madVR - General discussion thread [Archive] - Page 2

blu3wh0

3rd October 2018, 16:41

The downside of LG's PC mode is, that it only works in 60hz mode. If you switch to 23/24 Hz the full 4:4:4 support drops... so you can not play movies in 24p with full 4:4:4. You would have to use 60hz mode an maybe smooth motion to get the movie judder free with full 4:4:4 support. Thats the only thing left that is still annoying. Anything else is running good an I am really satisfied with my setup.

My latest and working setup:

Windows 10 1809
Kodi DS Player 17.6
AMD RX480, 8GB with driver 18.9.3
madvr 0.92.17
lav filters 0.72.0-15
xysubfilter 3.1.0.751

LG 65" C8 OLED (Frimware 04.10.05) / Black Level: high
AMD settings: RGB Full, 10 bit
madvr: 0-255

Have you actually tested this? When I tested 4K 4:4:4 PC mode on my TV, it displayed 4:4:4 properly on 23Hz, result was the same as 60Hz. I tested using "the quick brown fox" picture that is usually recommended, but you should be able to use any 4:4:4 test. Also, you know when it's properly active if most of the picture options are disabled, it will switch to a different mode on it's own if it doesn't work. FYI, 1080p 4:4:4 23Hz does not work under these conditions. You can also switch back and forth between any other mode and PC mode to easily tell the difference between these tests under the exact same settings.

Edit: I use Nvidia with it's funky 12 bit depth limitation for hdmi, so I have display resolution 4K RGB 4:4:4 12 bits as my setting, not as a TV resolution option that it might be setting for 1080p. I never investigated further since what I needed worked.

Warner306

3rd October 2018, 19:06

Yea I was trying to figure out if my C8 was disabling tone mapping or not if madvr was doing it and it doesn't appear to be the case. The TV is still tone mapping from what I can see. (Murder on the Orient Express was a much better test case for trying to figure that out in my case). What I don't understand is how the dynamic tone mapping on the Panasonic UB820 gets around this issue (if it actually does but in my testing it sure seems to). I tried with the LG's dynamic tone mapping on and off and it didn't seem to make any difference. I would have thought, and madshi agreed with the theory, that turning on dynamic tone mapping in the LG would mostly disable the tone mapping if it is already sent a properly tone mapped image for the max nit level for the display. That didn't appear to be the case, unfortunately.

Can you take some screenshots of the Panasonic Blu-ray player to show what it is and isn't doing vs. madVR. I know the Panasonic preset for an OLED is 1,000 nits. This would mean only titles over 1,000 nits would receive any tone mapping at all. That is about 20% of currently available HDR movies. Are you changing other sliders to alter the 1,000 nits OLED preset?

I question whether the madVR pixel shader is working at all. One person said 100 nits was working fine. In absolute PQ nits, this would make the image almost entirely black. The image should get darker and darker as the target nits is lowered. It isn't possible for the display to make the input brighter because PQ values are absolute in terms of luminance. So 100 nits peak should be shown at 100 nits; 500 nits at 500 nits, etc. Tone mapping might make the image darker (reduce the luminance of some input values to match the brightness capabilities of the display), but the majority of displays will not make the image brighter. I would expect the color would change in some way when you use pixel shader, but changes in brightness should be more apparent with lower target nits. If this isn't happening, I would think pixel shader is not working correctly.

mclingo

3rd October 2018, 19:20

just installing 1809 now, wish me luck !

.....although as i'm typing this i'm thinking... why are you doing this Mclingo...

SamuriHL

3rd October 2018, 23:51

Can you take some screenshots of the Panasonic Blu-ray player to show what it is and isn't doing vs. madVR. I know the Panasonic preset for an OLED is 1,000 nits. This would mean only titles over 1,000 nits would receive any tone mapping at all. That is about 20% of currently available HDR movies. Are you changing other sliders to alter the 1,000 nits OLED preset?

I question whether the madVR pixel shader is working at all. One person said 100 nits was working fine. In absolute PQ nits, this would make the image almost entirely black. The image should get darker and darker as the target nits is lowered. It isn't possible for the display to make the input brighter because PQ values are absolute in terms of luminance. So 100 nits peak should be shown at 100 nits; 500 nits at 500 nits, etc. Tone mapping might make the image darker (reduce the luminance of some input values to match the brightness capabilities of the display), but the majority of displays will not make the image brighter. I would expect the color would change in some way when you use pixel shader, but changes in brightness should be more apparent with lower target nits. If this isn't happening, I would think pixel shader is not working correctly.

