Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 15th October 2017, 17:01   #46581  |  Link
cork_OS
Registered User
 
cork_OS's Avatar
 
Join Date: Mar 2016
Posts: 160
Quote:
Originally Posted by foozoor View Post
I suspect madshi wanted to fool websites like majorgeeks or videohelp.
videohelp is pretty useful site and it still have 0.92.6 listed.
__________________
I'm infected with poor sources.
cork_OS is offline   Reply With Quote
Old 15th October 2017, 18:49   #46582  |  Link
mrcorbo
Registered User
 
Join Date: Jan 2005
Posts: 171
Quote:
Originally Posted by thewildsun View Post
I'm encountering a strange problem where video playback is smooth for the first 6-8 minutes or so, and then I start dropping frames like crazy. Pausing and then playing again doesn't help, but restarting the video player (PotPlayer, although I've also tried MPC-HC) fixes the problem immediately... for the next 6-8 minutes. Task manager doesn't show anything that stands out while frames are being dropped -- no spikes in CPU, GPU, or RAM.

It doesn't seem like a hardware issue, as madVR is the only thing that has seems to have any issues. I'm on the latest windows 10 insider build, so this might be the culprit, but it only happens on my NVIDIA GTX 1070 machine. I have another desktop with the AMD RX-480 that's also on the latest insider build, but the problem doesn't happen there. I've tried old NVIDIA drivers and old madVR builds, but had no luck.

Has anyone encountered anything like this before?
In the Nvidia Control panel, under Manage 3D settings, do you have Power Management Mode set to "Optimal Power"? If so, set it to "Adaptive" instead.
mrcorbo is offline   Reply With Quote
Old 15th October 2017, 19:01   #46583  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by mparade View Post
Is it possible to set up a separate setting profile for 3D frame-packed content?
Yes, there should be a working "3D" profile condition supported.

Quote:
Originally Posted by mparade View Post
Is it the correct behaviour of madVR to upscale BD sources/DVD sources to 4320x2430p?
I don't know. Show me a screenshot of the OSD (Ctrl+J), and I might be able to say more.

Quote:
Originally Posted by mparade View Post
Is it the correct behaviour of madVR to output at 8bit
even if chosing 10bit as "the native display bitdepth"?
My source is 8bit (BD mostly), but my TV can handle 10bit surely.
10bit output is only possible with D3D11 presentation, and in older OSs only in FSE mode.

Quote:
Originally Posted by mzso View Post
I tried... Here they are: https://drive.google.com/open?id=0By...DUySEFpdUQ3aVk
But the symptoms are not quite identical to Potplayer. The madVR OSD doesn't appear. I get a windows alert sound after a few secs, and sometimes MPC gets unhung and the playback starts. Also hanging seems to occur differently, when I clobber one file instead of when I change between files.
Thanks, will have a look at this later.

Quote:
Originally Posted by foozoor View Post
Where is the v0.92.7 changelog?
Haha, I didn't expect people to notice the website change so quickly. I was in the process of uploading v0.92.7, so I already updated the website, but then had to change something again. Upload coming "soon". The real version announcement is always in this thread. If it's not announced here, it's not up yet.

Quote:
Originally Posted by omw2h View Post
A sample of one distorted image might not be fair, but it shows my point where i find the end result of the new AS
lacking in comparison and i find no shorige of bad content.
I'm using the awesome reducing random noise and compression artefacts when needed, but in this situation i have to
use them, and this solution brings the questions of subtle harm and more GPU freedom being lost (then just having the previus AS checked).
If other options come to light more joy to me, i did caught on that the new AS is stronger,
so if i tone it down a bit and if the grain is not as visible i guess i'll survive the change.
Just give the next build a try and let me know what you think. I'm very willing to listen to feedback. Unfortunately not many people have commented on the new AdaptiveSharpen yet.

