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 13th May 2015, 13:56   #29841  |  Link
aufkrawall
Registered User
 
Join Date: Dec 2011
Posts: 1,716
Quote:
Originally Posted by huhn View Post
added in the post
i would say catmull-rom is better for this source.
but this isn't a perfect x4 upsacale so it is used for image scaling too. simply not a perfect comparison.
Hard to say what is caused by luma or chroma upscaling.
But with C-R, the blue robe of "Mangust" is clearly more aliased than with Jinc.
aufkrawall is offline   Reply With Quote
Old 13th May 2015, 14:19   #29842  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 5,984
Quote:
Originally Posted by aufkrawall View Post
Hard to say what is caused by luma or chroma upscaling.
But with C-R, the blue robe of "Mangust" is clearly more aliased than with Jinc.
yes agree. but look at the icon on the right jinc3ar is blurring them like crazy. for example look at the person in the red robe with the "sparkles" on top of it catmull-rom is a lot sharper there and shows more "details" for more aliasing. the icon in the top right with the blue dots is clearly better with catmull-rom ar.

this is not even the same frame i try to fix this.

Last edited by huhn; 13th May 2015 at 14:39.
huhn is offline   Reply With Quote
Old 13th May 2015, 15:00   #29843  |  Link
Werewolfy
Registered User
 
Join Date: Feb 2013
Posts: 135
Well in this picture catmull-rom ar is even better than Jinc... Look at the picture on top of the men around the fire. The blue part is blurry with Jinc and Nnedi3.

Edit : Ok I didn't notice you edited your post
__________________
Windows 8.1 x64 - Intel Core i5-4670K (4.2 GHz) - 8 GB DDR3 - MSI Geforce GTX 1080 8 GB - Sony KD-55A1

Last edited by Werewolfy; 13th May 2015 at 15:04.
Werewolfy is offline   Reply With Quote
Old 13th May 2015, 15:15   #29844  |  Link
kalston
Registered User
 
Join Date: May 2011
Posts: 164
Quote:
Originally Posted by mark0077 View Post
madshi, this is a question, not specifically related to the newer versions of madVR so absolutely no priority in a reply to this (FYI using the latest 0.88.5 version)

I just got a GTX 980 and have it overclocked as much as I can push it, so I have been experimenting with NNEDI 32 neurons for image doubling / quadrupling. Quadrupling seems to be working fine at 32 neurons on very low resolution files. I'm now testing of a movie which is 1280 x 720, so to get to 1920 x 1080 I can see the following is being used in Ctrl-J stats

chroma > Jinc3 AR
luma > Nnedi32 < Catmull-Rom AR
chroma > Catmull-Rom AR

This seems to be alot more demanding than the quadrupling of the lower resolution files, which is fine, so in gpu-z I can see GPU Load hitting ~90% but my question is.

I don't see any dropped frames, delayed frames or presentation glitches, however the queues are as follows

decoder queue 15-16 / 16
upload queue 7-8 / 8
render queue 1-2 / 8
present queue 0-2 / 8

My question is, although I know my gpu is literally at its absolute limits, how come some of the queues seem full, while others are nearly empty all the time, yet my gpu isn't under enough pressure to drop frames. Absolutely very low priority question, just thought it was worth mentioning as I'm not sure if the render and present queues not being full in this scenario may be something madVR is controlling and could potentially be improved to have them more full seeing as frames aren't being dropped.
Those stats blow, do you have a gsync monitor maybe? I get stats like that (including the GPU usage) only when I go FSE with gsync enabled (it's a bug FYI but I think it's nvidia's and not madvr's fault - by using less demanding settings you can fix the queues and get smooth playback). Otherwise I do NNEDI13 128 or 256 easily for 720p to 1440p on a gtx 970 (assuming 24/25 fps content - might have to go down to 32 for 60fps for example).

