- This topic has 17 replies, 3 voices, and was last updated 4 years, 5 months ago by
.
-
Posts
-
I’m having some problems calibrating my PA329Q Asus monitor. I’m using Nvidia’s new Studio Drivers which allow 10bit and Windows 10 x64 with a GTX 1070 graphics card. I also am using a Spyder5 Pro. Part of what I’m seeing is that I calibrate the RGB to my white point (6500K) and get it to below 0.5, but the next day it is off by up to 1 DE. I wasn’t sure if this is the monitor or the calibrator showing signs of age? I’ve had this Spyder5 Pro for more than a few years now so I wasn’t sure, and don’t have another one to use to compare.
Thanks.
nVidia have a levels bug in their newest drivers (436.02 and newer, last working driver was 431.60), so the first thing you should do is check if you are affected (using attached test picture, you should see two bars on each side). If you are, go into nVidia control panel, desktop color settings, if you have an option “content type” set it to any other setting than the currently selected, hit apply, then cancel. Otherwise, the only other workaround is a reboot.
Wow… after doing all those calibrations trying to find the best results and the drivers were bugged. pfffft…
Well that does explain why DCal would often say that the monitor was 8bit even though I let DCal handle the colour management. Sometimes it would also say that it couldn’t tell what the bit depth was, and the oddest log told me I had a “9bit display” wtf? lol
So, before I could see 2 bars on each side. I did a full uninstall of the old Studio Drivers and rebooted and did a clean install of 431.60 Studio Drivers. I can still see the bars, but they are quite a bit harder to see. I’m about to try quick report to see if it reads it as 10bit now.
Oh, BTW, do I set the “who handles colour controls, Nvidia or ‘other application’ to Nvidia to enable the option to pick 10bit output? What about the section for what handles video colours? It looks like the default of letting the player handle the colours limits the colour range. I have to let Nvidia handle video colour if I want to use full range. At least according to Nvidia’s control panel.
Also, is there a way to make a full range MadVR 3dLUT? I will probably have more questions, and doing all this has made me think of a couple ideas that I think could be a big help for people trying to get the best picture they can get from the monitor they have.
But it’s almost 3am… thanks for the help. I probably never would have figured that out.
And… good….. night! lol
Edit: Hrmm… still saying its 8bit……. It’s a PA329Q. If it doesn’t have a 10bit display, heads will roll…. rofl
Well that does explain why DCal would often say that the monitor was 8bit even though I let DCal handle the colour management.
Sometimes it would also say that it couldn’t tell what the bit depth was, and the oddest log told me I had a “9bit display” wtf? lolI doubt that DisplayCAL would say something about the bits in your monitor, or link betawen GPU and display… What displayCAL does (actually ArgyllCMS) is to test VIDEO LUT (= GPU calibration tables).
8bit => truncation in GPU tables, which could lead to banding
X bit with X>8 => ArgyllCMS test (noise in LUT I suppose) went OK, measurements show some dithering or high bit depth tables being used.
for “unable to find effecitve video LUT” (of something like that) ask in ArgyllCMS mailist but I saw it in AMD using dithering in LUT, 10bit-DP link (not OpenGL 10bit draw, just link) while rendering smooth bandless gradients like lagom 8bit PNG.If ArgyllCMS reports “effective video lut entry 8bits” it is likely that your system will have banding if you load a calibration in GPU. It’s a long issue with nvidias but the proof is an 8bit smooth gradient viewed in a non color managed enviroment, like lagom PNG in MS Paint. If it renders smooth while you have GPU calibration applied, all is fine, a false positive from ArgyllCMS LUT test.
Studio driver as you may have read only exposes OpenGL functionality to draw 10bit stuff while in Windows desktop mode (Photoshop). Before that driver they could do 10bit fullscreen… which seems to be what you wanted.
MadVR uses dithering too, to it can render 16bit gradients smooth in a 8bit link so all that “10bit stuff” is not as useful as a lot of people think. Also AFAIK madVR does not use OpenGL 10bit drawing on desktop, so Studio drivers offer no advantage for that application.Also, is there a way to make a full range MadVR 3dLUT?
You can do it with ArgyllCMS using commandline but it would be no use. MadVR 3dLUT expects video range LUT, so DisplayCAL makes them the way they need to be.
