Youtube does not work with ICC. The gpu driver will respect the gamma/rgb/whitepoint adjustment in the profile, but this has nothing to do with 3dLUT.

Only color-aware applications like photoshop, will do 3D correction from an ICC.

But again, you do not want 3d correction for the vast majority of the content, because it will clamp to srgb or rec709, your brand new 141% monitor will look like your grandma's 1984 CRT tv.

YOU DO NOT want rec709 or srgb. These are the edges of the container, that does not mean the artists WANT their content to appear this way. RED apples and firetrucks will look pink/salmon.

That ICC is WRONG for your monitor, it is no more right whether it's oled or ips. The variation between panel is enormous.


Another thing regarding calibration is, it's only accurate for 300 hours, AFTER 300 hours, the backlight burns away, and changes color, affecting the white point, the panel itself after many heat cycles and dries out, + under different temperature conditions will behave differently across the rgb curve and especially near black/near white.

When they market monitor accuracy Delta E<3, that's a SCAM, it's entirely meaningless because of backlight drift.

MADVR is also not a color aware application, it can only take a static 3dLUT table manually. So there is a very specific procedure.


If you don't want to buy a colorimeter, don't use an ICC or attempt 3Dlut, it doesn't make any sense.

The SRGB mode on this monitor is basically exactly what the 3DLUT will look like in terms of "Color clamping". The only difference the srgb mode will look more lifted/washed-out near black due to the srgb tone curve which always look trash on any monitor. But again, this container is irrelevant for content consumption, it is NOT meant for that.

You're better off just using the lower saturation hotkey set in Madvr hotkey.

[Sunspark]

Yes and no..

If things are horrifically off, then you actually do need an icc or at least the settings to enter in if it was professionally done from another display if it's the same device.

This TV I'm on was known to not have the right settings by quite a lot. If I didn't use the spectracal settings that were published in a review then everything would look worse. Even with that, it's not perfect, but it's pretty alright now.

I would like to buy a calibrator sometime though. I had an early Spyder years ago. It was so-so, I ended up not using it because I didn't feel it was all that good and the filters on colorimeters fade away. I'd like to try a spectrophotometer.

[tp4tissue]

For madvr, it doesn't make sense to spend money on a spectro.

Consumer displays don't have the hardware stability such that they'd HOLD very accurately to a profile.

The limit is the display, not the measurement device.

Legit spectros are expensive, $5000 used.

Spectros are also slow as hell if you use it for lots of patches.

Colorimeter filters are good for at least 10 years, longer if they're glass and kept in the dark.

Xrites probes are really cheap for what they can do, but it's almost a small hobby onto itself in the beginning, and can be confusing because those guys on displaycal forum are really fatigued from answering the same questions for the 9999th time.

The spyder is much slower and less consistent, and can't measure oled.

ICC from the internet rarely helps, because offsets are always quite different between monitors.

The only time it might help (despite being wrong), is if gamma curve is really really off, which is seldom the case with newer name brand monitors.


TV settings are good for things like MODES, and possibly contrast, but they're not good for rgb offsets, just asking for trouble.

[tp4tissue]

Madvr doesn't work on snapdragon cpu windows does it? NE1 know?

[Sunspark]

As far as I know there is no inline assembler code, so if snapdragon windows has a translation layer from x86, it should work fine since all madvr is for the most part is algorithms and calls to d3d apis.

[tp4tissue]

Those snapdragon laptops are $400 cheaper than intel/amd equivalents. Wonder if anyone's tried them with Madvr. :eek:

[mclingo]

been having MADVR drop frames and freezes, basically since I moved to win11, through several versions or win and madvr, nothing I tried would fix, driied drivers and a gazillion different settings in MADVR, LAV and GPU etc, literally lost weeks of my life trying to fix this.

This weekend my OS drive got corrupted and was irretrievable so I had to rebuild my entire HTPC/SERVER/Gaming machine, this took me 3 days in all which is why I hadnt attempted this before now. I suspect as I hadnt reimaged my machine since I installed win11 over the top of win10, something broke.

Thankfully i'm back to perfect playback in MADVR now :)

[tp4tissue]

You always want fresh install. Software these days are so bloated, ontop of the sweeping changes they're making to Win11 to become an "Agentic OS."

[mclingo]

if it wasn't such a massive ball ache to do it I would have done it But this is my server with 12 discs HTPC and a gaming machine with 10TB of steam games amongst family photos and videos and all sorts of stuff, I still have loads to reinstall but I'm just saving it now until I need it. I'm thinking about splitting off my HTPC into its own machine and just using this as a server NAS for movies.

[DragonQ]

I have a 6800 XT and updated my AMD GPU driver from 25.8.1 to 26.3.1 yesterday. Now if I play a video using madVR in MPC-HC, I get a black screen for ~25s (CTRL+J doesn't show stats but audio is fine), then the video starts playing properly. Switching to a different renderer fixes the issue.

Is this a known AMD GPU driver bug? If so is there a latest "working" version I can try? Or a particular madVR feature that is known to cause the issue?

[strumf666]

Do you have dithering set to error diffusion?

