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 21st October 2012, 17:34   #14941  |  Link
MasterMeNL
Registered User
 
Join Date: Jan 2012
Posts: 13
Quote:
Originally Posted by madshi View Post
Ouch. But Jinc3 AR is better, anyway. How much usage does your 680 have with Jinc3 AR?
Around 45-55%. Note that both upscaling algorithms are Jinc3 AR.
This is on the default clock (1111Mhz) on the default BIOS. I don't use the LN2 bios right now since it always has all power stages on = more power usage (no problem) = more heat (problem).
This card has a lot of OC potential but I don't need it yet.

I think the main reason that it uses so much is my 2560x1440 resolution. It's ~80% more than the average 1080p.

BTW: It uses 1750-1850 MiB VRAM, and I have desktop composition in full screen disabled. I want maxium quality for my videos.

Off-topic question: I use MPC-HC with the default subtitle renderer. However, it's slow so I set it to buffer 60 frames and disabled 'round power of two'. This sometimes causes flickering with certain complex animated subtitles.
But when I set it to render real time, it does lag the video when there is some really complex animation going on. Is there any way to fix this? CPU is an i7 3820, 24GB ram and it's around 5-10% when playing stuff. I refuse to use xy-VSFilter because it renders the subs at video resolution which is only 720p and I want sharp, crisp subs (a big difference between 720p subs and 1440p subs).
Is there any way to let it use more of my CPU so that it doesn't lag? I tried some multi-threaded version but results are the same.
MasterMeNL is offline   Reply With Quote
Old 21st October 2012, 17:53   #14942  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by madshi View Post
Are you talking about custom pixel shader support, similar to what MPC-HC supports for EVR? Or are you talking about a dedicated sharpening algorithm implemented in madVR, with comparable quality to good AviSynth sharpening scripts? Both might come some day, but custom pixel shader support will likely come first.
What is you stance on framerate multiplication? Or will it be possible with custom shaders?

Quote:
Originally Posted by madshi View Post
No, I'm not planning to create a 64bit version soon. Maybe some day, but not anytime soon. The reason? Because the possible benefit is not big enough compared to the development time it would cost me.
What is the source of the complexity? As I understand madVR works for the most part with the GPU which is separate from the AMD64 cpu architecture. (Just curious)
mzso is offline   Reply With Quote
Old 21st October 2012, 17:59   #14943  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by Gser View Post
win7 64bit, AMD 5870 driver 12.8

Sample 1
Sample 2
On a quick check EVR seems to behave identical? These files are not really native interlaced video content. Probably they were originally 50i, then converted to 60i by using field blending. With true native interlaced content, every interlaced field is from a different time, so there's motion between every field. With both of your samples, there are always 2 fields which are from the same time. So it's almost progressive. However, one of those fields has another field blended it. Basically the content is terrible. Please double check whether you get better results with EVR, but I don't think so.

Quote:
Originally Posted by MasterMeNL View Post
Off-topic question: I use MPC-HC with the default subtitle renderer. However, it's slow so I set it to buffer 60 frames and disabled 'round power of two'. This sometimes causes flickering with certain complex animated subtitles.
But when I set it to render real time, it does lag the video when there is some really complex animation going on. Is there any way to fix this? CPU is an i7 3820, 24GB ram and it's around 5-10% when playing stuff. I refuse to use xy-VSFilter because it renders the subs at video resolution which is only 720p and I want sharp, crisp subs (a big difference between 720p subs and 1440p subs).
Is there any way to let it use more of my CPU so that it doesn't lag? I tried some multi-threaded version but results are the same.
The xy-vsfilter developer, nevcairiel and I have worked on a new subtitle interface which will allow xy-vsfilter and madVR to work together in a similar way to the internal MPC-HC subtitle renderer - but with all media players. I think the new interface should avoid such problems as you're reporting, but we'll have to wait and see. Not sure when there will be an xy-vsfilter version available with support for the new interface. That depends on the xy-vsfilter developer.
madshi is offline   Reply With Quote
Old 21st October 2012, 18:01   #14944  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by mzso View Post
What is you stance on framerate multiplication? Or will it be possible with custom shaders?
It will not be possible with custom shaders. Not sure what you mean with "my stance". I don't plan that for madVR anytime soon, if that's what you mean.