Output levels has nothing to do with 8/10/12 bit. Wrong output levels will lead to clipping, irrespective of 8/10/12 bit, thus the test image. If you can see all four bars, levels are probably ok. If you cannot see the darkest bar, you’re sending full range 0-255 to a monitor that expects video 16-235 levels (or the monitor has been setup wrong, e.g. unsuitable contrast or digital black level setting).
Regarding higher than 8 bit output:
Sometimes it would also say that it couldn’t tell what the bit depth was, and the oddest log told me I had a “9bit display”
A measurement device will have a hard time distinguishing between 9, 10 or even 12 bits – the display + instrument noise are already higher than the instrument sensitivity. It is not a problem.
Oh, BTW, do I set the “who handles colour controls, Nvidia or ‘other application’ to Nvidia to enable the option to pick 10bit output?
In nV CPL, under “Change esolution”, set “Use nVidia color settings”. Under “Adjust desktop color settings”, select “Other applications control color settings”. Do NOT use optimizations (leave digital vibrance at default 50% and hue at 0%).
What about the section for what handles video colours?
Ignore (leave at “With Video Player”).
Also, is there a way to make a full range MadVR 3dLUT?
No. Full range madVR 3D LUT would be inherently wrong, always.
Edit: Hrmm… still saying its 8bit……. It’s a PA329Q
Higher than 8 bit output will be lost after resume from standby. A reboot fixes that. This is a Windows 10 and/or nVIdia driver bug that has existed forever (Windows 7 was fine…).
Btw, if the “report on uncalibrated display” test shows “couldn’t determine videoLUT bitdepth”, that is usually a sure-fire way of knowing that higher than 8 bit output is currently working 🙂
Studio driver as you may have read only exposes OpenGL functionality to draw 10bit stuff while in Windows desktop mode (Photoshop). Before that driver they could do 10bit fullscreen… which seems to be what you wanted.
10/12 bit windowed and fullscreen still works the same (for the few applications that use it, i.e. madVR) in this driver. The openGL 10-bit is the new part (since 436.02, but that driver introduced the levels bug I mentioned).
Studio driver as you may have read only exposes OpenGL functionality to draw 10bit stuff while in Windows desktop mode (Photoshop). Before that driver they could do 10bit fullscreen… which seems to be what you wanted.
10/12 bit windowed and fullscreen still works the same (for the few applications that use it, i.e. madVR) in this driver. The openGL 10-bit is the new part (since 436.02, but that driver introduced the levels bug I mentioned).
Great!
But there is a way to test it with Rec709 SDR? I mean madVR does dithering, you can open a 16bit TIFF gradients in some video players and render it in madVR.
Windowed and 8bit desktop (common AMD) shows smooth color managed image (by LUT3D). Full screen 10bit link with the same GPU shows the same smooth color managed gradient. I cannot distinguish visually.
I mean… is it useful? In Photoshop it is useful because it does not dither when doing color management, hence 10bit outout is an improvement. But Lightroom and Capture One do dither after color management stage before sending to GPU, hence they do not need it.
Same would apply to madVR (SDR content).Hi, thanks for the info.
I uninstalled the driver I was using with IOBit Uninstaller and rebooted with all Nvidia graphics software removed, and then did a clean install of the 430.86 Studio Drivers. Since the number is lower then the number you said had the bug, and these are still “Studio Drivers” I should theoretically be able to get 10bit. My Nvidia control panel was already set up as you described Florian. The 1 option that I always wondered about, and maybe this just isn’t your area, but, I never know if I should have “Scaling” done by the monitor or the graphics card? I have a GTX 1070.