Quote:
Originally Posted by Asmodian View Post
It looks like madshi actually controls that website now?!
Yes, have been for quite a while already. The original owner was very nice when I contacted him. He transferred the domain to me without any hassle whatsoever.

Quote:
Originally Posted by ashlar42 View Post
You know what? I'm gonna say a simple thing.

I hope everybody here realizes how lucky we have to have you.
Thanks!

Quote:
Originally Posted by austinminton View Post
I mean what setup is he using to reach the conclusion that 8 bit is better than 10 bit for him. By keeping madvr at 8 bit hes defeating the whole purpose of doing this test? His tv will never get the 10 bit signal?
Well, I suppose he should ideally answer himself.
madshi is offline   Reply With Quote
Old 15th October 2017, 19:10   #46584  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
@madshi

Quote:
10bit output is only possible with D3D11 presentation, and in older OSs only in FSE mode.
I am on D3D11 path (set in madVR) and on Win10 64 bit but in JRiver the crtl+J
says I am in window mode. In window mode 10 bit is also possible?

Quote:
Yes, there should be a working "3D" profile condition supported.
Excellent.
mparade is offline   Reply With Quote
Old 15th October 2017, 19:13   #46585  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
madVR v0.92.7 released

http://madshi.net/madVR.zip

Code:
* added experimental deblock option "don't reduce texture detail"
* added experimental AdaptiveSharpen option "linear light"
* added new lowest "reduce compression artifacts" strength
* revisited "reduce random noise" strength levels, now from 1-12
* updated AdaptiveSharpen to latest igv version
1) I'd like to get rid of the new "linear light" option for AdaptiveSharpen, so please everybody who uses/likes AdaptiveSharpen try the "linear light" option and let me know if you prefer it checked or unchecked.

2) Would be extra nice if you could also compare the new AdaptiveSharpen to the old one (v0.92.4 and older).

3) "reduce random noise" has new strength levels from 1-12. The very lowest setting doesn't seem to do much, if anything at all. Does anybody find it useful? If nobody says he wants to keep it, I'll drop it in the next build. I think strength level 2 will probably be the best starting point?

4) I've added a new lowest "reduce compression artifacts" strength, the other strength settings got moved up one. But to be honest, the new lowest strength 1 doesn't seem to do much, so I'm not sure if it's really useful. What do you think?

5) There's a new option called "don't reduce texture detail", which activates a slightly modified "reduce compression artifacts" algorithm. The original algo included some random noise reduction over the whole video frame. The modified algo tries to do noise reduction only near edges. The difference is not overly dramatic, though. Which option do you like more? I think I'm going to switch to the "don't reduce texture detail" algo variant and remove the option in the next build, but wanted to give you the chance to compare first.
madshi is offline   Reply With Quote
Old 15th October 2017, 19:27   #46586  |  Link
PurpleMan
Registered User
 
Join Date: Oct 2003
Posts: 273
Quote:
Originally Posted by madshi View Post
First of all, are we talking about bitmap type subtitles (VOBSUB, SUP, Blu-Ray m2ts included subs etc)? Or are we talking about text type subtitles (SRT, ASS, etc)? Can you please also test "the other type" to see if those are also affected?

Finally, a small video sample which shows all these problems you mentioned would be helpful. Or if it's hard to find one sample which reproduces all these problems, then I'm also fine with multiple smaller video samples, one for each problem.
These are SRT text subtitles, not bitmap/vobsub. I don't have a bitmap/vobsub one handy so couldn't test to reproduce.

I think I managed to depict this using screenshots and adding position markers. Here are 4 different subtitles, with the 'move to bottom' feature and without the 'move to bottom' feature:

Subtitle 01 || with move-to-bottom || without move-to-bottom
Subtitle 02 || with move-to-bottom || without move-to-bottom
Subtitle 03 || with move-to-bottom || without move-to-bottom
Subtitle 04 || with move-to-bottom || without move-to-bottom

You can see that there are 6 different horizontal line positions in the "with move-to-bottom" screenshots but only 2 fixed positions (one for each line) in the "without move-to-bottom" screenshots.