Quote:
Originally Posted by mzso View Post
What is the source of the complexity? As I understand madVR works for the most part with the GPU which is separate from the AMD64 cpu architecture. (Just curious)
Pixel shaders are only a small part of madVR, unfortunately. I wish I could write pixel shaders all day instead of fighting with DirectShow, Direct3D, MSVC++ and all sorts of other things.
madshi is offline   Reply With Quote
Old 21st October 2012, 18:03   #14945  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by MasterMeNL View Post
Around 45-55%. Note that both upscaling algorithms are Jinc3 AR.
This is on the default clock (1111Mhz) on the default BIOS. I don't use the LN2 bios right now since it always has all power stages on = more power usage (no problem) = more heat (problem).
This card has a lot of OC potential but I don't need it yet.

I think the main reason that it uses so much is my 2560x1440 resolution. It's ~80% more than the average 1080p.

BTW: It uses 1750-1850 MiB VRAM, and I have desktop composition in full screen disabled. I want maxium quality for my videos.
Actually, I think the reason your numbers are so high, is because you turned the queues up as high as they could go. Reset to the defaults and you should see more reasonable numbers.
6233638 is offline   Reply With Quote
Old 21st October 2012, 18:18   #14946  |  Link
Gser
Registered User
 
Join Date: Apr 2008
Posts: 418
Quote:
Originally Posted by madshi View Post
On a quick check EVR seems to behave identical? These files are not really native interlaced video content. Probably they were originally 50i, then converted to 60i by using field blending. With true native interlaced content, every interlaced field is from a different time, so there's motion between every field. With both of your samples, there are always 2 fields which are from the same time. So it's almost progressive. However, one of those fields has another field blended it. Basically the content is terrible. Please double check whether you get better results with EVR, but I don't think so.
Yeah seems your right, it is completely unbelievable that a broadcast station would air and shoot their material like this.
Gser is offline   Reply With Quote
Old 21st October 2012, 18:28   #14947  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
High queues should *not* increase average GPU consumption. GPU consumption will be higher only while the queues are filling up. Once they're filled, GPU consumption should be identical to smaller queues.
madshi is offline   Reply With Quote
Old 21st October 2012, 18:52   #14948  |  Link
MasterMeNL
Registered User
 
Join Date: Jan 2012
Posts: 13
Quote:
Originally Posted by madshi View Post
The xy-vsfilter developer, nevcairiel and I have worked on a new subtitle interface which will allow xy-vsfilter and madVR to work together in a similar way to the internal MPC-HC subtitle renderer - but with all media players. I think the new interface should avoid such problems as you're reporting, but we'll have to wait and see. Not sure when there will be an xy-vsfilter version available with support for the new interface. That depends on the xy-vsfilter developer.
This is great news, if that gets implemented that my setup is perfect!
Quote:
Originally Posted by 6233638 View Post
Actually, I think the reason your numbers are so high, is because you turned the queues up as high as they could go. Reset to the defaults and you should see more reasonable numbers.
Quote:
Originally Posted by madshi View Post
High queues should *not* increase average GPU consumption. GPU consumption will be higher only while the queues are filling up. Once they're filled, GPU consumption should be identical to smaller queues.
Queues only increase the use of VRAM and RAM if I'm correct, so if something happens there are plenty of frames before it'll actually drop stuff so it has plenty of time to catch up.

Update: GPU usage is anywhere between 20%-60% on a 720p video with Jinc3 AR. I guess it depends on the video's complexity.

Off-topic question again: Is there any way for ReClock to just read the media speed instead of interrupting the audio when it's measuring stuff and applying it.

Last edited by MasterMeNL; 21st October 2012 at 19:26. Reason: Update Added
MasterMeNL is offline   Reply With Quote
Old 21st October 2012, 20:00   #14949  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
For what it's worth, I've just checked by downloading Kill Bill off iPlayer, opening it in madVR, and comparing that to the Blu-ray, and those files are also being wrongly detected as SMPTE-C. (colour is obviously wrong) It's an 832x468 25fps FLV.

I'd be surprised if there's anything below 720p at 25/50fps that is not supposed to be EBU/PAL. (or possibly BT.709)


Quote:
Originally Posted by madshi View Post
High queues should *not* increase average GPU consumption. GPU consumption will be higher only while the queues are filling up. Once they're filled, GPU consumption should be identical to smaller queues.
Sorry I wasn't clear with my post, I had meant that the reason VRAM usage was so high, was the high number of queues. (and usage definitely spikes until they are full)

Quote:
Originally Posted by MasterMeNL View Post
Off-topic question again: Is there any way for ReClock to just read the media speed instead of interrupting the audio when it's measuring stuff and applying it.
Can you clarify this? ReClock should not be "interrupting the audio".