I'll probably be this weekend before I can get to that. Yes, I knew they use 1000 nits for the OLED setting. No, I'm not messing with the other sliders. However, the presets for HDR on the panasonic definitely change the picture.

That situation is what led me to the conclusion that the TV's tone mapping is not disabled. If I go down to 100 nits it's the same brightness as 1000. Maybe my conclusion is wrong and that the pixel shading isn't working when HDR mode is enabled. No matter what I change in madvr with pixel shader enabled in HDR mode, the picture does not seem to change so your assertion could be more correct than mine.

SamuriHL

3rd October 2018, 23:52

just installing 1809 now, wish me luck !

.....although as i'm typing this i'm thinking... why are you doing this Mclingo...

I've had it installed for a few days now. It's fine.

mclingo

4th October 2018, 00:34

yeah, its went fairly well all things considered, It did cause the colour saturation issue on mine again which I now cant fix, I can only workaround it. Its really really odd that its only MADVR/SDR/23HZ material, everything else is ok. whats equally or even more odd is that when I calibrate in MADVR for BT2020 to fix this, it doesnt have any knock on effect on the other videos which were playing HDR / SDR @60hz, you;d think these would be changed in some way by MADVRs calibration for BT2020. odd.

This is why I think its a MADVR bug but MADSHI does not so it will probably never get fixed unless more people report the same issue, that doesnt look like its going to happen, either it is my system or there just arent that many people with similar enough setups.

The over saturation can be quite subtle in some movies though, could be some people just arent noticing it.

I could very happily live with the workaround permanently though, it has no other negative impact luckily.

alps006

4th October 2018, 06:59

I asked this same question elsewhere.Is there a difference between "passthrough HDR to dispaly" and "output video in HDR format" options in latest version of madVR? Thanks!

huhn

4th October 2018, 07:03

"output video in HDR format" where is this option?

alps006

4th October 2018, 07:24

It is under "devices" "hdr". I am using v0.92.17 with a recent test build."output video in HDR format" where is this option?

huhn

4th October 2018, 07:34

the option under "tone map using pixel shaders" is most likely a renamed from send HDR meta data to this because it make more sense.

the difference is that madVR will alter the image before sending it to the display but with that option it is still "HDR" type not SDR like with passthrough.

alps006

4th October 2018, 08:19

Thanks, huhn. That makes sense. I have great results enabling that option. Good natural colors without oversaturation. the option under "tone map using pixel shaders" is most likely a renamed from send HDR meta data to this because it make more sense.

the difference is that madVR will alter the image before sending it to the display but with that option it is still "HDR" type not SDR like with passthrough.

mclingo

4th October 2018, 14:13

off topic WINDOWS 1809 issue - you cant make the task bar black that I can find, you can get it very dark grey but not black, this unfortunately means that even if you hide the task bar to stop burn in on OLEDS there is still a line of grey pixels along the bottom of the screen which are there all the time, this is bad for OLEDS.

The only fix i've found so far is to install CLASSIC SHELL where you can set the task bar as fully black, not ideal.

Warner306

4th October 2018, 16:12

This is why I think its a MADVR bug but MADSHI does not so it will probably never get fixed unless more people report the same issue, that doesnt look like its going to happen, either it is my system or there just arent that many people with similar enough setups.

The over saturation can be quite subtle in some movies though, could be some people just arent noticing it.

I could very happily live with the workaround permanently though, it has no other negative impact luckily.

Did you ever determine if your display is calibrated for BT.709 in its color settings? If you select BT.2020, madVR would send BT.2020 to the display. You can confirm this in the OSD. The video driver would have to be lucky to accidentally convert BT.2020 to BT.709 and send it to the display without corrupting the image.

Warner306

4th October 2018, 16:14

I'll probably be this weekend before I can get to that. Yes, I knew they use 1000 nits for the OLED setting. No, I'm not messing with the other sliders. However, the presets for HDR on the panasonic definitely change the picture.

That situation is what led me to the conclusion that the TV's tone mapping is not disabled. If I go down to 100 nits it's the same brightness as 1000. Maybe my conclusion is wrong and that the pixel shading isn't working when HDR mode is enabled. No matter what I change in madvr with pixel shader enabled in HDR mode, the picture does not seem to change so your assertion could be more correct than mine.