[DragonQ]

Do you have dithering set to error diffusion?
Yes, option 2. Switching to Ordered Dithering fixes it, thanks.

So annoying that video features get constantly broken in GPU drivers, they clearly don't test them properly at all.

[mclingo]

i had that exact same issue, Strumf66 is correct that should fix, I just use ordered now and do see any difference.

[tp4tissue]

Nooooooo..... My Dithering Quality !!

Still on 6.1.1, Dithering still works fine.

[clsid]

That AMD bug has existed for long time now already. And since hardly anybody is reporting it to AMD, it likely ain't going to be fix soon either.

[tp4tissue]

The dithering bug?

6.1.1 is from january 18th. My dithering is working fine.

Does it have anything to do with win11 24H2 perhaps which I'm still using instead of going to 25? But forced update is coming up.

[tp4tissue]

Is there a way to run madvr on Linux?

[mclingo]

yes, inside a virtual machine running windows :) - no

[strumf666]

The dithering bug?

6.1.1 is from january 18th. My dithering is working fine.

Does it have anything to do with win11 24H2 perhaps which I'm still using instead of going to 25? But forced update is coming up.
Update your drivers and tell us ;)
I had issues with ED before 25.2.1, not sure about win11 version at the time. I believe, Huhn told me to change it to ordering as a workaround, but haven't been checking after each driver update if it has been fixed.

[clsid]

I think this issue exists since 24.9.1

I have no issues on my laptop with oct25 driver. So it also depends on the GPU model and maybe other factors.

Error Diffusion uses DirectCompute API. So maybe find a DirectCompute benchmark tool and check if that works correctly.

[tp4tissue]

I remember something like black screen bug a long time ago, when I first got 6900xt 2 years ago.

I am on 26.1.1 now, 7900xtx, with win 11 24h2, it does not blackscreen. But the mouse will freeze if i use dx11/copyback, works ok switched to dxva2copyback.

Mouse can still freeze on dxva2copyback, but much more rarely, acceptable side of things.

[mclingo]

I ws toying with getting an XTX until they refused to add FSR 4.0 support, still a great card otherwise and a lot better than me 7800xt which is great for MADVR, awful for VR

I'd go green but no 3D support

[tp4tissue]

Download the optiscaler client, it auto installs optiscaler to your game and it supports int8 version of FSR4.0 for RNDA2 (6xxx) AND RDNA3 (7xxx) cards, fully works with most big titles,

[mclingo]

sadly it doesnt work with VR games, 2d games play fine :)

[Klaus1189]

2D or 3D Games?

[mclingo]

neither, VR Games

[Klaus1189]

I mean what 2D Game? Almost every Game is 3D.

[mclingo]

I think you misunderstood I'm saying that VR games don't play very well with this card It's probably not powerful enough, 2D games non VR games run perfectly fine so I don't really need FSR for those

[Klaus1189]

I understood what you mean, I just wanted to ask if you know the difference between 2D and 3D games. Maybe you mixed it up. This has nothing to do with VR. Look here:
https://videochums.com/article/2d-platformers-vs-3d-platformers-1.jpg

[mclingo]

you've totally lost me I don't really know what you are asking me at this point

[huhn]

you called none VR 2D games which makes little sense.

[tp4tissue]

Has anyone measured one of these new 32" Oled PC Monitors through MadTPG?

I'm getting super weird results. The Image looks more/less fine, and Youtube HDR looks as you'd expect, but measuring through HCFR+Madtpg, WEIRRRD curves.

I know on the forums, people who measure via Calman, they seem to track EOTF correctly.

Something's up with the EDID?

Using latest firmware, and I tried both Nvidia and AMD GPU, same result. Tried both HDMI and DP

This is on a Dell S3225QC, 4K120hz, QDoled.

https://i.imgur.com/9n21Scl.png
https://i.imgur.com/FY5FU2r.png
https://i.imgur.com/LWWga56.png
https://i.imgur.com/Axn3y18.png

[tp4tissue]

Having some troubles with DSC and exclusive mode, has anyone figured out a way to detect if DSC is turned on?

[huhn]

there is one way to beat DSC go into the service menu and disable it.

[tp4tissue]

there is one way to beat DSC go into the service menu and disable it.

Very rarely do we get access to service menus.

Unless huhn's got a connnnect? LOL.. Can you ask him about Dell S3225-QC

[tp4tissue]

Woah, actually in the Service menu, seems the new joystick ones are made by TPV oem.

While PC is turned on, Press joystick, hold, double click power button, press joystick up^.

Unfortunately, no DSC option, unless there's a soooper secret menu somewhere. (T . T)>

[huhn]

if you have no access to it try your luck with CRU it has an option to disable DSC this usually doesn't work.

[tp4tissue]

CRU method works with AMD gpu, but not Nvidia.

[Swede]

It's time to close this thread and start new ones. There is no use having +66.000 posts in one thread not to mention how it just kills the server when all AI-bots reloads the *entire* thread time and time again.
So, please start new ones, preferably with a descriptive titel. I'd rather see one to many thread than *mega* ones.

[huhn]

rip

it's been barely used for madVR anyway.