Last edited by 6233638; 21st October 2012 at 20:07.
6233638 is offline   Reply With Quote
Old 21st October 2012, 20:30   #14950  |  Link
SamKook
Registered User
 
Join Date: Mar 2011
Posts: 216
Quote:
Originally Posted by 6233638 View Post
Can you clarify this? ReClock should not be "interrupting the audio".
When the framerate of the video get detected, reclock reinitialize or something similar and the audio cut for a fraction of a second.

I deactivated the portion of reclock that change the fps since it bothered me too much and only kept it for the WASAPI audio.

As far as I know, it's not possible to avoid this little interruption but it would be great if there was a way.
SamKook is offline   Reply With Quote
Old 21st October 2012, 21:23   #14951  |  Link
pankov
Registered User
 
Join Date: Mar 2002
Location: Sofia, Bulgaria
Posts: 661
Quote:
Originally Posted by madshi View Post
Hmmmmm... Can you try whether pressing F2 works? You know, F2 is available if you toggle some of the settings. F2 is handled identically in v0.84.0 vs. v0.83.7, so I'm wondering if that still works. Also, can you try changing the settings to use some very simple keyboard shortcuts? E.g. just "Ctrl+Something"? Does that also not work?
Yes, I've just tried it and F2 does work.
When I initially reported the problem I was using exactly such "simple" shortcuts - Ctrl+J and Ctrl+R to show and reset the OSD stats.
I was just about planning to add a new one (toggle the different deinterlacing states) when I discovered that none of the shortcuts worked with Girder.
btw
can you tell me more about how you hook the keyboard events?
which process is actually doing it? madHcCtrl.exe or the madVR.ax itself and which "window" inside of them? I'm asking this because Girder has the ability to "target" a specific window (found by window Name and/or ClassName and/or executable name) not only the foreground one and it's possible that there might be a solution that doesn't require you to change anything.
__________________
Z370M Pro4 | i3-8100 | 16GB RAM | 256GB SSD + 40TB NAS
NVIDIA GTX 1060 6GB (385.28) | LG OLED65B7V
Win 10 64bit 1803 + Zoom Player v14

Last edited by pankov; 21st October 2012 at 21:26.
pankov is offline   Reply With Quote
Old 21st October 2012, 21:36   #14952  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by MasterMeNL View Post
Update: GPU usage is anywhere between 20%-60% on a 720p video with Jinc3 AR. I guess it depends on the video's complexity.
GPU usage should not change much with different video files or scenes. It might change if it has to fill up the queue. So I guess those 60% is when the queue is empty.

Quote:
Originally Posted by pankov View Post
Yes, I've just tried it and F2 does work.
When I initially reported the problem I was using exactly such "simple" shortcuts - Ctrl+J and Ctrl+R to show and reset the OSD stats.
Can you please create a log, then press Ctrl+J, Ctrl+R with Girder?

Quote:
Originally Posted by pankov View Post
can you tell me more about how you hook the keyboard events?
which process is actually doing it? madHcCtrl.exe or the madVR.ax itself and which "window" inside of them? I'm asking this because Girder has the ability to "target" a specific window (found by window Name and/or ClassName and/or executable name) not only the foreground one and it's possible that there might be a solution that doesn't require you to change anything.
madVR.ax inside of the media player process is doing the hooking, by using SetWindowsHookEx(WH_KEYBOARD_LL).
madshi is offline   Reply With Quote
Old 21st October 2012, 21:38   #14953  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by SamKook View Post
When the framerate of the video get detected, reclock reinitialize or something similar and the audio cut for a fraction of a second.

I deactivated the portion of reclock that change the fps since it bothered me too much and only kept it for the WASAPI audio.

As far as I know, it's not possible to avoid this little interruption but it would be great if there was a way.
I can't say that I have experienced that, but ReClock is detecting the framerate on videos almost instantaneously for me. By the time I can open up the dialogue box, the framerate has been detected.

I have madVR set up with both the "delay playback start" options enabled though. I'm not sure if that will affect it, but I suggest trying that.