I did one of the short reports on the display while calibrated and it didn’t say what it thought the bit depth was. I then did one for uncalibrated and it said this,
15:00:41,148 Setting up the instrument
15:00:41,148 Instrument Type: Datacolor Spyder5
15:00:41,148 Serial Number: 50123929
15:00:41,148 Hardware version: 0x0a0f
15:00:41,148 Uncalibrated response:
15:00:41,148 Black level = 0.2328 cd/m^2
15:00:41,148 50% level = 40.50 cd/m^2
15:00:41,148 White level = 186.34 cd/m^2
15:00:41,148 Aprox. gamma = 2.20
15:00:41,148 Contrast ratio = 800:1
15:00:41,148 White chromaticity coordinates 0.3141, 0.3304
15:00:41,148 White Correlated Color Temperature = 6416K, DE 2K to locus = 4.6
15:00:41,148 White Correlated Daylight Temperature = 6419K, DE 2K to locus = 0.1
15:00:41,148 White Visual Color Temperature = 6257K, DE 2K to locus = 4.4
15:00:41,148 White Visual Daylight Temperature = 6421K, DE 2K to locus = 0.1
15:00:41,148 Effective Video LUT entry depth seems to be 8 bits
15:00:41,148 Black drift was 0.000000 DE
15:00:41,148 White drift was 0.000000 DE’As you can see, DCal thinks the panel is 8bit. I don’t have time to try the “smooth rainbow” test mentioned above but I will later.
Also, I’ve had this Spyder5 Pro for a fair number of years. How long (a guess is fine) do Spyder 5s last, and if I don’t have a 2nd hardware profiler, is there any way to tell if it is degrading unless its malfunction makes it give wildly varying readings which are different each time? If it’s always “off” by 2 DeltaE, a verification report is going to give me 2 thumbs up! lol
Edit: Was about to post when I remembered that I have the box for this calibrator on a shelf. I looked on the back and it says 2016, and it does list Windows 10 as an OS, so it actually isn’t as old as I thought. (Wish I could say the same thing about me. hahaha… *sigh*)
Also, I’ve tried Asus ProArt Calibration software, and then made sure the graphics card LUT was reset and the loader disabled, because this monitor actually stores the information in a chip IN the monitor. This would all be an amazing thing, especially since you can use it to improve the uniformity, except that DCal’s verification reports always show DeltaE numbers that are way above what DCal gives me when I do a profile with your program. Since a verification report is just sending colours to the monitor and the calibrator records what it actually sees, the verification function should give an accurate reading of how good the Asus Calibration is shouldn’t it? The part that makes it a shame is that it seems, for the moment at least, you can’t JUST improve the uniformity. You have to also let the program calibrate the colours. :/
I find it hilariously not funny that ArgyllCMS and DisplayCal seems to be a far FAR better calibration package then one put out by a (probably) multi-hundred billion dollar a year multi-national corporation. lol
It’s to bad that there probably is no way to tap into the monitors screen uniformity even if you wanted to add such a function due to the nature of trying to reverse engineer something like that. 🙁
Hmm looks like I can’t edit twice. I wanted to know why a full range MadVR 3DLUT would be “wrong?” I use SVP 4 which uses MadVR, but I often watch the shows on my computer which can output full range. If we are talking about a movie, or maybe an anime, which doesn’t really conform well to the colour standards that TV was designed for since it is an animation. I’m not saying an animated program shown on a TV is “out of spec” for Rec709 or Rec2020, only that a full range colour probably wouldn’t hurt it…… would it? lol
Higher than 8 bit output will be lost after resume from standby. A reboot fixes that. This is a Windows 10 and/or nVIdia driver bug that has existed forever (Windows 7 was fine…).
15:00:41,148 Effective Video LUT entry depth seems to be 8 bits
[…]As you can see, DCal thinks the panel is 8bit. I don’t have time to try the “smooth rainbow” test mentioned above but I will later.
DisplayCAL does not say that, read again what DisplayCAL says. It’s the GPU LUT… DisplayCAL says nothing about panel & its bits. We’ve explanied this before.
Hmm looks like I can’t edit twice. I wanted to know why a full range MadVR 3DLUT would be “wrong?” I use SVP 4 which uses MadVR, but I often watch the shows on my computer which can output full range. If we are talking about a movie, or maybe an anime, which doesn’t really conform well to the colour standards that TV was designed for since it is an animation. I’m not saying an animated program shown on a TV is “out of spec” for Rec709 or Rec2020, only that a full range colour probably wouldn’t hurt it…… would it? lol
You misunderstand “GPU to display” configuration be it limited or full, with madVR input data. Second one needs limited range. After MadVR has done his stuff it will scale to whatever configuration you have
I did one of the short reports on the display while calibrated and it didn’t say what it thought the bit depth was.
