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 11th August 2013, 19:06   #19821  |  Link
michkrol
Registered User
 
Join Date: Nov 2012
Posts: 167
Quote:
Originally Posted by ThurstonX View Post
the "Exclusive" banner doesn't clear. It does clear on my other PCs running madVR on Radeon HD 5670 & 6670.
It's driver's bug. For now disable "use separate device for presentation" in rendering->general settings. Or use older drivers.

Also I hope, you've mixed image and chroma algorithms, since using jinc3 for chroma is pointless if it means sacrificing image scaling quality. You'll only spot chroma upscaling differences on very specific images, like red object on light blue background and similar. Try checking it using "nearest neighbor" as kind of reference for low quality, both for chroma and image, you'll spot the difference instantly with image, not so with chroma. That said you should be ok with anything else for chroma (even bilinear). My preference is softcubic100 for chroma, but that's me.

Also MPC-HC nightly builds are worth checking out ( nightly.mpc-hc.org ), they're quite stable and "the" stable release is a little outdated.
michkrol is offline   Reply With Quote
Old 11th August 2013, 23:27   #19822  |  Link
ThurstonX
Registered User
 
Join Date: Mar 2006
Posts: 58
Quote:
Originally Posted by michkrol View Post
It's driver's bug. For now disable "use separate device for presentation" in rendering->general settings. Or use older drivers.

Also I hope, you've mixed image and chroma algorithms, since using jinc3 for chroma is pointless if it means sacrificing image scaling quality. You'll only spot chroma upscaling differences on very specific images, like red object on light blue background and similar. Try checking it using "nearest neighbor" as kind of reference for low quality, both for chroma and image, you'll spot the difference instantly with image, not so with chroma. That said you should be ok with anything else for chroma (even bilinear). My preference is softcubic100 for chroma, but that's me.

Also MPC-HC nightly builds are worth checking out ( nightly.mpc-hc.org ), they're quite stable and "the" stable release is a little outdated.
Thanks a lot for the info. I didn't mix the scaling, and I know Image is more "important," thus my sad face re: Jinc 3+AR I couldn't tell much difference with the various Chroma algorithms. I just tried Jinc 3+AR to see how it affected the render stats. Definitely increased it. For some older SD content, Lanczos for Image was too much, and Soft Cubic, which I prefer for such things, was better. I prefer something a little sharper for better content.

Anyway, I'll try unchecking "use separate device for presentation" and see how it goes.

Cheers.
ThurstonX is offline   Reply With Quote
Old 12th August 2013, 01:20   #19823  |  Link
SyrupBuccaneer
Registered User
 
Join Date: Dec 2012
Posts: 6
Getting occasional freezing when transitioning from window to full-screen in Potplayer. Log is below.

-Opened file
-Browsed to bookmark
-Double click to fullscreen
-Completely frozen video/player, audio still playing.
Attached Files
File Type: 7z madVR - log.7z (139.4 KB, 13 views)
SyrupBuccaneer is offline   Reply With Quote
Old 12th August 2013, 01:21   #19824  |  Link
Gagorian
Registered User
 
Join Date: Jul 2013
Posts: 27
I'm trying to verify the 3DLut created by ArgyllCMS. I'm using HCFR with madVR as pattern generator. As I'm running the greyscale series, is it normal that the grey screens are filled with some clear artifacting (posterization banding?)? Shouldn't it be homogenous? When playing the greyscale screens as .mp4 files from the AVSHD set there are no issues. There seems to be some compatibility issues as well, often the test patterns just freeze at 0/n. The ArgyllCMS patches in the madVR pattern generator also exhibit similar artifacting.

I also seem to be getting better results using DispcalGUI and generating a .icm profile and loading it than I did using the command lines to create a 3Dlut for MadVR. What are the advantages in madVR of a 3DLut compared a loaded .icm profile? I'm at least having issues with the full screen patterns, sometimes the deltaE's can go up to 15, but sometimes they are within 0-2 dE without doing any changes to the monitor or the profile. Is this an issue with the .icm profile not being applied?