ReClock is set up like this for me:


I'm going straight from my PC to the display using HDMI though, and am not trying to bitstream. (bitstreaming is detrimental to audio playback on a PC)

It seems like this discussion belongs in the ReClock forums rather than madVR though.
6233638 is offline   Reply With Quote
Old 21st October 2012, 21:50   #14954  |  Link
pankov
Registered User
 
Join Date: Mar 2002
Location: Sofia, Bulgaria
Posts: 661
Quote:
Originally Posted by madshi View Post
Can you please create a log, then press Ctrl+J, Ctrl+R with Girder?

madVR.ax inside of the media player process is doing the hooking, by using SetWindowsHookEx(WH_KEYBOARD_LL).
Here is the log
http://www.mediafire.com/file/2j7403jel8zdxst/madVR_-_log_(Girder_-_keyboard).rar
I've opened it myself and found the problem
madVR gets Ctrl+Shift+J instead of Ctrl+J. This reminded me that I've configured a capital J in Girder. When I changed it to the small letter it started working. I guess this is not something you can fix on your end but you never stop to surprise me so I wait for your conclusion.
__________________
Z370M Pro4 | i3-8100 | 16GB RAM | 256GB SSD + 40TB NAS
NVIDIA GTX 1060 6GB (385.28) | LG OLED65B7V
Win 10 64bit 1803 + Zoom Player v14
pankov is offline   Reply With Quote
Old 21st October 2012, 22:19   #14955  |  Link
SamKook
Registered User
 
Join Date: Mar 2011
Posts: 216
Quote:
Originally Posted by 6233638 View Post
I can't say that I have experienced that, but ReClock is detecting the framerate on videos almost instantaneously for me. By the time I can open up the dialogue box, the framerate has been detected.

I have madVR set up with both the "delay playback start" options enabled though. I'm not sure if that will affect it, but I suggest trying that.

ReClock is set up like this for me:


I'm going straight from my PC to the display using HDMI though, and am not trying to bitstream. (bitstreaming is detrimental to audio playback on a PC)

It seems like this discussion belongs in the ReClock forums rather than madVR though.
He did say it was off-topic(and this will be my last reply on the subject here). I tried your settings and I still get a delay of about 4 seconds before detection(unless it detects previously set framerate even though I cleared the DB, then it's instantaneous). If you display the icon in the tray, you can see when it gets detected since it will change color.

I also have both delay options enabled in madvr(it doesn't change a thing), my pc hooked to my tv via hdmi(dvi to hdmi to be exact) and not trying to bitstream the audio.

Since my TV only support 60Hz and most of the stuff I watch is 23.976, I never bothered asking in the reclock forum since it wouldn't change much for me anyway.
SamKook is offline   Reply With Quote
Old 22nd October 2012, 01:41   #14956  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by mzso View Post
framerate multiplication [..] will it be possible with custom shaders?
I believe Seb.26(an HLSL PS script coder from the HCFR forum) made very clear that frame interpolation wasn't possible in MPC because it wouldn't allow access to N-1 frames, and Casimir666 never found the time to implement it either.

Hopefully, madshi will polish the PS script capabilities with the usual excellence he has now used us to, allowing before/after scaling processing, automatic rules based on resolution/frame rate and N-1 frames access(which would allow all kinds of temporal manipulations)

BTW, chambolle made a script that cuts the picture in 6 zones so you can finetune their respective convergence in order to "fix" a misconverged projector(anything with 3 discrete panels is misconverged to some extent) for instance(HR also provides this feature). Really cool stuff when you have a semi-busted display! Seb.26 improved it by adding even more zones but I can't find it atm.

Needless to say that I'm eagerly looking forward PS scripts support in mVR(any rough ETA please? ), so I can do levels conversion/mirroring/hardcore deinterlacing/super resolution upscale and what-not in full blown 32fp \o/

yesgrey3 also repeatedly told me that post-processing should preferably be applied on a color corrected picture and not prior to gamut mapping, so PS scripts support would pretty much kill all birds with one stone

Quote:
Originally Posted by madshi View Post
Pixel shaders are only a small part of madVR, unfortunately. I wish I could write pixel shaders all day
PS scripts coders definitely seem to really enjoy writing them: they get all the good sides of coding without being nagged by a million of other annoying things that essentially end up killing the fun

Quote:
Originally Posted by madshi View Post
There's no real benefit over LAV at the moment.
OK, thank you for the detailed explanation. I thought it might have had to do with jitter and so, but CUVID LAV + mVR truly make up for a stunning duo! And processing jitter doesn't seem to be as much of an issue with video as it can be with audio.