Do you see what I mean?
PurpleMan is offline   Reply With Quote
Old 15th October 2017, 19:31   #46587  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by mzso View Post
I tried... Here they are: https://drive.google.com/open?id=0By...DUySEFpdUQ3aVk
But the symptoms are not quite identical to Potplayer. The madVR OSD doesn't appear. I get a windows alert sound after a few secs, and sometimes MPC gets unhung and the playback starts. Also hanging seems to occur differently, when I clobber one file instead of when I change between files.
The logs say v0.91.10 instead of v0.92.6/7, though. And unfortunately the PDB debug symbols don't seem to work at all... Are they in the right folder? Meaning the PDB symbols for MPC-HC are in the same folder and have the same file name (except for the extension, obviously) as the MPC-HC main exe file?

Quote:
Originally Posted by mparade View Post
I am on D3D11 path (set in madVR) and on Win10 64 bit but in JRiver the crtl+J says I am in window mode. In window mode 10 bit is also possible?
Depends on the GPU. With an Nvidia or AMD GPU it should be possible, but it must be fullscreen windowed mode, not just windowed mode. FSE mode would be best, of course.

Quote:
Originally Posted by PurpleMan View Post
These are SRT text subtitles, not bitmap/vobsub. I don't have a bitmap/vobsub one handy so couldn't test to reproduce.

I think I managed to depict this using screenshots and adding position markers. Here are 4 different subtitles, with the 'move to bottom' feature and without the 'move to bottom' feature:

Subtitle 01 || with move-to-bottom || without move-to-bottom
Subtitle 02 || with move-to-bottom || without move-to-bottom
Subtitle 03 || with move-to-bottom || without move-to-bottom
Subtitle 04 || with move-to-bottom || without move-to-bottom

You can see that there are 6 different horizontal line positions in the "with move-to-bottom" screenshots but only 2 fixed positions (one for each line) in the "without move-to-bottom" screenshots.

Do you see what I mean?
Yes, I see what you mean. Good screenshots. Now can you provide me with a small sample with which I could easily reproduce this? Thanks!
madshi is offline   Reply With Quote
Old 15th October 2017, 19:38   #46588  |  Link
j82k
Registered User
 
Join Date: Jun 2017
Posts: 155
Quote:
Originally Posted by austinminton View Post
I mean what setup is he using to reach the conclusion that 8 bit is better than 10 bit for him. By keeping madvr at 8 bit hes defeating the whole purpose of doing this test? His tv will never get the 10 bit signal?
I basically compared:

- Nvidia 10/12 bit, madVR 10 bit

to

- Nvidia full RGB 8 bit, madVR 8 bit

MadVR was correctly showing "D3D11 fullscreen windowed (10bit)" while testing 10/12 bit.

I think I tried all available 10/12bit modes. (RGB, 4:2:0, 4:2:2, 4:4:4) and RGB 8 bit still had the best distinction between the different grayscale steps. With 10/12 bit there were always some steps bunched together as if they were just one color.

Also I just played this 10-bit pattern through usb using the TV's internal mediaplayer and it looked worse than madVR's dithered 8-bit. I have to say this isn't surprising, as LG Oled TVs are known to have color banding issues.
j82k is offline   Reply With Quote
Old 15th October 2017, 19:48   #46589  |  Link
PurpleMan
Registered User
 
Join Date: Oct 2003
Posts: 273
Quote:
Originally Posted by madshi View Post
Yes, I see what you mean. Good screenshots. Now can you provide me with a small sample with which I could easily reproduce this? Thanks!
Yay! Glad this helps!
Here's a sample SRT file with the same 4 lines I used above:

test.srt

You can change the timestamps in it to fit your needs.

Thank you!
PurpleMan is offline   Reply With Quote
Old 15th October 2017, 19:49   #46590  |  Link
Anima123
Registered User
 
Join Date: Jun 2005
Posts: 504
Wow, some high-compressed videos really need as strong deblock as like 8, to get a nice result.

