View Full Version : madVR - high quality video renderer (GPU assisted)
70MM
22nd March 2016, 03:54
You can reset madVR to its default settings by finding the madVR plugin folder in JRiver and running the "reset settings" file.
To improve the performance of smooth motion, try increasing the size of the GPU queue size in general settings. This will provide greater protection against these glitches.
For your refresh rate problem, madshi mentioned you can use madVR profile rules to create profiles for devices -> display modes. You could set your entire pre-show content to be played at 60 Hz and the feature at 24 Hz.
Select devices.
Click create profile group.
Add display modes to the profile group.
Create two profiles: name one folder something like "Pre-show" and the other "Feature."
if (fileName = "introduction.mkv") "Pre-show"
else if (fileName = "trivia.mkv") "Pre-show"
else if (fileName = "trailers.mkv") "Pre-show"
else "Feature"
Leave the display modes field blank under "Pre-show." Under "Feature," enter "1080p23."
One profile will play without refresh rate matching. The other will switch to 1080p24 when the feature starts.
Hi Warner, ok this all sounds very interesting but when it switches from the 60 content to the feature at 24 Im back to the same old problem with the SLOW SYNC of the JVC RS600 projector, arnt I?
The proj takes up to 17+ secs to sync between 60 > 24, it plays the audio but clips all the video until it finally hand shakes. You miss all the logos and intro of the feature, its horrible!
This was the whole reason to remain on 60 for all the files, trailers, shorts we play before the feature at our film club.
The 24 feature was then to remain also on 60 so everything transitions nice and smoothly for files and feature, nothing clipped or missing on the screen....
AngelGraves13
22nd March 2016, 06:41
you can't use that on chroma.
Hmmm...a limitation of chroma upscaling?
Magik Mark
22nd March 2016, 07:36
Madshi,
The author of Adaptive Sharpen recommends linear light when using this. Is it being followed?
Werewolfy
22nd March 2016, 09:56
Unfortunately the download link doesn't work.
Ok... So now I know what this website really worths ^^
Here's a new link for my videos about deinterlacing and my issues with it : http://www46.zippyshare.com/v/BiIrl8f9/file.html
I hope it will work this time. If it doesn't, please tell me what website I need to use.
Oh, did you mention that before? How fast is your CPU? Have you tried DXVA copyback decoding as an alternative to software decoding?
Are you using automatic black bar detection? If so, does disabling that workaround the issue?
I think you misunderstood me, I'm talking here only when I use deinterlacing and not in general.
I have a Core i5 4670k @ 4.0 Ghz. Disabling automatic black bar detection doen't change anything.
I tested in the past with DXVA decoding enabled but I still experienced dropped frames with interlaced content.
madshi
22nd March 2016, 11:13
will ripping movies with MakeMKV keep the subtitle positions or do I need to rip them to .iso or .m2ts with tsmuxer in order for MadVR to playback 3D subtitles properly?
The 3D depth information is partially lost. The MakeMKV devs are *VERY* sluggish to respond. See here:
http://www.makemkv.com/forum2/viewtopic.php?f=10&t=13892
Thanks for replying :) Aero was to blame, had not activated it.
Went from bad to worse though. When checking DX11 and playing a movie now now i get a black screen (file gets opened though, I can see the timer going off )
But you do have the Windows 7 Platform Update installed? And you don't get a picture, neither in windowed nor in fullscreen exclusive mode?
Also, has any further work been done on the debanding filter to target pixelation?
Pixelation? What pixelation? I've no idea what you're talking about.
The author of Adaptive Sharpen recommends linear light when using this. Is it being followed?
Where does he say that? Link?
Here's a new link for my videos about deinterlacing and my issues with it : http://www46.zippyshare.com/v/BiIrl8f9/file.html
I hope it will work this time. If it doesn't, please tell me what website I need to use.
That's a good host, thanks. As is e.g. mediafire.
I think you misunderstood me, I'm talking here only when I use deinterlacing and not in general.
I have a Core i5 4670k @ 4.0 Ghz. Disabling automatic black bar detection doen't change anything.
I tested in the past with DXVA decoding enabled but I still experienced dropped frames with interlaced content.
Does the decoder queue fill with the same video if you disable deinterlacing completely?
ryrynz
22nd March 2016, 11:24
Also, has any further work been done on the debanding filter to target pixelation?
No work has been done on debanding. I believe it's current state is good enough and is unlikely to receive much of the way of changes in future.. if anything.
Anyway, what you mentioned would normally fall under denoising, which madVR is likely to tackle at some point..
huhn
22nd March 2016, 11:35
Hmmm...a limitation of chroma upscaling?
there is no "light" that can be "changed" to linear light because it is chroma.
