Display Calibration - Page 23

nevcairiel · 11th December 2021, 00:29

Its quite possible that D3D9 in madVR is a factor. D3D9 is quite limited as it comes to resource management, and the driver/API certainly plays a role there. There aren't enough caches to cache the entire 3DLUT in any case, or a full video texture for that matter, so it has to do a reasonable job at accessing it from video memory directly, or nothing would work.
Like I hinted at above, texture sampling should be a bread and butter operation for any GPU, even if 3D textures are a bit rarer.

huhn · 11th December 2021, 01:18

the vulkan numbers are very convincing if that's even the reason this happens.

this alone is a good reason to move on from dx9 the 19 years of age are also a good reason.

aufkrawall · 11th December 2021, 15:34

I guess the 3D LUT format for madVR can be described as "maxed out". When you generate in IRIDAS format, it is already just a fraction in size.
But probably even ~600KB ReShade format would look fine when properly applied in fp16 precision + dithering.

Balling · 15th February 2022, 16:38

Quote:

Originally Posted by aufkrawall

Well, also on Android you can just enforce sRGB gamut mapping globally for it seems every UI element without exception(?). Windows (but also Linux Xorg) is just excessively stupid in that regard. Hope it gets better with Wayland on Linux, the corresponding protocol has been in the making for ages...

You cannot. That debug setting does not work in many cases.

Balling · 15th February 2022, 16:39

Quote:

Originally Posted by aufkrawall

I think you can't on Windows, afaik neither the Netflix/Disney+ etc. store apps make use of the installed ICC profile, or any browser when it comes to video or even Widevine decoding. So such content on a wide gamut display will have goofy LSD-like colors, at least with Rec709 content. The TV would have to do the color conversion, but then you probably can just use the integrated apps of a SmartTV instead of a HTPC anyway. Well, and with a regular monitor, you're screwed. It's really pathetic.

Though I gotta say I really also dislike brightness locks or other artificial limitations.

ICC profile is not to be used when you output to hdr screen using hdr swapchain.

speedy · 11th March 2022, 22:55

Has anyone figured out what type of performance hit there is from running dwm_lut? Disabling/closing and then re-opening it doesn't seem to yield a noticeable difference in the benchmarks I've ran.

aufkrawall · 5th April 2022, 15:13

We talked about colorimeters a few months ago in the madVR thread. In the meantime, I temporarily got hands on an X-Rite i1 Pro and Spyder X in addition to my old Spyder 4. Observations:
-the X-Rite is basically unusable with generic spectral correction profiles, white is off and relatively low gamut coverage scores with all displays tested with all generic correction profiles
-with a spectral correction profile specifically targeted for a given device, much better results can be achieved. But they apparently need to fit extra well too, else white can still be a bit off
-so I think you need to be extra lucky with correction profiles with the i1 Pro for your given display, else you can buy a spectrometer too
-the SpyderX is much more forgivable with its generic LCD white LED correction (as is the Spyder 4, if not even more)
-you need to update ArgyllCMS to a newer version than shipped by DisplayCal to make the SpyderX work correctly on Windows (including unsigned driver)
-it works out of the box on Linux distributions that ship recent Argyll, such as Arch-based ones (Manjaro live boot environment does the trick)
-both the X-Rite and the SpyderX fail when it comes to correctly measuring gamma of the two VA displays tested, it is measured ~0.2 too bright
-the old Spyder 4 can measure it correctly (i.e. 2.2 for both displays' default setting)
-the old Spyder 4 seems to be less tolerant toward the green part of white, but target of 6500k determined by it looks well to me with all displays tested. Though in reality, it might be a bit colder than that number.
-despite of the Spyder 4 failing miserably in detecting both white and black brightness, this really seems to be of little to no consequence for the actual calibration result in comparison to the newer colorimeters with typical gamma 2.2/2.4 curves
-for whatever reason, the old Spyder 4 scores by far the highest DCI-P3 scores when comparing with generic spectral correction profiles