It is more than a problem of the display disabling its tone mapping if 100 nits from madVR is the same brightness as 10,000 nits. When you select 100 nits, the entire range is supposed to be compressed to 100 nits. This would make the image very dark. The display can't reverse engineer these altered RGB values to restore the original source brightness.

mclingo

4th October 2018, 16:48

Did you ever determine if your display is calibrated for BT.709 in its color settings? If you select BT.2020, madVR would send BT.2020 to the display. You can confirm this in the OSD. The video driver would have to be lucky to accidentally convert BT.2020 to BT.709 and send it to the display without corrupting the image.

Hi, unfortunately my TV doesnt have any such settings, it set as stock with minor changes to low IRE colour settings only, I used to have a Sony LCD which did so I know what you mean.

However, this also happens with my Samsung TV so i've pretty much ruled Mt TV and AVR out of the equation.

I'm probs going to have to re-image my HTPC but I'm really reluctant to do that as its a 3 day build at least as its also a server with 12 disks in it - but also, if I dont know what is causing it, if it comes back I'll be back to square one and 3 days down.

I was hoping MADSHI had an AMD contact but because there are only a couple of people with this problem, and thats its defeatable, he's not really got time to look at it, which is fair enough.

I was just hoping that with the way it manifested itself it would be obvious to someone what was going on.

Thanks for your interest though :)

SamuriHL

4th October 2018, 16:55

It is more than a problem of the display disabling its tone mapping if 100 nits from madVR is the same brightness as 10,000 nits. When you select 100 nits, the entire range is supposed to be compressed to 100 nits. This would make the image very dark. The display can't reverse engineer these altered RGB values to restore the original source brightness.

That's quite true, but, the result I'm getting is exactly what I described. So something is very wrong and possibly with my setup. I'll see if I can look at it and figure out what's going on. Cause I agree with you, 100 nits should NOT be displaying the same brightness.

EDIT: Yea something's very wrong here. I went back to madvr 0.9.17 and I'm seeing the same problem so it's not the latest test builds. I'm getting the exact same output from the tone mapping option as I do with passthrough. Doesn't matter what I change on the tonemapping option.

SamuriHL

4th October 2018, 23:53

mclingo

5th October 2018, 09:59

blu3wh0

5th October 2018, 11:39

Ok, got some help from madshi on this. First, the assumption that 100 nits with HDR turned on would automatically be dark was wrong. There is simply no way to know what the TV is going to do. In my LG C8's case, it cranks up the brightness. Second, he gave me some excellent source material to show what's going on with. One is his 10000 nit color test pattern (32 bit players only) and the other is the spear in BvS. HDR tone mapping is *DEFINITELY* active and working. And with those 2 different source materials it's a huge difference. WAY more detail in the spear scene and even his test pattern with tone mapping enabled in madvr. Now I'll have to play some more tomorrow.

When I initially tested this on a couple of movies, like you, I really couldn't tell the difference between passthrough and madVR tone mapping, no matter what peak brightness I chose. So, it's good to hear that there is actually a more subtle improvement. It sounds like we have no control over the brightness output of our TVs, but color/detail is not touched. In this case, what are the best options for madVR tone mapping with HDR output? What peak brightness should we set if this actually has no impact on our TVs? Or does this have some other subtle impact that we just haven't seen yet? I always have dynamic tone mapping on since it better represents what the brightness should be during the scene. Do these two things, dynamic tone mapping and madVR ton mapping, actually work together properly? I may do some testing this weekend since this sounds more promising now, meaning we could actually see some benefit from the HDR to SDR processing work. I just can't see myself watching HDR movies in SDR on our TVs due to the significantly limited brightness output.

SamuriHL

5th October 2018, 14:24

Your questions are very difficult to answer because it depends on a lot of variables. And I'm not entirely sure myself what looks "best" at the moment. Now that I have scenes which definitely show the differences it's a matter of sitting and testing. I'd start by using a target nit close to your TV max. For oled between 700 and 800. Then the restore details you'll have to play with. With hdr on I prefer a higher setting. Also make sure to turn on the color gamit option in the trade performance options. I highly recommend testing it on the BvS spear scene as you will see quite clearly what's happening with the picture with those options.

Sent from my Pixel XL using Tapatalk

SamuriHL

5th October 2018, 14:28

no fair, I cant turn my tone mapping off on my TV :(That doesn't mean the option is entirely useless. Even if your TV is tone mapping you could benefit. Think about it this way... Your TV tone maps when the movie is above the limits of the TV. In theory, and I stress the theory part of this because as madshi said we don't actually know what the TV does internally, if you make it closer to the TV parameters the internal tone mapping should have less to do. Try what I recommend in my previous post and see what happens. It'll become very clear if there's a benefit or not.