PS: bah, now I crave a new GPU so I can use jinc3 AR for 720p to 1080p luma upscale...my 96SP 8800GS only allows jinc3 AR for chroma and lanczos/spline AR for luma. This said, lanczos3 AR looks quite good and maybe you'll manage to optimize jinc3 AR for luma at some point

Last edited by leeperry; 22nd October 2012 at 04:44.
leeperry is offline   Reply With Quote
Old 22nd October 2012, 05:07   #14957  |  Link
jmone
Registered User
 
Join Date: Dec 2007
Posts: 652
Quote:
Originally Posted by madshi View Post
If you compare the data, the 660 is clearly on another level:

550Ti: 28.8 GT/s, 98.5 GB/s, 691 GFLOPS
660: 78.5 GT/s, 144.2 GB/s, 1881 GFLOPS
Great - got a Asus ENGTX660-DC2-2GD5 GF GTX660 DirectCU II PCI-E 3.0, 2GB on order (hope it gets here for the weekend) and will repost the 60i results.
jmone is offline   Reply With Quote
Old 22nd October 2012, 06:18   #14958  |  Link
cyberbeing
Broadband Junkie
 
Join Date: Oct 2005
Posts: 1,859
@SamKook

Reclock's DirectShow Frame-rate detection is extremely fragile (it uses a special source filter to automatically build a temporary Directshow graph at the beginning of playback). Certain combinations of splitters, decoders, and transform filters will cause it to stop working. When Reclock's DirectShow Frame-rate detection breaks, it falls back to Built-in Estimator which takes a few seconds of playback to determine the frame-rate. It would be nice if James @ Slysoft would fix this, but it seems he's finished doing improvements to Reclock.

If you set a default frame-rate like '23.976' instead of 'unknown' in Reclock, you should be able to prevent an initialization pop/gap when the estimator and your default frame-rate match. Depending on your audio hardware, if you set Reclock's 'Sound pre-buffer' low enough, you may also be able to reduce or eliminate the initialization pop/gap effect.

Last edited by cyberbeing; 22nd October 2012 at 06:25.
cyberbeing is offline   Reply With Quote
Old 22nd October 2012, 06:59   #14959  |  Link
MasterMeNL
Registered User
 
Join Date: Jan 2012
Posts: 13
Quote:
Originally Posted by 6233638 View Post
{snip}Post about ReClock{/snip}
Quote:
Originally Posted by cyberbeing View Post
{snip}Post about ReClock{/snip}
I took over your settings except for the buffer and a few other options and assuming a 23.976 frame rate did it.
It assumes 23.976 and after a few seconds it has measured the frame rate which is 99% of the time 23.976 fps so nothing changes. Thanks a lot for your info.

I'm really surprised by this forum, it's very active and a lot of helpful people in here. I've been following this for a long time but I only started posting yesterday. Great stuff!
MasterMeNL is offline   Reply With Quote
Old 22nd October 2012, 07:00   #14960  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by pankov View Post
madVR gets Ctrl+Shift+J instead of Ctrl+J. This reminded me that I've configured a capital J in Girder. When I changed it to the small letter it started working. I guess this is not something you can fix on your end but you never stop to surprise me so I wait for your conclusion.
I could "fix" it, but I don't think I should. After all there's a difference between Ctrl+J and Ctrl+Shift+J, and we want to have the option to use these two key combinations for different things, don't we?

Quote:
Originally Posted by leeperry View Post
Hopefully, madshi will polish the PS script capabilities with the usual excellence he has now used us to, allowing before/after scaling processing, automatic rules based on resolution/frame rate and N-1 frames access(which would allow all kinds of temporal manipulations)
In the first step I'll probably just duplicate what MPC-HC offers. A more complete solution might come at a later date. Similar to how madVR's screenshot functionality is currently very simplistic and will get more features sometime in the future.

Quote:
Originally Posted by leeperry View Post
yesgrey3 also repeatedly told me that post-processing should preferably be applied on a color corrected picture and not prior to gamut mapping
I don't think I agree with that.

Quote:
Originally Posted by leeperry View Post
This said, lanczos3 AR looks quite good and maybe you'll manage to optimize jinc3 AR for luma at some point
I wouldn't expect speed improvements for Jinc3 AR. It might even get a little slower, after some more tweaks to the AR algorithm.
madshi 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 02:07.


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