Note that this is seen through consumer goggles, I don't care for lowest deviations that meet cleanroom criteria.

huhn · 6th April 2022, 08:30

you just compared two meter with each other. there is nothing you can compare it to without a spectrometer.

this is just random number you have here without a reverence you have nothing to compare it too.

these meter have been more than once been compared to spectrometer.

aufkrawall · 6th April 2022, 15:26

You missed the point. Only few enthusiasts buy a spectrometer with their colorimeter. That's why there are generic correction profiles and the online database for specific correction profiles with a particular colorimeter and display. When whitepoint is off in an obvious way, that's a sure sign that the spectral correction profile isn't close to optimal at all. My post was about how usable the different colorimeters proved to be without (!) spectrometer. No need for comments like "But spectrometer!!" or "random numbers" (typical huhn hyperbole). Of course you can always fund me a spectrometer...

huhn · 6th April 2022, 20:55

but you don't have a clue how white looks like without a Spectro meter.

Quote:

-both the X-Rite and the SpyderX fail when it comes to correctly measuring gamma of the two VA displays tested, it is measured ~0.2 too bright
-the old Spyder 4 can measure it correctly (i.e. 2.2 for both displays' default setting)

why even calibrate if the display is already better then your meter wow...

and you wouldn't believe how little it matters if the white point is off because your eye is made to correct it the important part part is that the error is steady and doesn't change over the gamma curve.

d65 is kinda red. you just go and say the white point of a spyder on my display with the correction(without even stating them) i have here looks more white to me tested on two VA panels.
if you think the white point is better ( by eye) just read the calibrated white point with a i1 d3 and use the read white point you should end up with a very similar one.

BTW. the i1 pro(i1pro 2/3) is a spectrometer the usually colorimeter is the i1d3(tons of names i1 display or pro colormunki for example).

if you want to know why the i1d3 is considered to be a much better meter do some close to black reading on i don't know a cheap oled mobile phone or something.

Damien147 · 16th April 2022, 15:58

I think I know the answers but just to confirm.

Can you calibrate HDR?

My colorimeter(colormunki display) is 1000 nits.If my display is higher than 1000 nits I need a new colorimeter?

huhn · 16th April 2022, 18:50

i never tried such a think but i guess so.

aufkrawall · 23rd April 2022, 00:21

Quote:

Originally Posted by huhn

but you don't have a clue how white looks like without a Spectro meter.

Both 6500K and 6800K are "white", to throw some random numbers. It is impossible to guess what you actually intend to say, so I'll leave it there.

Quote:

Originally Posted by huhn

why even calibrate if the display is already better then your meter wow...

Of course the measurements are very likely wrong in this case, the chances that both office PC displays are gamma 2.4 with their default gamma setting are unrealistic.

Quote:

Originally Posted by huhn

d65 is kinda red.

It is not.

huhn · 23rd April 2022, 09:31

you have 2 modern meters at and and they agree and you take the results of an old known to drift meter and say the other too are wrong seriously...

aufkrawall · 23rd April 2022, 15:59

Seriously, if you can't distinguish gamma 2.2 from 2.4, please don't reply to my posts.

huhn · 23rd April 2022, 20:38

smart...
so can i distinguish gamma 2.4 and 2.2. yes i can... i can make you even images of that but it has nothing todo with this.
but your absolute disregarded of calibration and understanding of it the issue here is now i can't see the difference between gammas because you have to point your finger at something when you have zero arguments.

you have 3 meters one telling it is gamma 2.2 and two tell you it is 2.4 and now you talk about seeing the difference i can't see your stuff and even if i do.
you don't have a clue which is correct if you just take the data
just read your own stuff:

Quote:

-despite of the Spyder 4 failing miserably in detecting both white and black brightness, this really seems to be of little to no consequence for the actual calibration result in comparison to the newer colorimeters with typical gamma 2.2/2.4 curves

do you even know how gamma is measured at all?
by reading brightness which you confirm (based on what is yet another mystery) to be bad.