Last edited by kalston; 13th May 2015 at 15:19.
kalston is offline   Reply With Quote
Old 13th May 2015, 15:16   #29845  |  Link
cca
Anime Otaku
 
Join Date: Oct 2002
Location: Somewhere in Cyberspace...
Posts: 437
Quote:
Originally Posted by madshi View Post
Stupid me. I forgot that the refresh rate fix also needs new madHcNetXx.dlls. You'll find them here:

http://madshi.net/madVR885refreshRateFix2.zip

These files together with the modified madVR(64).ax should make the refresh rate work again.
Yay, it works perfectly now!
__________________
AMD FX8350 on Gigabyte GA-970A-D3 / 8192 MB DDR3-1600 SDRAM / AMD R9 285 with Catalyst 1.5.9.1/ Asus Xonar D2X / Windows 10 pro 64bit
cca is offline   Reply With Quote
Old 13th May 2015, 15:44   #29846  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by huhn View Post
ok quadrupling is not that huge sorry but placebo? not sure.chroma doubling is quiet something!

chroma upscale
jinc3AR: http://abload.de/img/nnedi3jinc3arx2syl.png
nnedix2: http://abload.de/img/nnedi3chromax2posri.png
nnedix4: http://abload.de/img/nnedi3chromax424swl.png
editv2: catmull-rom AR http://abload.de/img/nnedi3catv2misgl.png

and with some more jinc3ar scaling on top:
jinc3AR: http://abload.de/img/nnedi3jinc3arg2sj3.png
nnedix2: http://abload.de/img/nnedi3chromax2rtsql.png
nnedix4: http://abload.de/img/nnedi3chromax4z7ssj.png

i wouldn't be shock if scaling to UHD would show some more potential for quadrupling on this source. but this source is just ...

edit2:added the right frame
Ok, Jinc has less aliasing but is also softer than Catmull-Rom. At least we can see that there *is* a difference. However, there's zero difference I can see between enabling and disabling chroma quadrupling. Which may suggest that only the first chroma upsampling operation is important. For 4:2:0 that would be 4:2:0 -> 4:4:4 upsampling. After that, probably chroma doubling would show no visible differences.

So I'm somewhat torn now: With this PC game, chroma doubling helps, and also Jinc looks "different" to Catmull-Rom. So maybe I should put both chroma doubling and Jinc back in? But with a 4:2:0 source maybe there's no difference at all. However, "upscaling refinement" could result in sharpening the chroma channel so much that doubling after refinement might make a difference, once again. So maybe that's a good argument for putting it all back in? Opinions?

Quote:
Originally Posted by Dogway View Post
I have a 60fps video and I don't want madvr to do anything if source framerate is not 24p.
edit: And to answer before you ask, no, I don't want to add 1080p60 since that switches to 60p screen mode, and I want to play the file at "1080 @ 60Hz" screen mode (aka in PC mode).
What is the difference between 60p screen mode and "1080 @ 60Hz PC mode"?

Quote:
Originally Posted by detmek View Post
No custom flush settings. I tried with defaults, disabling FSE, changing queues, scaling algorithms... Problem remains. Its worse if I activate Smooth Motion and disable FSE and its a bit worse on Windows 7. Then it stutters the all the time in full screen windowed mode. In FSE it stops after few seconds because presentation queue goes full after few seconds.

Another thing, presetation gliches do not rise always. Sometimes those do not change even if video stutters with out of order frames.
Ok, I have a guess at what might be wrong. I'll create a test build for you or two to test my theory. I really wish I could reproduce it myself, but for some reason I can't...

Quote:
Originally Posted by pirlouy View Post
It is during playback. Each time I enter Fullscreen, I wait 2-3 seconds then press Ctrl+R to reset stats.
They are full. Then 1 frame dropped, and they are full just after.
I was testing on a monitor which does not have 59Hz, only 60 (even if it's 59,88 in reality). The video was 25Hz interlaced, with smooth motion. So indeed, 50 does not match 60, but with smooth motion, it should be ok. And I managed to have 0 dropped frame/glitche during these 2-3 minutes.
If I take a classic 23,976 Hz video, I have the same dropped frames/glitches ratio.
At the moment when the frame drop occurs, are the queues near empty or are they still full right at that moment? In any case, a log file might clear this up. With smooth motion enabled, there really should not be any frame drops, as long as your GPU is fast enough.
madshi is offline   Reply With Quote
Old 13th May 2015, 16:04   #29847  |  Link
Shiandow
Registered User
 