Sent from my Pixel XL using Tapatalk

mclingo

5th October 2018, 14:34

i tried it, it seemed to made no difference I could see, should I try a different value than 400 nits, what else can i try / change?

SamuriHL

5th October 2018, 14:52

What scene did you try it on? In motion it's hard to see the benefits clearly. Paused in the right scene the impact becomes crystal clear. I typically target about 800 nits.

Sent from my Pixel XL using Tapatalk

mclingo

5th October 2018, 15:04

just what I had available, interstellar 4K HDR, the bit when they visit the cloud planet, coming down through the clouds, thought that would be good to show any highlights that were brought out.

My TV does 400 nits max at 2%, drops off horribly after that.

SamuriHL

5th October 2018, 15:35

Then keep the target nits at 400. As for the scene, it might be a good one to show or it might not. If you look at the AVS thread, there's a post from april that has a few of the test scenes that they're using to evaluate the changes in each test build. One of them is what they call the green spear scene from BvS and that's the one I used to see exactly what kind of differences I get from madvr tone mapping vs pass-through. I would recommend that. Or madshi's 10000 nit color test pattern. For that you'll have to install a 32 bit player if you don't have one, and then install his filter. I will tell you that is an amazing test pattern to see the differences. As soon as I loaded that up my mind was blown.

mclingo

5th October 2018, 15:41

my chain is still 32 bit as KODI DS was always more stable in 32 bit, it might be ok now, just havent tried it. I have a look at those test patterns then.

mclingo

5th October 2018, 15:48

mclingo

5th October 2018, 16:07

Warner306

5th October 2018, 16:18

Ok, got some help from madshi on this. First, the assumption that 100 nits with HDR turned on would automatically be dark was wrong. There is simply no way to know what the TV is going to do. In my LG C8's case, it cranks up the brightness. Second, he gave me some excellent source material to show what's going on with. One is his 10000 nit color test pattern (32 bit players only) and the other is the spear in BvS. HDR tone mapping is *DEFINITELY* active and working. And with those 2 different source materials it's a huge difference. WAY more detail in the spear scene and even his test pattern with tone mapping enabled in madvr. Now I'll have to play some more tomorrow.

You should post some screenshots, so others can see what is supposed to be happening. It still sounds wacky when 100 nits and 1,000 nits produces the same brightness. I know the 2018 LG OLEDs have very dynamic tone mapping, but it sounds a little too dynamic. The PQ gamma curve is supposed to be predictable.

For example, you may have an RGB value of R(668) G(668) B(668). This is white to be shown at exactly 400 nits.

madVR might tone map this to R(519) G(519) B(519). This is white to be shown at exactly 100 nits.

Dynamic tone mapping is supposed to show the 400 nits white as close as possible to 400 nits to reduce compression whenever possible based on the brightness of the scene. Increasing the brightness of any value would destroy the intent of the original HDR grade. So the 100 nits white should also be shown at 100 nits. No HDR content would be graded to 100 nits. I don't know how the display would know to make this brighter. The image only gets darker to preserve detail, never brighter, unless the display is butchering the source.

Like I said, some screenshots would be helpful. Many will be confused by your results and probably won't know what to look for.

Warner306

5th October 2018, 16:26

Hi, unfortunately my TV doesnt have any such settings, it set as stock with minor changes to low IRE colour settings only, I used to have a Sony LCD which did so I know what you mean.

However, this also happens with my Samsung TV so i've pretty much ruled Mt TV and AVR out of the equation.

I'm probs going to have to re-image my HTPC but I'm really reluctant to do that as its a 3 day build at least as its also a server with 12 disks in it - but also, if I dont know what is causing it, if it comes back I'll be back to square one and 3 days down.

I was hoping MADSHI had an AMD contact but because there are only a couple of people with this problem, and thats its defeatable, he's not really got time to look at it, which is fair enough.

I was just hoping that with the way it manifested itself it would be obvious to someone what was going on.

Thanks for your interest though :)

I just looked up your display at Rtings.com. The color gamut should be set to "Standard" for BT.709 and "Wide" for BT.2020.

Quote:

"You can get a slightly wider color gamut when 'Color Gamut' is set to wide, but this isn't recommended for normal content because the colors will be over saturated."

SamuriHL

5th October 2018, 16:27

just had a go at MADSHIs test patters. Ran this installer, nothing seemed to happen, opened one of he files with MPC-HC which has MADVR set as renderer, "failed to render", i'm trying this over an RDP connection from work to home, i'll try it again when i get home later.