Last edited by Gagorian; 12th August 2013 at 12:20.
Gagorian is offline   Reply With Quote
Old 12th August 2013, 01:23   #19825  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 5,814
you can use jinc 3 ar with an hd 4000

just set the cpu queue to 9-16 gpu to 9 (or to backbuffer +1) backbuffer to 8 don't use smoothmotion and don't use fullscreen exclusive mode and ofc don't use jinc 3 for chroma.

i didn't tryed it for 24hz +but u can try it.
huhn is offline   Reply With Quote
Old 12th August 2013, 01:58   #19826  |  Link
ThurstonX
Registered User
 
Join Date: Mar 2006
Posts: 58
Quote:
Originally Posted by huhn View Post
you can use jinc 3 ar with an hd 4000

just set the cpu queue to 9-16 gpu to 9 (or to backbuffer +1) backbuffer to 8 don't use smoothmotion and don't use fullscreen exclusive mode and ofc don't use jinc 3 for chroma.

i didn't tryed it for 24hz +but u can try it.
Yep, as michkrol noted, unchecking "use separate device for presentation" did the trick. I'd forgotten about it, as I rarely use this laptop for video playback, but I use the same setting on it. And as you note, I was able to get Jinc 3+AR for Image Scaling, but I left the buffer settings at their defaults. And Smooth Motion is just fine, as is FSE. But yeah, Jinc 3 for Chroma just dogged it, and made no sense.

Soft Cubic still rules for crappy encodes, esp. things like VCR captures and ancient AVIs. But Jinc 3+AR is da bomb for better quality videos.

I played back a DVD rip that's 23.976 fps using Soft Cubic 70+AR and Jinc 3+AR. Smooth Motion activated, and it was all glorious.

I'm still mucking about with Chroma settings, but without getting into a detailed method of comparison, just viewing a few files and changing them, I can't tell much difference. Soft Cubic 70+AR seems just fine.

Anyway, thanks for all the help

Last edited by ThurstonX; 12th August 2013 at 02:02.
ThurstonX is offline   Reply With Quote
Old 12th August 2013, 09:12   #19827  |  Link
THX-UltraII
Registered User
 
Join Date: Aug 2008
Location: the Netherlands
Posts: 850
Quote:
Originally Posted by ƒreeman View Post
why do you wanna remove it..??
it's a good sign that madVR is working..
I have to disagree with you on this. My HTPC is stable as hell and madVR works flawlessly and perfect and I know 1000% sure that exclusive mode is ALWAYS working. So there is really no need for me for the exclusive notification.

I (and I expect other too) use my HTPC and I don t want to notice ANYTHING about it beeing a PC. When I startup my system I immediately go to XBMC and I use MPC-BE as external player. The exclusive notification is the last esthetic thing I would like to see removed.
THX-UltraII is offline   Reply With Quote
Old 12th August 2013, 09:23   #19828  |  Link
dansrfe
Registered User
 
Join Date: Jan 2009
Posts: 1,212
Ok, so my situation is probably not a common one but just thought I'd mention it. When using different refresh rates with my particular monitor, the gamma curve of the display and/or luminance level darkens/lightens. Due to this I have to create separate ICC profiles and 3DLUTs.

However, madVR doesn't support loading of different 3DLUTs upon refresh rate switch therefore I just leave the monitor at 96Hz and disable the display mode switcher. However, that turns on smooth motion in cases where the refresh rate could have been switched to a multiple of the frame rate.

Last edited by dansrfe; 12th August 2013 at 09:26.
dansrfe is offline   Reply With Quote
Old 12th August 2013, 11:59   #19829  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by huhn View Post
i found a resize bug in in madvr
lanczos8ar video codec is ZMBV output is rgb24 http://s3.imgimg.de/uploads/lanczos8ar325f536apng.png
unscaled it looks like this http://s3.imgimg.de/uploads/schweif0...877b96bpng.png
bilinear and dxva look very similar and flicker with my gtx 760
jinc 3-8 without ar/with ar looks bad too
jinc 3 ar: http://s3.imgimg.de/uploads/jinc3ar6af6b90dpng.png
a lot of details are destroyed
i hope there isn't something wrong with my 760
I can't reproduce this on my system.

That said, generally increasing the number of taps will increase sharpness and reduce aliasing, but greatly reduce image quality due to the increase in ringing. (even with the AR filter enabled) I recommend sticking to the 3-tap variants of any algorithm.