Didn't notice any difference between "don't reduce texture detail" on and off with real-life content.

Edit: Also I found that the NGU-AA is not as good as NGU Soft with the deblock algorithm, NGU Soft 4x doesn't do any better than NGU Soft 2x.

Edit2: With stronger as far as 9, the "don't reduce texture detail" did result in better quality. It's okay for me to become the default algorithm.

Last edited by Anima123; 15th October 2017 at 22:00.
Anima123 is offline   Reply With Quote
Old 15th October 2017, 19:57   #46591  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
@madshi

Quote:
I don't know. Show me a screenshot of the OSD (Ctrl+J), and I might be able to say more.

here is the screenshot:

http://s1.toldacuccot.hu/letoltes?si...creenshot1.png

thank you!
mparade is offline   Reply With Quote
Old 15th October 2017, 20:20   #46592  |  Link
thewildsun
Registered User
 
Join Date: Oct 2015
Posts: 10
Quote:
Originally Posted by mrcorbo View Post
In the Nvidia Control panel, under Manage 3D settings, do you have Power Management Mode set to "Optimal Power"? If so, set it to "Adaptive" instead.
It's set to max performance. I tried changing it to adaptive anyway, just in case, but still no luck.
thewildsun is offline   Reply With Quote
Old 15th October 2017, 21:12   #46593  |  Link
Neo-XP
Registered User
 
Neo-XP's Avatar
 
Join Date: Mar 2016
Location: Switzerland
Posts: 140
Quote:
Originally Posted by madshi View Post
madVR v0.92.7 released
1) Checked, because without linear light it produces/accentuates too much artifacts near the edges.

2) Not tested.

3) Is it working ? levels 1-2 do nothing (?) and 3 way too much. Not a setting for me I think.

4) It is already doing a lot at strength 1 without the "don't reduce texture details" checked.

5) With the "don't reduce texture details" option checked, there are a lot of very ugly compression artifacts left in the image that are not processed at all. Why not keep this option ?

Just a test I did (Source / RCA@2 with the option enabled / RCA@2 with the option disabled) :



And zoomed :



It is quite obvious if you switch from one image to another. For instance on the chin of the guy in the left side, on the books behind him and on his shirt.

Thank you for letting us test this.

Last edited by Neo-XP; 15th October 2017 at 21:15.
Neo-XP is offline   Reply With Quote
Old 15th October 2017, 21:45   #46594  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by madshi View Post
The logs say v0.91.10 instead of v0.92.6/7, though. And unfortunately the PDB debug symbols don't seem to work at all... Are they in the right folder? Meaning the PDB symbols for MPC-HC are in the same folder and have the same file name (except for the extension, obviously) as the MPC-HC main exe file?
Oh, yeah... I forgot about that I downgraded because of the error at bookmarking. (But it shouldn't make much of a difference since the new version hung just the same)

The PDB-s I got from the mpc-hc download page, which had an identical folder structure to the main program, and included LAV PDBs. So I copied all to the program folder. (I just checked and the names are right.) So I don't have any idea why wouldn't it work. (Whatever it's supposed to be doing.)

So what now?

Edit:
Anyway, I made some new ones (#8-11) with lates madVR, maybe (if unlikely) it helps.

Last edited by mzso; 15th October 2017 at 21:56.
mzso is offline   Reply With Quote
Old 15th October 2017, 21:49   #46595  |  Link
ABDO
Registered User
 
Join Date: Dec 2016
Posts: 65
Quote:
Originally Posted by madshi View Post
haha, i didn't expect people to notice the website change so quickly.
we do, I visit posts py madshi tens times Per day to find any news from you, because you the best one in the world for video fans like me.
form the bottom of my heart thank you madshi for this stunning work
ABDO is offline   Reply With Quote
Old 15th October 2017, 22:25   #46596  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by PurpleMan View Post
Here's a sample SRT file with the same 4 lines I used above:

test.srt

You can change the timestamps in it to fit your needs.
Thanks.