It's not going to work over RDP. But beyond that, DO NOT run the installer as admin. I made that mistake. Just run the bat file. "no fancy right click tricks" as madshi said to me. LOL

SamuriHL

5th October 2018, 16:30

You should post some screenshots, so others can see what is supposed to be happening. It still sounds wacky when 100 nits and 1,000 nits produces the same brightness. I know the 2018 LG OLEDs have very dynamic tone mapping, but it sounds a little too dynamic. The PQ gamma curve is supposed to be predictable.

For example, you may have an RGB value of R(668) G(668) B(668). This is white to be shown at exactly 400 nits.

madVR might tone map this to R(519) G(519) B(519). This is white to be shown at exactly 100 nits.

Dynamic tone mapping is supposed to show the 400 nits white as close as possible to 400 nits to reduce compression whenever possible based on the brightness of the scene. Increasing the brightness of any value would destroy the intent of the original HDR grade. So the 100 nits white should also be shown at 100 nits. No HDR content would be graded to 100 nits. I don't know how the display would know to make this brighter. The image only gets darker to preserve detail, never brighter, unless the display is butchering the source.

Like I said, some screenshots would be helpful. Many will be confused by your results and probably won't know what to look for.

I can post some screenshots later. This is a quote from madshi when I asked about the 100 nits vs 1000 nits being darker:

Why would it be quite dark? If the LG is detecting that the brightest pixel only has 100nits, the LG might pump up image brightness, making the overall image much brighter than when using a higher target peak nits setting in madVR. Or it might not. It's virtually impossible to predict how any TV will react to madVR doing tone mapping and then sending the video in HDR format. Virtually *anything* can happen.

In any case, yea, I'll try to do some screenshots a bit later.

huhn

5th October 2018, 16:32

out of interest, why dont NVIDIA provide 23,976 native like AMD, is this hard coded into the architecture or something, seems really odd they've ignored this for so long, it would surely solve lots of CRU and driver issues?

this is unbelievable unimportant for nvidia. HTPC is simple not important why do something that's still "working" if changing breaks something that is important or why should they waist time doing something for HTPC if they could improve FPS in what ever popular game about this and if you would really read into that topic you would very fast learn that a "perfect refreshrate" will not fix this issue because the real issue comes from the fact there is more than one (usually 3)clock generator and a they are far far from perfect.

the display HZ in madVR is relative to the system clock let's call it wrongly "CPU" clock but the repeated dropped frames come form a GPU and audio clock differences.

mclingo

5th October 2018, 16:38

I just looked up your display at Rtings.com. The color gamut should be set to "Standard" for BT.709 and "Wide" for BT.2020.

Quote:

"You can get a slightly wider color gamut when 'Color Gamut' is set to wide, but this isn't recommended for normal content because the colors will be over saturated."

Thanks yeah, I've got it set to WIDE for my HDR preset, I guess I should have known thats what they stood for sorry :)

mclingo

5th October 2018, 16:46

this is unbelievable unimportant for nvidia. HTPC is simple not important why do something that's still "working" if changing breaks something that is important or why should they waist time doing something for HTPC if they could improve FPS in what ever popular game about this and if you would really read into that topic you would very fast learn that a "perfect refreshrate" will not fix this issue because the real issue comes from the fact there is more than one (usually 3)clock generator and a they are far far from perfect.

the display HZ in madVR is relative to the system clock let's call it wrongly "CPU" clock but the repeated dropped frames come form a GPU and audio clock differences.

I guess their actions back up what you are saying, its unimportant to them but when they are selling low level cards like the GT3030 which is clearly aimed at HTPC's you would have thought they'd want to address this.

You are right of course but how hard can it be if AMD can do it and still provide much of the same gaming abilities.

SamuriHL

5th October 2018, 16:50

This is with madshi's hdr10000nitcolored test pattern:

Pass-through

https://i.postimg.cc/SxwrgJ2m/pass-through.png

madvr 800 nit target, high detail recovery

https://i.postimg.cc/15VMwGXv/madvr_tonemap.png

huhn

5th October 2018, 17:28

mclingo you only read the part you want to hear didn't you?

even fixing it doesn't fix it. it's not that complicated. even through the review don't understand it and post stuff like this: https://images.anandtech.com/doci/12572/23.png

https://www.anandtech.com/show/12572/the-intel-hades-canyon-nuc8i7hvk-review-kaby-lakeg-benchmarked/7

with a repeated frame ever 21 mins very accurate...

and it is no secret that nvidia has a far bigger driver development team than amd...