And as for Jinc 3 destroying details - it is a scaling algorithm that is best suited to video footage rather than computer graphics. (especially pixel art) One of the things which Jinc does compared to the other scaling algorithms available in madVR is that it will smooth out diagonal lines - which is normally great for video, but can be destructive with pixel art as you have found out.

In some cases, you may even want to switch to Nearest Neighbor scaling when viewing pixel art. You can set up keyboard shortcuts to manually switch algorithms in madVR's preferences. (because you would not want to use Nearest Neighbor with other video content)

I do think it is interesting that Jinc 3 turns the dither into the intermediate shade that it was trying to approximate though. This is something that CRTs did naturally (well, they did when using a poor connection at least - not via RGB) and is something that emulators are only just starting to tackle.

Quote:
Originally Posted by ThurstonX View Post
madVR's render times are great with Jinc 3+AR for Chroma and Soft Cubic 70+AR or Lanczos 3+AR for Image Scaling (Jinc 3+AR / Lanczos3+AR gives me about 11ms).
I don't recommend using the anti-ringing filter with SoftCubic scaling, and you may wish to reconsider using Jinc scaling for Chroma if it allows you to use better Luma settings. (though SoftCubic is still my preferred algorithm for DVD content anyway)

Most people will not see a difference between Bicubic 75 AR, Lanczos 3 AR, or Jinc 3 AR chroma.
It also depends what your display is doing - many process in 4:2:2, so anything above Bicubic 75 AR is wasted.

Quote:
Originally Posted by ThurstonX View Post
I'm still mucking about with Chroma settings, but without getting into a detailed method of comparison, just viewing a few files and changing them, I can't tell much difference. Soft Cubic 70+AR seems just fine.
If you search through my post history, you should be able to find some detailed comparisons that illustrate the worst-case scenarios for chroma. (red text on black)

The short version is to use one of the three options I listed above. Use Jinc 3 if your display is capable of showing 4:4:4 and you have the GPU power spare, but in most cases, you won't even see the difference there. I would be happy using Bicubic 75 AR.
6233638 is offline   Reply With Quote
Old 12th August 2013, 13:32   #19830  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 5,814
i just tryd it on my second pc (amd 6770) and i can't reproduce it too.

it is still happening on my main pc (gtx 760) plus the bilinear/dxva flicker problem. so there is something wrong with my pc or nvidia ...

Quote:
In some cases, you may even want to switch to Nearest Neighbor scaling when viewing pixel art. You can set up keyboard shortcuts to manually switch algorithms in madVR's preferences. (because you would not want to use Nearest Neighbor with other video content)
in this case i would love to have an resize option to 1600x1000 (source is 320x200) with nearest neighbor and then a different resize to 1080p
huhn is offline   Reply With Quote
Old 12th August 2013, 19:20   #19831  |  Link
ThurstonX
Registered User
 
Join Date: Mar 2006
Posts: 58
Quote:
Originally Posted by 6233638 View Post
I don't recommend using the anti-ringing filter with SoftCubic scaling, and you may wish to reconsider using Jinc scaling for Chroma if it allows you to use better Luma settings. (though SoftCubic is still my preferred algorithm for DVD content anyway)

Most people will not see a difference between Bicubic 75 AR, Lanczos 3 AR, or Jinc 3 AR chroma.
It also depends what your display is doing - many process in 4:2:2, so anything above Bicubic 75 AR is wasted.

If you search through my post history, you should be able to find some detailed comparisons that illustrate the worst-case scenarios for chroma. (red text on black)

The short version is to use one of the three options I listed above. Use Jinc 3 if your display is capable of showing 4:4:4 and you have the GPU power spare, but in most cases, you won't even see the difference there. I would be happy using Bicubic 75 AR.
Thanks, I was hoping you'd throw out a few tips I always enjoy and certainly appreciate all the testing and posts you've made re: scaling.

You've given me good food for thought. I have Jinc 3+AR for both on my main HTPC (core i5, Radeon HD 6670 w/ 1GB GDDR 5). I'll turn off AR for Chroma and see how that looks.

On the Intel HD Graphics system in question Jinc 3+AR for Chroma disrupted the rendering, but I haven't had a chance to try it without AR. When I've time, I'll try it and the others you mention. It's mostly a TV PC (US digital through a Ceton InfiniTV 4), and most video files watched are low-res, so I end up with Soft Cubic ##+AR for Image.