by using your magical gamma seeing eyes.

wow smart but can i see the difference between 2.2 and 2.4 calibrated from meter i don't have here on displays i don't have here?
no i obviously can't.

you are claiming a spider x and a i1d3 can't do a gamma ramp at all because the objectively worse meter looks better to you in a calibration thread...

aufkrawall · 23rd April 2022, 21:52

Quote:

Originally Posted by huhn

you are claiming a spider x and a i1d3 can't do a gamma ramp at all because the objectively worse meter looks better to you in a calibration thread...

No, I said this is what happens exclusively with those particular two VA panel monitors. You read what your confirmation bias wants you to read. The old Spyder 4 correctly measures gamma as 2.2 with OLED devices (no difference in this case vs. the two newer colorimeters), so don't make up even more bullsh*t you want to be true to fit your point.

huhn · 23rd April 2022, 22:45

getting a wrong gamma means your meter reads brightness wrong depending on the level.

do you know what happens when your black/white is read to high/low? this changes your effective measured gamma. you sad your spyder reads brightness for black and white wrong this literally means it is reading the gamma for these wrong. how you confirmed that i don't know. why would it do a better job at the other brightness level then the other meter sorry that's hard to understand.
yes the spyder x and i1d3 could be wrong here i have no clue how old they are but that's not my point my point is you can not properly proof that like this.
you could also show the gamma reading?

btw. feel free to read how long the life expectancy of a spyder 4 is about 3 years that's it.
i have to sadly replace my i1d3 because it could have drifted too. some people will tell you they are not like old meter that life about 3 years which is true.. that doesn't mean they will live forever or even close to.

i was not in the boat of the spyder is useless because the spyder x is a fine meter.

there are also test done compared to spectrometer:
https://www.chromapure.com/newgear_i...20Accuracy.asp
https://www.chromapure.com/newgear-new.asp

aufkrawall · 23rd April 2022, 23:08

Quote:

Originally Posted by huhn

getting a wrong gamma means your meter reads brightness wrong depending on the level.

do you know what happens when your black/white is read to high/low? this changes your effective measured gamma.

Doesn't make sense. The x-rite measured contrast of 1100:1 for my IPS device and reports gamma 2.2. The Spyder 4 measures 770:1 and reports gamma 2.2 too...

I suspect you are greatly exaggerating brightness measurements as a factor for pure power gamma 2.2/2.4 curves. Even BT.1886 still is similarly bright for dark shades with the x-rite/spyderx compared to spyder 4 and I dare to say hardly anybody would notice the difference in a double blind test...

huhn · 24th April 2022, 01:14

i don't have your number.
black offset is unlikely to change it but can. it highly depends on the formula and how cleaver it ignores close to black numbers or if you did it yourself.
if you try to measure stimilus 10 % but your device can't read the difference between it and stimulus 5% you would get odd gamma values.
stimulus 10% for gamma 2.2 is 0.63 % for 2.4 it is 0.40 %.
depending on how deep your screen can go it could produce values the spyder 4 can't read and it will see a very low gamma.

white will be drastic.
for 90% stimulus the difference between gamma 2.2 and gamma 2.4 is ~1.6 % brightness(77.66% and 79.31). just assume what 5 or 10 % would do to the calculated gamma.
if the device is actually 100 nits but it reports 110 nits i leave the rest up to your math.
without accurate white you can't even calculate correct gamma numbers it's pretty much end here. even if all other stimulus are read correctly the calculated gamma is just plain wrong.
but i don't know how off your spyder reported white is by 5% 20% no clue. 1% shouldn't matter to do a gamma difference of 0.2 but i didn't calculated it.

but i will do the math for 100 nits 2.4 read as 2.2 it is:
97.9 nits that's not much...
for stimulus 90% an error of 2.1 nits stimulus 100% is a gamma difference of 0.2... assuming stimulus 90% is read correctly.

in the end you just had to compare the stimulus that it can read accurately to see where they disagree and where they agree.

and if you think i'm exaggerating the important of brightness gamma is calculated from the peak white. looks like i was heavily underestimating it feel free to check the math it's not like it do this every day.