Join Date: Dec 2013
Posts: 752
Quote:
Originally Posted by madshi View Post
So I'm somewhat torn now: With this PC game, chroma doubling helps, and also Jinc looks "different" to Catmull-Rom. So maybe I should put both chroma doubling and Jinc back in? But with a 4:2:0 source maybe there's no difference at all. However, "upscaling refinement" could result in sharpening the chroma channel so much that doubling after refinement might make a difference, once again. So maybe that's a good argument for putting it all back in? Opinions?.
Not sure if it helps, but have you tried using SuperChromaRes for the extra chroma scaling? It should be quite good at making the chroma match the luma, and however many passes you need it's probably still going to be faster than NNEDI3. Of course there's not much point if the quality is worse, but I think it could work quite well.
Shiandow is offline   Reply With Quote
Old 13th May 2015, 16:07   #29848  |  Link
Werewolfy
Registered User
 
Join Date: Feb 2013
Posts: 135
Quote:
Originally Posted by madshi View Post
Ok, Jinc has less aliasing but is also softer than Catmull-Rom. At least we can see that there *is* a difference. However, there's zero difference I can see between enabling and disabling chroma quadrupling. Which may suggest that only the first chroma upsampling operation is important. For 4:2:0 that would be 4:2:0 -> 4:4:4 upsampling. After that, probably chroma doubling would show no visible differences.

So I'm somewhat torn now: With this PC game, chroma doubling helps, and also Jinc looks "different" to Catmull-Rom. So maybe I should put both chroma doubling and Jinc back in? But with a 4:2:0 source maybe there's no difference at all. However, "upscaling refinement" could result in sharpening the chroma channel so much that doubling after refinement might make a difference, once again. So maybe that's a good argument for putting it all back in? Opinions?
I think that if we have found a scenario where we can see the difference, there are possibly other scenarios. You should put back these options at least to judge if it's useful for upscaling refinement or anything else. To test your new version, it's better to have all the cards
__________________
Windows 8.1 x64 - Intel Core i5-4670K (4.2 GHz) - 8 GB DDR3 - MSI Geforce GTX 1080 8 GB - Sony KD-55A1
Werewolfy is offline   Reply With Quote
Old 13th May 2015, 16:08   #29849  |  Link
detmek
Registered User
 
Join Date: Aug 2009
Posts: 475
Quote:
Originally Posted by madshi View Post
Ok, I have a guess at what might be wrong. I'll create a test build for you or two to test my theory. I really wish I could reproduce it myself, but for some reason I can't...
Thanks madshi. I will test those builds when you post it.
__________________
Intel Pentium G3220, Asus H81M-K, 4GB DDR3@1600MHz (1333MHz working), LG W1934S, Windows 8.1 Pro x64
detmek is offline   Reply With Quote
Old 13th May 2015, 16:12   #29850  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 5,984
Quote:
Originally Posted by madshi View Post
Ok, Jinc has less aliasing but is also softer than Catmull-Rom. At least we can see that there *is* a difference. However, there's zero difference I can see between enabling and disabling chroma quadrupling. Which may suggest that only the first chroma upsampling operation is important. For 4:2:0 that would be 4:2:0 -> 4:4:4 upsampling. After that, probably chroma doubling would show no visible differences.

So I'm somewhat torn now: With this PC game, chroma doubling helps, and also Jinc looks "different" to Catmull-Rom. So maybe I should put both chroma doubling and Jinc back in? But with a 4:2:0 source maybe there's no difference at all. However, "upscaling refinement" could result in sharpening the chroma channel so much that doubling after refinement might make a difference, once again. So maybe that's a good argument for putting it all back in? Opinions?
i would say let it as it is and wait for more UHD screen.
upscaling SD to UHD should make the picture very soft with a default scaler and things like chroma doubling may help there a lot.
huhn is offline   Reply With Quote
Old 13th May 2015, 16:21   #29851  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by Shiandow View Post
Not sure if it helps, but have you tried using SuperChromaRes for the extra chroma scaling? It should be quite good at making the chroma match the luma, and however many passes you need it's probably still going to be faster than NNEDI3. Of course there's not much point if the quality is worse, but I think it could work quite well.
You're right, that would be a good option. However, the problem is how to make all this configurable for the user in a good way. That's the hard part. I'll have to think about it.
madshi is offline   Reply With Quote
Old 13th May 2015, 16:37   #29852  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
So here are 2 test builds.