Ver Greeneyes
22nd March 2016, 11:38
Pixelation? What pixelation? I've no idea what you're talking about.
I think they mean using it for deblocking. I do think having a deblocking filter (maybe separately for chroma, which often seems to be very heavily compressed on things like Youtube videos) would be nice, but yeah.. different kind of filter.
madshi
22nd March 2016, 11:54
I think they mean using it for deblocking. I do think having a deblocking filter (maybe separately for chroma, which often seems to be very heavily compressed on things like Youtube videos) would be nice, but yeah.. different kind of filter.
Oh ok. I thought (s)he meant that debanding would *cause* pixelation.
ryrynz
22nd March 2016, 11:56
I think they mean using it for deblocking. I do think having a deblocking filter (maybe separately for chroma, which often seems to be very heavily compressed on things like Youtube videos) would be nice, but yeah.. different kind of filter.
Yeah makes sense. He'll be able to take care of that when madshi adds Vapoursynth support. Once that's done many often used filters could make their way into the options without hassle.. we'd basically have a mini ffdshow inside madVR.
Magik Mark
22nd March 2016, 14:00
Madshi
http://forum.doom9.org/showthread.php?p=1721881#post1721881
If we are talking about the same adaptive Sharpen. The author recommends gamma light as input.
I'm sorry my initial posts stated "linear light" its gamma light
madshi
22nd March 2016, 14:03
http://forum.doom9.org/showthread.php?p=1721881#post1721881
If we are talking about the same adaptive Sharpen. The author recommends gamma light as input.
I'm sorry my initial posts stated "linear light" its gamma light
I'm feeding gamma light to the shader.
XTrojan
22nd March 2016, 16:36
Anyone noticing some aliasing/artifacting with Super-Xbr 150 Doubling from 1080->2160p?
CarlosCaco
22nd March 2016, 17:13
Yeah makes sense. He'll be able to take care of that when madshi adds Vapoursynth support. Once that's done many often used filters could make their way into the options without hassle.. we'd basically have a mini ffdshow inside madVR.
Its the second time that i read someone talking about madshi adds vapoursynth support on madvr but i never read madshi talking about this... is this true? Madshi have this on mind to madvr future?
XMonarchY
22nd March 2016, 17:42
VapourSynth? As in Disappearing Features?
huhn
22nd March 2016, 17:52
Anyone noticing some aliasing/artifacting with Super-Xbr 150 Doubling from 1080->2160p?
ringing at lot of it.
dansrfe
22nd March 2016, 18:10
@madshi
Turning off exclusive mode and keeping to overlay prevents the OpenCL crash from happening. No frames are dropped so I think I'll keep it on overlay only.
aufkrawall
22nd March 2016, 18:55
Radeon R9 390 on Windows 10:
The queues don't seem to get filled nicely in fullscreen or borderless mode and so frames drop. When I keep an UI element of the player opened, playback is fine.
Is this a known problem? Used driver is Crimson 16.3.1.
baii
22nd March 2016, 19:54
Radeon R9 390 on Windows 10:
The queues don't seem to get filled nicely in fullscreen or borderless mode and so frames drop. When I keep an UI element of the player opened, playback is fine.
Is this a known problem? Used driver is Crimson 16.3.1.
Typical driver problem probably, usually power saving features? There is a power mode switch thing since 16.3
Sent from my 306SH
leeperry
22nd March 2016, 19:56
ringing at lot of it.
yeah with my current settings sxbr definitely doesn't work anymore due to excessive ringing so I stick to NNEDI3 for lores and NEDI for 720p@1080p, hopefully next build with the improved AR will be a game changerhttp://forum-images.hardware.fr/images/perso/zigg.gif
vapoursynth support on madvr but i never read madshi talking about this... is this true?
http://www.svp-team.com/forum/viewtopic.php?pid=53211#p53211
what is "EE"?
Edge Enhancement (http://forum.doom9.org/showpost.php?p=1753284&postcount=35338)
aufkrawall
22nd March 2016, 20:14
Typical driver problem probably, usually power saving features? There is a power mode switch thing since 16.3
I'm already using that switch since frametime quality in games else suffers.
However, I found out what was causing the madVR problem for me:
MSI Afterburner, or to be more precise: a certain setting of it.
I disabled low level driver access, since I thought it would help to prevent compatibility issues. But the opposite is true: After enabling low level driver access with user-mode driver, the queue issue disappeared. :cool:
They get nicely filled now.
The R9 390 definitely offers a lot of performance for NNEDI3, it seems to be easily competing with a GTX 980 OC.