Quote:
Originally Posted by Anima123 View Post
Edit2: With stronger as far as 9, the "don't reduce texture detail" did result in better quality. It's okay for me to become the default algorithm.
Good to know.

Quote:
Originally Posted by mparade View Post
Looks weird. It seems to me that your media player has sized the madVR window much too big. Your screenshot is 1920x1080, but according to the OSD, the madVR rendering window is sized to something bigger than 4K.

Which media is that? Have you tried another one, e.g. MPC-HC/BE?

Quote:
Originally Posted by thewildsun View Post
It's set to max performance. I tried changing it to adaptive anyway, just in case, but still no luck.
Other possible causes could be 3rd party software accessing the GPU, like GPU-Z, f.lux or similar? If that doesn't help, make a screenshot of the OSD (Ctrl+J) in the moment when frame drops occur. Don't attach is to this forum, just link to it here.

Quote:
Originally Posted by Neo-XP View Post
1) Checked, because without linear light it produces/accentuates too much artifacts near the edges.
Interesting. Do you have a good screenshot for that?

Quote:
Originally Posted by Neo-XP View Post
3) Is it working ? levels 1-2 do nothing (?) and 3 way too much. Not a setting for me I think.
The "reduce random noise" strength setting defines the threshold below which noise can be cleaned, and above which it's considered detail and kept. So depending on your image, a certain strength setting may do nothing, if the film grain is higher than the threshold, while increasing the strength settings by 1 might suddenly result in total smoothing. This behaviour may feel somewhat weird. An alternative approach would be to offer 2 separate controls: 1) noise vs detail threshold (current strength setting) and 2) reduction level of noise (currently always 100%). But then, this would be hard to understand. So I'll probably keep it as it is.

Quote:
Originally Posted by Neo-XP View Post
4) It is already doing a lot at strength 1 without the "don't reduce texture details" checked.
So you consider strength 1 useful?

Quote:
Originally Posted by Neo-XP View Post
5) With the "don't reduce texture details" option checked, there are a lot of very ugly compression artifacts left in the image that are not processed at all. Why not keep this option ?

Just a test I did (Source / RCA@2 with the option enabled / RCA@2 with the option disabled)
Hmmmm... I can't manage to reproduce your test results. The ugly artifacts on the chin are pretty much the same with RCA@2 for me, regardless of whether the option is enabled or disabled. I think your good looking screenshot must have had either a higher RCA setting, or a combination of RCA + RRN? Can you double check?

Quote:
Originally Posted by mzso View Post
Oh, yeah... I forgot about that I downgraded because of the error at bookmarking. (But it shouldn't make much of a difference since the new version hung just the same)

The PDB-s I got from the mpc-hc download page, which had an identical folder structure to the main program, and included LAV PDBs. So I copied all to the program folder. (I just checked and the names are right.) So I don't have any idea why wouldn't it work. (Whatever it's supposed to be doing.)
Could it be possible that you download the 32bit PDB files instead of the 64bit ones? Just a wild guess. Maybe you could try 32bit MPC?

Quote:
Originally Posted by mzso View Post
Anyway, I made some new ones (#8-11) with lates madVR, maybe (if unlikely) it helps.
Same as before, unfortunately.

Quote:
Originally Posted by ABDO View Post
we do, I visit posts py madshi tens times Per day to find any news from you, because you the best one in the world for video fans like me.
form the bottom of my heart thank you madshi for this stunning work
madshi is offline   Reply With Quote
Old 15th October 2017, 22:53   #46597  |  Link
thewildsun
Registered User
 
Join Date: Oct 2015
Posts: 10
Quote:
Originally Posted by madshi View Post
Other possible causes could be 3rd party software accessing the GPU, like GPU-Z, f.lux or similar? If that doesn't help, make a screenshot of the OSD (Ctrl+J) in the moment when frame drops occur. Don't attach is to this forum, just link to it here.
Thanks. I'm not using any 3rd party software like f.lux.