11th December 2021, 00:29	#441 \| Link
nevcairiel Registered Developer Join Date: Mar 2010 Location: Hamburg/Germany Posts: 10,348	Its quite possible that D3D9 in madVR is a factor. D3D9 is quite limited as it comes to resource management, and the driver/API certainly plays a role there. There aren't enough caches to cache the entire 3DLUT in any case, or a full video texture for that matter, so it has to do a reasonable job at accessing it from video memory directly, or nothing would work. Like I hinted at above, texture sampling should be a bread and butter operation for any GPU, even if 3D textures are a bit rarer. __________________ LAV Filters - open source ffmpeg based media splitter and decoders

5th April 2022, 15:13	#447 \| Link
aufkrawall Registered User Join Date: Dec 2011 Posts: 1,812	We talked about colorimeters a few months ago in the madVR thread. In the meantime, I temporarily got hands on an X-Rite i1 Pro and Spyder X in addition to my old Spyder 4. Observations: -the X-Rite is basically unusable with generic spectral correction profiles, white is off and relatively low gamut coverage scores with all displays tested with all generic correction profiles -with a spectral correction profile specifically targeted for a given device, much better results can be achieved. But they apparently need to fit extra well too, else white can still be a bit off -so I think you need to be extra lucky with correction profiles with the i1 Pro for your given display, else you can buy a spectrometer too -the SpyderX is much more forgivable with its generic LCD white LED correction (as is the Spyder 4, if not even more) -you need to update ArgyllCMS to a newer version than shipped by DisplayCal to make the SpyderX work correctly on Windows (including unsigned driver) -it works out of the box on Linux distributions that ship recent Argyll, such as Arch-based ones (Manjaro live boot environment does the trick) -both the X-Rite and the SpyderX fail when it comes to correctly measuring gamma of the two VA displays tested, it is measured ~0.2 too bright -the old Spyder 4 can measure it correctly (i.e. 2.2 for both displays' default setting) -the old Spyder 4 seems to be less tolerant toward the green part of white, but target of 6500k determined by it looks well to me with all displays tested. Though in reality, it might be a bit colder than that number. -despite of the Spyder 4 failing miserably in detecting both white and black brightness, this really seems to be of little to no consequence for the actual calibration result in comparison to the newer colorimeters with typical gamma 2.2/2.4 curves -for whatever reason, the old Spyder 4 scores by far the highest DCI-P3 scores when comparing with generic spectral correction profiles Note that this is seen through consumer goggles, I don't care for lowest deviations that meet cleanroom criteria. Last edited by aufkrawall; 5th April 2022 at 15:28.

11th December 2021, 01:18	#442 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	the vulkan numbers are very convincing if that's even the reason this happens. this alone is a good reason to move on from dx9 the 19 years of age are also a good reason.

11th December 2021, 15:34	#443 \| Link
aufkrawall Registered User Join Date: Dec 2011 Posts: 1,812	I guess the 3D LUT format for madVR can be described as "maxed out". When you generate in IRIDAS format, it is already just a fraction in size. But probably even ~600KB ReShade format would look fine when properly applied in fp16 precision + dithering.

11th March 2022, 22:55	#446 \| Link
speedy Registered User Join Date: Nov 2019 Posts: 31	Has anyone figured out what type of performance hit there is from running dwm_lut? Disabling/closing and then re-opening it doesn't seem to yield a noticeable difference in the benchmarks I've ran.

6th April 2022, 08:30	#448 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	you just compared two meter with each other. there is nothing you can compare it to without a spectrometer. this is just random number you have here without a reverence you have nothing to compare it too. these meter have been more than once been compared to spectrometer.