It seems my 390 is stable with -100mV undervoltage, which keeps it very quiet, combined with a custom fan curve.
Seems like a good deal if much NNEDI3 and SuperRes bang is needed for few bucks.
Edit: Got a MSI R9 390 Gaming.
Warner306
22nd March 2016, 21:27
Anyone noticing some aliasing/artifacting with Super-Xbr 150 Doubling from 1080->2160p?
super-xbr100 + AR with SuperRes doesn't ring a whole lot. I used to use a setting of 150, but the extra detail seemed less important than the added ringing. super-xbr is already very sharp at 100.
aufkrawall
22nd March 2016, 21:42
With cartoons, the ringing of super-xbr 100 is very obvious and SuperRes makes it even worse. However, this might change when madshi introduces Hylian's improved anti-ringing filter.
ryrynz
22nd March 2016, 22:15
Its the second time that i read someone talking about madshi adds vapoursynth support on madvr but i never read madshi talking about this... is this true? Madshi have this on mind to madvr future?
madshi has expressed interest in doing so, I expect it'll be a nearer to 1.0 feature. He doesn't tend to disclose his feature work list.. but it does make a lot of sense to add support for Avisynth/Vapoursynth plugins considering where madVR positions itself in the market.
VapourSynth? As in Disappearing Features?
A whole boat load of features..
cork_OS
22nd March 2016, 22:33
@madshi
Hello.
According to this thread, madVR seems to be generally used for rendering modern high-bitrate HD video, including 4k & 3D.
Instead, I'm watching a lot of old MPEG-2 & H.263 encoded video. Such video often suffers from blocking, ringing/mosquito noise, aliasing or improper deinterlace.
madVR already contain some filters to improve such video (SuperRes, sharpen filters), but it would be great to see even more video processing and enhancing options in madVR (EDIT: including AVS/VS plugins support, as alternative).
Since improving old videos isn't main madVR goal, could you please tell, are adding deblocking/de-mosquitoing/anti-aliasing filters etc. possible/planned at all?
Thanks for your work!
ryrynz
22nd March 2016, 22:55
madVR already contain some filters to improve such video (SuperRes, sharpen filters), but it would be great to see even more video processing and enhancing options in madVR (EDIT: including AVS/VS plugins support, as alternative).
I just covered this a few posts back at the top of the page. At this point you can use Avisynth and ffdshow to achieve what you want, so either read about about how to set that up or sit back and wait.
bluechris
22nd March 2016, 23:32
Do you have them always, or only sometimes? Always right from the start of the movie? Or does it start after some minutes of playback?
Sorry for the late reply but yes it happens always.
Anima123
23rd March 2016, 00:06
@madshi
Hello.
According to this thread, madVR seems to be generally used for rendering modern high-bitrate HD video, including 4k & 3D.
Instead, I'm watching a lot of old MPEG-2 & H.263 encoded video. Such video often suffers from blocking, ringing/mosquito noise, aliasing or improper deinterlace.
madVR already contain some filters to improve such video (SuperRes, sharpen filters), but it would be great to see even more video processing and enhancing options in madVR (EDIT: including AVS/VS plugins support, as alternative).
Since improving old videos isn't main madVR goal, could you please tell, are adding deblocking/de-mosquitoing/anti-aliasing filters etc. possible/planned at all?
Thanks for your work!
I would like to second that.
There's one thing that I'd like to cite, although technologies has been advanced to HD-video, 3D, H265 etc., a lot of my favored contents are still SD, with a lot of such artefacts as mentioned by cork_OS.
It really would be very zealous features for us, if madshi can considering adding, at least some of these, post-processing features to madVR, in an appropriate time.
DragonQ
23rd March 2016, 00:24
I would like to second that.
There's one thing that I'd like to cite, although technologies has been advanced to HD-video, 3D, H265 etc., a lot of my favored contents are still SD, with a lot of such artefacts as mentioned by cork_OS.
It really would be very zealous features for us, if madshi can considering adding, at least some of these, post-processing features to madVR, in an appropriate time.
Heh, a few of my favourite comedies are essentially 288p/25 because they were poorly converted from 576i/25 in post-production. I often wonder how much better they'd look if I manually stripped out every other line from the source and reconstructed them with an algorithm like NNEDI3.
Mano
23rd March 2016, 03:26
what is the latest stable AMD driver for madvr? is it still 13.2?
ryrynz
23rd March 2016, 05:12
what is the latest stable AMD driver for madvr? is it still 13.2?