TV is an LG 42LD550.

Anyway, thanks again
ThurstonX is offline   Reply With Quote
Old 13th August 2013, 00:29   #19832  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by ThurstonX View Post
Thanks, I was hoping you'd throw out a few tips I always enjoy and certainly appreciate all the testing and posts you've made re: scaling.
No problem. I really need to go back through this topic and compile a list of the posts which still have useful examples in them. madVR has changed so much that some of them may not be relevant any more.

Quote:
Originally Posted by ThurstonX View Post
You've given me good food for thought. I have Jinc 3+AR for both on my main HTPC (core i5, Radeon HD 6670 w/ 1GB GDDR 5). I'll turn off AR for Chroma and see how that looks.
I would actually suggest dropping down to a "lower level" of chroma scaling (preferably Lanczos 3 or Bicubic 75) that you can run with the AR filter, than using Jinc 3 with the AR filter disabled - in most cases, it's a better trade-off.

It's really very rare that you need anything more than Bicubic 75 AR. http://forum.doom9.org/showpost.php?...ostcount=16513
This comparison was done against Jinc 8, but it is not a recommendation for people to use Jinc 8 - Jinc 3 AR actually looks better than Jinc 8 99% of the time. If I recall correctly, I left Lanczos 3 out because it's almost indistinguishable from Bicubic 75.

I probably wouldn't recommend anyone increase Chroma above Bicubic 75 AR unless they are already using Jinc 3 AR Luma, have a display capable of showing 4:4:4 chroma, and have GPU power to spare.

Quote:
Originally Posted by ThurstonX View Post
When I've time, I'll try it and the others you mention. It's mostly a TV PC (US digital through a Ceton InfiniTV 4), and most video files watched are low-res, so I end up with Soft Cubic ##+AR for Image.
Again, I don't recommend SoftCubic with the anti-ringing filter.

I think the anti-ringing filter has been improved since this comparison was made, but these images illustrate what happens when you use SoftCubic+AR:
Quote:
Originally Posted by 6233638 View Post
Pay attention to the edges of the X in particular.
I don't recall there being problems using the AR filter with other algorithms.

Quote:
Originally Posted by ThurstonX View Post
TV is an LG 42LD550.
Chroma Resolution Test Pattern
4:4:4 and 4:2:2 examples using this pattern

If your display processes chroma in 4:2:2, there's no need to use anything more than Bicubic 75 AR - even that may be more than you need.
6233638 is offline   Reply With Quote
Old 13th August 2013, 00:55   #19833  |  Link
ThurstonX
Registered User
 
Join Date: Mar 2006
Posts: 58
Quote:
Originally Posted by 6233638 View Post
No problem. I really need to go back through this topic and compile a list of the posts which still have useful examples in them. madVR has changed so much that some of them may not be relevant any more.
That would be a great service to the community.

Quote:
Originally Posted by 6233638 View Post
Again, I don't recommend SoftCubic with the anti-ringing filter.

I think the anti-ringing filter has been improved since this comparison was made, but these images illustrate what happens when you use SoftCubic+AR:

Pay attention to the edges of the X in particular.
I don't recall there being problems using the AR filter with other algorithms.


Chroma Resolution Test Pattern
4:4:4 and 4:2:2 examples using this pattern

If your display processes chroma in 4:2:2, there's no need to use anything more than Bicubic 75 AR - even that may be more than you need.
Yep, the 'X' caught my eye immediately. I need to make a file of tests to run. Too much to keep in me 'ead

Cheers.
ThurstonX is offline   Reply With Quote
Old 13th August 2013, 08:46   #19834  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,201
Quote:
Originally Posted by ThurstonX View Post
That would be a great service to the community.
The man has done so much already!