@detmek:
http://madshi.net/madVR885Haswell.zip
Please try them all and let me know which fix the problem, if any. Thanks.

@everyone with the D3D11 minimize / Aero Peak problem:
http://madshi.net/madVR885minimize.zip
Does this one fix the problem?
madshi is offline   Reply With Quote
Old 13th May 2015, 16:43   #29853  |  Link
clsid
Registered User
 
Join Date: Feb 2005
Posts: 5,014
Yes, that fixes D3D11 minimize issues.

Suggestion: open the settings when double-clicking the tray icon.

Last edited by clsid; 13th May 2015 at 16:46.
clsid is offline   Reply With Quote
Old 13th May 2015, 16:43   #29854  |  Link
Dogway
Registered User
 
Join Date: Nov 2009
Posts: 1,009
Quote:
Originally Posted by madshi View Post
What is the difference between 60p screen mode and "1080 @ 60Hz PC mode"?
On a SmartTV "1080 @ 60Hz" (PC mode) is the screen mode you get when plugging your HTPC to the TV with HDMI interface and set the input as PC, this mode has some image/tv settings disabled and those that aren't are saved within this mode. "1080 24p" mode is the mode you get when you set 1080p24 in madvr, you have more TV image settings enabled and modifications are stored within this mode and (here comes the oddity) any other "p" mode, namely "1080 60p", "1080 50p", etc. What this means is that when I enable "LED Clear Motion" (a marketing word for black frame insertion) for "1080 24p" in order to deal with the implicit 24p motion blur issues, I will also get it for "1080 60p" mode which doesn't need it. That setting also dims screen a bit so I thought on playing non 24p content as "PC Mode", problem is I can't seem to do it once I write 1080p24 in madvr, it wants to turn my PC mode from where I'm launching my videos into some of the "p" modes.

I don't know if you have a smartTV so sorry if this sounds confusing to you.



PD: btw, on your chroma discussion IHMO you should use nnedi3 chroma where you use nnedi3 luma, otherwise you get horrible chroma mismatches on saturated borders.
Dogway is offline   Reply With Quote
Old 13th May 2015, 16:53   #29855  |  Link
pirlouy
_
 
Join Date: May 2008
Location: France
Posts: 634
Quote:
Originally Posted by madshi View Post
At the moment when the frame drop occurs, are the queues near empty or are they still full right at that moment? In any case, a log file might clear this up. With smooth motion enabled, there really should not be any frame drops, as long as your GPU is fast enough.
They seem full for me: (7-8), then frame dropped, then (0-8), and in 1ms (7-8) again, until next dropped frame.

Here is a log. Hope it will help. I double click the file which opens the player, go into fullscreen, reset stats (Ctrl+R), let the problems occurs (for 1 minute i guess), then pause, exit fullscreen, closes player.
http://dl.free.fr/jHUVQcG4y
pirlouy is offline   Reply With Quote
Old 13th May 2015, 16:54   #29856  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by clsid View Post
Suggestion: open the settings when double-clicking the tray icon.
Funny, I was thinking exactly the same a couple of days ago...