6th April 2022, 15:26	#449 \| Link
aufkrawall Registered User Join Date: Dec 2011 Posts: 1,812	You missed the point. Only few enthusiasts buy a spectrometer with their colorimeter. That's why there are generic correction profiles and the online database for specific correction profiles with a particular colorimeter and display. When whitepoint is off in an obvious way, that's a sure sign that the spectral correction profile isn't close to optimal at all. My post was about how usable the different colorimeters proved to be without (!) spectrometer. No need for comments like "But spectrometer!!" or "random numbers" (typical huhn hyperbole). Of course you can always fund me a spectrometer...

16th April 2022, 15:58	#451 \| Link
Damien147 Registered User Join Date: Mar 2011 Posts: 380	I think I know the answers but just to confirm. Can you calibrate HDR? My colorimeter(colormunki display) is 1000 nits.If my display is higher than 1000 nits I need a new colorimeter?

16th April 2022, 18:50	#452 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	i never tried such a think but i guess so.

23rd April 2022, 09:31	#454 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	you have 2 modern meters at and and they agree and you take the results of an old known to drift meter and say the other too are wrong seriously...

23rd April 2022, 15:59	#455 \| Link
aufkrawall Registered User Join Date: Dec 2011 Posts: 1,812	Seriously, if you can't distinguish gamma 2.2 from 2.4, please don't reply to my posts.

23rd April 2022, 22:45	#458 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	getting a wrong gamma means your meter reads brightness wrong depending on the level. do you know what happens when your black/white is read to high/low? this changes your effective measured gamma. you sad your spyder reads brightness for black and white wrong this literally means it is reading the gamma for these wrong. how you confirmed that i don't know. why would it do a better job at the other brightness level then the other meter sorry that's hard to understand. yes the spyder x and i1d3 could be wrong here i have no clue how old they are but that's not my point my point is you can not properly proof that like this. you could also show the gamma reading? btw. feel free to read how long the life expectancy of a spyder 4 is about 3 years that's it. i have to sadly replace my i1d3 because it could have drifted too. some people will tell you they are not like old meter that life about 3 years which is true.. that doesn't mean they will live forever or even close to. i was not in the boat of the spyder is useless because the spyder x is a fine meter. there are also test done compared to spectrometer: https://www.chromapure.com/newgear_i...20Accuracy.asp https://www.chromapure.com/newgear-new.asp

24th April 2022, 01:14	#460 \| Link
huhn Registered User Join Date: Oct 2012 Posts: 7,926	i don't have your number. black offset is unlikely to change it but can. it highly depends on the formula and how cleaver it ignores close to black numbers or if you did it yourself. if you try to measure stimilus 10 % but your device can't read the difference between it and stimulus 5% you would get odd gamma values. stimulus 10% for gamma 2.2 is 0.63 % for 2.4 it is 0.40 %. depending on how deep your screen can go it could produce values the spyder 4 can't read and it will see a very low gamma. white will be drastic. for 90% stimulus the difference between gamma 2.2 and gamma 2.4 is ~1.6 % brightness(77.66% and 79.31). just assume what 5 or 10 % would do to the calculated gamma. if the device is actually 100 nits but it reports 110 nits i leave the rest up to your math. without accurate white you can't even calculate correct gamma numbers it's pretty much end here. even if all other stimulus are read correctly the calculated gamma is just plain wrong. but i don't know how off your spyder reported white is by 5% 20% no clue. 1% shouldn't matter to do a gamma difference of 0.2 but i didn't calculated it. but i will do the math for 100 nits 2.4 read as 2.2 it is: 97.9 nits that's not much... for stimulus 90% an error of 2.1 nits stimulus 100% is a gamma difference of 0.2... assuming stimulus 90% is read correctly. in the end you just had to compare the stimulus that it can read accurately to see where they disagree and where they agree. and if you think i'm exaggerating the important of brightness gamma is calculated from the peak white. looks like i was heavily underestimating it feel free to check the math it's not like it do this every day.