Here's the OSD screenshot: https://i.imgur.com/s2tMJWk.png

Edit: It might be a RAM-related issue. I noticed that RAM only ever increases while playing a video. That explains why restarting the video player fixes the problem temporarily. I'm not sure what to try next, though...

Edit2: Resetting madVR to default settings did seem to alleviate the issue (I was able to play a 20 minute video with no frame drops) but if the issue is RAM-related, then that may just be because the default settings are less RAM intensive.

Edit3: For chroma upscaling, I had been using Recon-soft with SuperRes 3. Turning off SuperRes seems to fix the issue -- I need to do more testing to see if this is a true fix or not.

Last edited by thewildsun; 16th October 2017 at 18:24.
thewildsun is offline   Reply With Quote
Old 15th October 2017, 23:10   #46598  |  Link
ABDO
Registered User
 
Join Date: Dec 2016
Posts: 65
Quote:
Originally Posted by madshi View Post
madVR v0.92.7 released
1) unchecked,with linear light it reduce Sharpen effect.

2) New AdaptiveSharpen better than the old one, the old did ahuge sharpen edges with visible artifacts, so i never used it, but the new one do enhance detail very nice and sharpen it with no visible ringing in the same time.

3) Not tested.

4) I agree strength 1 doesn't do much.

5) as @Neo-XP notice, With the "don't reduce texture details" option checked, there are some compression artifacts left in the image that are not processed.

Source


RCA 10 with the option disabled


RCA 10 with the option enabled


edit:
it seem (don't reduce texture detail ) option save some texture detail in normal life content, and it give good result to the whole picture.

look to the pullover line detail

Source


RCA 7 with the option disabled


RCA 7 with the option enabled


@madshi could you please keep this option for us, seem good for 3d content, and not to 2d cartoon.

Last edited by ABDO; 16th October 2017 at 02:29.
ABDO is offline   Reply With Quote
Old 15th October 2017, 23:38   #46599  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
@madshi

Quote:
Looks weird. It seems to me that your media player has sized the madVR window much too big. Your screenshot is 1920x1080, but according to the OSD, the madVR rendering window is sized to something bigger than 4K.

Which media is that? Have you tried another one, e.g. MPC-HC/BE?
JRiver latest 32-bit version. I haven't tried any other player yet (I have paid for it, so there shouldn't be any bug).

I have problems with "3D" profile rule as well.
Could you please have a look at this:

Code:
if (3D) "3D"

else if (srcWidth > 1920) and (not 3D)"2160p"
else if (srcWidth <= 1920) and (srcHeight > 1080) and (not 3D) "2160p"

else if (srcWidth > 1280) and (srcWidth <= 1920) and (not 3D) "1080p"
else if (srcWidth <= 1280) and ((srcHeight > 720) and (srcHeight <= 1080)) and (not 3D) "1080p"

else if (srcWidth > 960) and (srcWidth <= 1280) and (not 3D) "720p"
else if (srcWidth <= 960) and ((srcHeight > 540) and (srcHeight <= 720)) and (not 3D) "720p"

else if (srcWidth <= 960) and (srcHeight <= 540) and (not 3D) "SD"
When playing 3D ISO files madVR is using my "1080p" profile instead of the "3D" one.

I would really appreciate your help.

Last edited by mparade; 15th October 2017 at 23:53.
mparade is offline   Reply With Quote
Old 16th October 2017, 00:33   #46600  |  Link
Anima123
Registered User
 
Join Date: Jun 2005
Posts: 504
Madshi, what is the order of artifact removal algorithms, if over one single options have been used?

Well, as the order of them might have impacts on the final quality of the processed image sequences, will you plan to open the setting order selection for us, or you have a way to determine which order is optimal in most cases?
Anima123 is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 18:44.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.