Madshi may I suggest adding whatever 6233638 manages to cook up into your second post for stickiness?
ryrynz is offline   Reply With Quote
Old 13th August 2013, 09:05   #19835  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,463
Yeah, kudos for the 4:4:4 split screen test pattern that allowed me to find out that only HD DVB-T is processed in full chroma resolution on my new flat screen. The trick I've found to get 4:4:4 HDMI is to enable PIP(with an HD DVB-T channel) then instantly turn if off et voilą! 4:4:4 mVR goodness

I'm now annoying their tech support to have its firmware fixed("oh my, it's going way over my head, nobody ever called us for this kind of problem" ), and they should take me seriously or I'll call their head office

It's a shame this pattern can't easily be located and will end up drowned into this gigantic thread within a few weeks/months
leeperry is offline   Reply With Quote
Old 13th August 2013, 10:44   #19836  |  Link
Qaq
AV heretic
 
Join Date: Nov 2009
Posts: 422
Quote:
Originally Posted by ThurstonX View Post
TV is an LG 42LD550.
LG must support 4:4:4 no problem. What you need is to set HDMI input to PC (RGB) and change black level to High (or Low, can't say for sure) for 0-255.
Qaq is offline   Reply With Quote
Old 13th August 2013, 10:56   #19837  |  Link
Raylan Givens
Registered User
 
Raylan Givens's Avatar
 
Join Date: Jul 2013
Location: Macedonia, Greece
Posts: 43
I have some mkv movies and when the upstream says "full range" then I get "limited range" on my PC monitor ( too bright, no deep black ). I guess this is a bug or something, right? MPC HC without MadVr does show the same clips with correct, full range no matter what the upstream says. Incidentally, when MadVr gets "limited range" it outputs the movie correctly in full range...

Also, is there a way to get MPCHC's native OSD? I find it truly nice whilst on the other hand, MadVr's could get a few improvements in design...
__________________
CPU: i5 4690 ~ M/B: Z97 G45 ~ GPU: GTX 960 ~ RAM: 8GB 1,6Ghz ~ SSD: Samsung 850 Pro + EVO Sound: ALC 1150 ~ Speakers: Z5500 ~ OS: Win 10 x64 ~ MPCHC x64
Raylan Givens is offline   Reply With Quote
Old 13th August 2013, 11:29   #19838  |  Link
Skankee
Registered User
 
Join Date: Dec 2012
Posts: 17
@6233638:

As already mentioned by some people, your linked test-pattern does not work if your TV does 4:2:0 subsampling (4:2:0 and 4:4:4 looks identical).

I personally dont like test-patterns that can show you wrong results, so i would thank you if you add this information to your (old) post.

(even if only very old TVs are affected, we dont know how many people with a 4:2:0 TV are reading this thread )
Skankee is offline   Reply With Quote
Old 13th August 2013, 16:37   #19839  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by Skankee View Post
As already mentioned by some people, your linked test-pattern does not work if your TV does 4:2:0 subsampling (4:2:0 and 4:4:4 looks identical).

I personally dont like test-patterns that can show you wrong results, so i would thank you if you add this information to your (old) post.
That's true - although, as far as I know, the only display that seems to apply to is a Pioneer Kuro. I don't know of anything else which processes chroma in 4:2:0, and I would think that most people that have a PC setup capable of running madVR will not be using a display old enough where that is going to be a problem.

Without having access to a display which processes in 4:2:0, it's difficult for me to create a pattern which will test for that.

Can someone with a Kuro (or another display known to process in 4:2:0?) test this?


If your display is only capable of 4:2:0 resolution, you should not be able to see any lines in the left pattern, and I think the whole image will appear to be the same color.
With a 4:2:2 or 4:4:4 display, the pattern on the left should appear to be distinct red and blue alternating lines.

If your display is only capable of 4:2:2 resolution, the right pattern will appear to be noticeably darker than the center pattern, and seem to be a solid color. (example here)

If your display is capable of 4:4:4 resolution, the patterns in the center and on the right should appear to be roughly the same color. They may differ slightly because one is made from horizontal lines and the other is a checkerboard pattern.
6233638 is offline   Reply With Quote
Old 13th August 2013, 17:48   #19840  |  Link
ThurstonX
Registered User
 
Join Date: Mar 2006
Posts: 58
Quote:
Originally Posted by Qaq View Post
LG must support 4:4:4 no problem. What you need is to set HDMI input to PC (RGB) and change black level to High (or Low, can't say for sure) for 0-255.
Yep, and how quickly I forget things I've already posted! Must be gittin' old

Quite a few advanced features get lost when setting the input to 'PC'. May have to do some comparisons again, once the TV is free (oh when will that be? ;-)
ThurstonX 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 23:02.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.