VideoLUT bit depth can only be determined when uncalibrated.
As you can see, DCal thinks the panel is 8bit. I don’t have time to try the “smooth rainbow” test mentioned above but I will later.
Not panel, VideoLUT.
How long (a guess is fine) do Spyder 5s last
If you take a dremel to it, not very long 😉
is there any way to tell if it is degrading
There is not much point in telling if it’s degrading if the performance is sub-par to begin with 😉 (I think you can tell by now I don’t have very high opinion on Spyder devices up to 5, I just normally don’t go out of my way to say they suck)
I looked on the back and it says 2016, and it does list Windows 10 as an OS, so it actually isn’t as old as I thought. (Wish I could say the same thing about me. hahaha… *sigh*)
Oh come on now. I was born ’76. Not old, just (somewhat) experienced 🙂
Hmm looks like I can’t edit twice
Intended, to avoid edits being applied too liberally. The timeframe for edits from post creation is 5 minutes or so.
I wanted to know why a full range MadVR 3DLUT would be “wrong?”
Because it has all to do with video encoding range and madVR’s internal processing, and nothing with output range (which is separate from encoding range).
Ya I ignored the SpyderX when I heard that it was faster, but not necessarily more accurate. If I got a better one, is there a good newer one that you would recommend that would completely kill my bank account? 😉 Or would you wait because you know more “new tech” calibrators are coming down the line sometime soon? I figure you have your ear to the ground on that sort of thing way more than me.
Oh come on now. I was born ’76. Not old, just (somewhat) experienced
hah that’s actually funny because so was I. I remember when I used to think 40s was old. Then I moved it to 50, then 60. At this point I just hope I age gracefully and hopefully, health permitting, I’m only at about the 50% point of my life, and not 70%. 🙂
Thanks for all the info.
DM
SpyderX Pro on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.X-Rite has announced their newest re-branding of the i1D3 line a while ago. If you want to be somewhat more future-proof for HDR, you could get the i1Display Pro Plus (up to 2000 cd/m² compared to the i1Display Pro’s 1000 cd/m²), but whether that device works with the current version of ArgyllCMS is somewhat unclear – it wouldn’t be a big deal to get it supported though. Otherwise, the i1Display Studio is the same as ColorMunki Display, and is confirmed to work.
hah that’s actually funny because so was I. I remember when I used to think 40s was old. Then I moved it to 50, then 60. At this point I just hope I age gracefully and hopefully, health permitting, I’m only at about the 50% point of my life, and not 70%.
That seems like a reasonable attitude to have 🙂 Twenty years ago, I couldn’t even imagine what it would be like when I was 40. In retrospect, not that much different – slightly more wrinkly, a lot less hair and a lot more (easily) tired mainly, haha. Perspective on life has changed a bit, realizing that the time that already passed grows continually larger than the amount that might be left – but here’s to us having a blast as the journey goes on 🙂
Thanks for the suggestions! And ya, my thought processes now are so different to what they were when I thought I was an adult, at, 14, 16, 18, 21, etc etc. I finally realized the real truth. The is no finish line to “being an adult.” You grow and change all through your life. The only real difference is that getting older usually isn’t as fun as being younger, and I don’t know about you, but I find that as I get older it seems like time just moves faster and faster and faster. I figure by the time I’m 60 I’ll blink and I’ll be in a nursing home. lol
BTW you’ve been a big help over the past couple years. Even though I think I’ve probably read your entire, VERY VERY long and in depth readme. lol
Do you know if the 1000cd/m2 version is less accurate or a lot slower then the newest one? And also, I was planning to make a bit of a donation as a thank-you for all your help. I mean, your software is better than Asus’s… My lordy…. lol But I wondered if you also had a link to those 2 calibrators for someplace that I could buy from (I’m in Canada) and you would get a percentage?
Thanks! BTW I did see some things that may help you understand my issues in the very detailed log that shows a lot more info. Can I attach one later? But for now, I noticed 1 thing that I wasn’t sure what it meant. It was looking at each R.G.B. channel and it shows like, R++ G- B-. Does that mean (as I already suspect) that red is over-saturated and being clipped, and that green and blue are not reaching full saturation?
Thank-you!
DM