Quote:
Originally Posted by Dogway View Post
On a SmartTV "1080 @ 60Hz" (PC mode) is the screen mode you get when plugging your HTPC to the TV with HDMI interface and set the input as PC, this mode has some image/tv settings disabled and those that aren't are saved within this mode. "1080 24p" mode is the mode you get when you set 1080p24 in madvr, you have more TV image settings enabled and modifications are stored within this mode and (here comes the oddity) any other "p" mode, namely "1080 60p", "1080 50p", etc. What this means is that when I enable "LED Clear Motion" (a marketing word for black frame insertion) for "1080 24p" in order to deal with the implicit 24p motion blur issues, I will also get it for "1080 60p" mode which doesn't need it. That setting also dims screen a bit so I thought on playing non 24p content as "PC Mode", problem is I can't seem to do it once I write 1080p24 in madvr, it wants to turn my PC mode from where I'm launching my videos into some of the "p" modes.
Ok, but PC modes vs. non-PC mode sounds like a setting in your TV, not on your HTPC. So why does your TV switch into non-PC mode if you let madVR switch to 1080p60? And how do *you* manually switch your HTPC into a mode which turns on PC mode on your TV? Are there two different 1080p60 modes on your HTPC, one for TV PC mode and one not? And madVR switches into the wrong 1080p60 mode? That's the part I don't understand. In theory you should get exactly what you want by adding both "1080p24" and "1080p60" into the madVR display mode edit box.
madshi is offline   Reply With Quote
Old 13th May 2015, 16:54   #29857  |  Link
James Freeman
Registered User
 
Join Date: Sep 2013
Posts: 919
Quote:
Originally Posted by madshi View Post
So here are 2 test builds.

@everyone with the D3D11 minimize / Aero Peak problem:
http://madshi.net/madVR885minimize.zip
Does this one fix the problem?
YES it does!
Can we please try a 64 build?
__________________
System: i7 3770K, GTX660, Win7 64bit, Panasonic ST60, Dell U2410.

Last edited by James Freeman; 13th May 2015 at 17:02.
James Freeman is offline   Reply With Quote
Old 13th May 2015, 17:15   #29858  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by pirlouy View Post
They seem full for me: (7-8), then frame dropped, then (0-8), and in 1ms (7-8) again, until next dropped frame.

Here is a log. Hope it will help. I double click the file which opens the player, go into fullscreen, reset stats (Ctrl+R), let the problems occurs (for 1 minute i guess), then pause, exit fullscreen, closes player.
http://dl.free.fr/jHUVQcG4y
Ok, I can see that every 5-10 seconds there's a "gap" in the log where nothing happens for about 100-150ms. Then suddenly everything works again, but those 100-150ms are enough to run the queues dry, which eventually results in a frame drop then.

Do you have any sort of application/service running in the background which might be using/blocking the GPU? Try closing any such application/service, and maybe also your browser. Does that help in any way?

Can you please create another log with this build:

http://madshi.net/madVRextendedLog.zip

Quote:
Originally Posted by James Freeman View Post
YES it does!
Can we please try a 64 build?
Later.
madshi is offline   Reply With Quote
Old 13th May 2015, 17:15   #29859  |  Link
GCRaistlin
Registered User
 
GCRaistlin's Avatar
 
Join Date: Jun 2006
Posts: 262
madshi, can you please add the possibility to change the current device's refresh rate from the tray icon menu? Sometimes it's better to watch PAL videos at native 25 Hz, for example SopCast translations, sometimes not. It's easy to change playing speed in ReClock but not the refresh rate with madVR. So sometimes I have to uncheck "treat 25p movies as 24p" setting in madVR and then relauch the player - just to get the refresh rate I need now.
__________________
Magically yours
Raistlin
GCRaistlin is offline   Reply With Quote
Old 13th May 2015, 17:23   #29860  |  Link
detmek
Registered User
 
Join Date: Aug 2009
Posts: 475
Quote:
Originally Posted by madshi View Post
So here are 2 test builds.

@detmek:
http://madshi.net/madVR885Haswell.zip
Please try them all and let me know which fix the problem, if any. Thanks.
Hi. I tried all four versions and none of those fixes a problem. I used MPC-BE x86 Nightly for test.

I don't know if I mentioned before but I don't have dropped or delayed frames or glitches. Its just render queue is not full and frames are presented out of order.

The thing is, problem almost desapires if I turn of smooth motion or turn on FSE. It flickers just a 1-2 seconds. With FSE render queue goes full. And that is even with regular build. Without FSE and with Smooth Motion on it flickers all the time.
__________________
Intel Pentium G3220, Asus H81M-K, 4GB DDR3@1600MHz (1333MHz working), LG W1934S, Windows 8.1 Pro x64

Last edited by detmek; 13th May 2015 at 17:52.
detmek 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 22:37.


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