AMD broke deinterancing with crimson and it took them month if not a year to fix it because HTPC are unimportant for all GPU vendor.

and nvidia, AMD and intel will break there stuff again and again and if it only affects HTPC user they mostly will take there time to do anything about it.

polaris can't decode VP9 in hardware not even 8 bit which can be done with nvidia cards since 2015.
so AMD actions they say a lot too....

mclingo

5th October 2018, 17:31

steady on fella, lets dial this down.

I actually agreed with you didnt i, sure you arent reading what you want to read?

mclingo

5th October 2018, 17:37

I couldnt live with 21 mins though mate, I cant have any issues at all watching a movie or I get taken right out of the immersion, this is one of the main reasons I use AMD cards.

yeah AMD have had their issues no doubt, just cant help feeling if NVIDIA sorted this 23,976 issue it would solve a few things for you guys, unreasonable statement?

huhn

5th October 2018, 17:37

You are right of course but how hard can it be if AMD can do it and still provide much of the same gaming abilities.

sorry hard to believe.

mclingo

5th October 2018, 17:42

sorry hard to believe.

Which bit, we're specifically discussing 23,976 support which AMD have had for ever.

I'm actually pretty open minded about hardware, I dont have any preferences, green/ red I dont care, if NVIDIA had native 23,976 i've be more inclined to go green and get the extra grunt.

AMD does seem to be more stable with 3D switch though too.

huhn

5th October 2018, 17:51

ok let's try again.

the display HZ in madVR is relative to the system clock let's call it wrongly "CPU" clock but the repeated dropped frames come form a GPU and audio clock differences.

in short matching A/V is luck nothing more and can totally change by just replacing the mainboard or soundcard hell even the GPU.

https://images.anandtech.com/doci/12572/23.png

this is a screen of a supposed accurate clock from intel i guess which is known to have the closes clock and guess what it doesn't matter... it is still repeating frames.

did you checked your OSD?

mclingo

5th October 2018, 18:02

ah right, not saying AMD have it perfect, they just have it setup in such a way the drops occur outside way outside the length of a movie so you'll never see any. I check my OSD every time I change something on my system or find a problem.

Usually for a movie, when it settles down it reports several hours, but often it can show days between drops, surely AMD cant just be lucky, they must be doing something here that NVIDIA just arent, surely?

huhn

5th October 2018, 18:11

they the number of nvidia are worse getting days out of the box is very unlikely.

but do yu want to tell me 21 mins are better? and thats not there fails there are way to many variables.

why do you think there is no general CRU setting that everyone is using with nvidia? you will get a better number on your madVR OSD but if or if you are not dropping repeating frames is a different story.

your clocks are so unstable that it ranges from hours to days?

and AMD is not the lucky one you are the lucky one.

Klaus1189

5th October 2018, 19:15

I tested several different Nvidia cards not only the 1080Ti, also got an 1050Ti of a very different date and pcb manufacturer and I always got very bad 23 Hz frame drop/repeat timings- every 4 .. 5 minutes out of the box.
A few years back I used an Sapphire Radeon and got also better 23 Hz out of the box without complicated custom timings which also didn't work or were rejected by the Nvidia driver (some versions).
My current RX 480 has about 11,6 hours (23 Hz) out of the box, which is enough for my needs.
24 Hz no framedrops, 50 Hz not framedrops, 59 Hz 6.8 hours :D
Everything out of the box

But again, if situation of Nvidia is getting better, I might buy one again. But for now - not today and not this week ;)

Since I am still interested in a Nvidia card, which driver version(s) are known to work, so
1.) HDR switching without FSE in madvr (latest version) works
2.) custom timings generated by madvr are not rejected by driver
3.) has to work on Win10 Pro x64 1809
4.) The selected bit depth of Desktop isn't lost after reboot.
5.) Did I miss any other issue?

iSeries

5th October 2018, 19:15

SamuriHL

5th October 2018, 19:58

That's because madshi's not done with the tone mapping yet. :) What it's showing is that at 800 nit target, we have detail recovery on high and you can see that it recovers quite a bit of detail from the image. Saturation is what they're working on now in the new test builds. The likely reason you prefer pass-through in my images is that the saturation has not been boosted. What I'm trying to show is that I definitely am getting madvr's tone mapping vs pass-through. The quality is going to improve with each build until he gets dynamic tone mapping complete.