AFAIA they're all stable as long as you're not deinterlacing. I've seen most recommend 13.12 still, It takes a few minutes to test so why not just update to the latest crimson
and if there's any issues then just post about them? Can't have people asking about what works every time a new driver is released.
Warner306
23rd March 2016, 06:14
@madshi
Hello.
According to this thread, madVR seems to be generally used for rendering modern high-bitrate HD video, including 4k & 3D.
Instead, I'm watching a lot of old MPEG-2 & H.263 encoded video. Such video often suffers from blocking, ringing/mosquito noise, aliasing or improper deinterlace.
madVR already contain some filters to improve such video (SuperRes, sharpen filters), but it would be great to see even more video processing and enhancing options in madVR (EDIT: including AVS/VS plugins support, as alternative).
Since improving old videos isn't main madVR goal, could you please tell, are adding deblocking/de-mosquitoing/anti-aliasing filters etc. possible/planned at all?
Thanks for your work!
I don't think I'd be interested in any denoise/deblock filters. However, the paper below does a good job of outlining the various artifacts found in low-quality video (for those who are curious):
Characterizing Perceptual Artifacts in Compressed Video Streams (https://ece.uwaterloo.ca/~z70wang/publications/HVEI14.pdf)
ryrynz
23rd March 2016, 06:30
I don't think I'd be interested in any denoise/deblock filters. [/URL]
I think you'd be surprised how much macroblocking there is in video. Particularly busy areas can have quite extreme artifacting that you wouldn't realize is there until you pause the video.
XMonarchY
23rd March 2016, 11:11
A bit off-topic, but is Rec. 2020 the official standard for UHD 4K-8K standard? Some say its DCI P3, but I think they are misinformed.
madshi
23rd March 2016, 11:25
Turning off exclusive mode and keeping to overlay prevents the OpenCL crash from happening. No frames are dropped so I think I'll keep it on overlay only.
Didn't you plan to analyze which exact madVR build introduced the problem?
According to this thread, madVR seems to be generally used for rendering modern high-bitrate HD video, including 4k & 3D.
Instead, I'm watching a lot of old MPEG-2 & H.263 encoded video. Such video often suffers from blocking, ringing/mosquito noise, aliasing or improper deinterlace.
madVR already contain some filters to improve such video (SuperRes, sharpen filters), but it would be great to see even more video processing and enhancing options in madVR (EDIT: including AVS/VS plugins support, as alternative).
Since improving old videos isn't main madVR goal, could you please tell, are adding deblocking/de-mosquitoing/anti-aliasing filters etc. possible/planned at all?
I don't really like to comment on future plans much. But for version 1.0 I do hope to have all important areas covered.
A bit off-topic, but is Rec. 2020 the official standard for UHD 4K-8K standard? Some say its DCI P3, but I think they are misinformed.
Which standard? Broadcasting has a different standard than UHD Blu-Ray. UHD Blu-Ray allows using a BT.2020 container, and the studios can then use metadata to signal which exact subset of the container is used. This is very often DCI-P3. But the studio can use up to BT.2020, if they want to.
madshi
23rd March 2016, 11:41
madVR v0.90.16 released
http://madshi.net/madVR.zip
* added adjustable anti-bloating filter for sharpening
* added adjustable anti-bloating filter for SSIM downscaling
* added adjustable anti-bloating filter for SuperRes
* added anti-ringing filter for SuperRes
* added Hyllian's super-xbr anti-ringing (for image doubling)
* reworked super-xbr chroma upscaling anti-ringing algorithms
* improved sharpen anti-ringing filter
* sharpen anti-ringing filter now also includes AdaptiveSharpen
* changed SSIM anti-ringing method
* added "relaxed" vs "strict (soft)" option for downscale anti-ringing
* "crispen edges" now tries to avoid amplifying grain & noise
* reduced some "crispen edges" aliasing problems
* added separate image doubling supersampling option
* added trade quality option "scale chroma separately if it saves performance"
* increased AdaptiveSharpen max value from 1.5 to 3.0
* increased "thin edges" max value from 4.0 to 8.0
* previously optional DX11 "alternative glitch handling mode" is now always on
* multiple profile groups can now share the same keyboard shortcut
* double clicking tray icon now opens "best" settings instead of always local
* fixed: Jinc/SSIM 2D downscaling was sometimes activated when not needed
* fixed: 4-taps Spline was broken
* fixed: some problems with 3D side-by-side and line/column alternative
* fixed: 10bit content could make DXVA scaling fail
* fixed: bitmap subtitles were not always moved up when cropping black bars
* fixed: when downscaling, some chroma algos were used although they shouldn't
* fixed: sometimes media player froze during DVD playback
* fixed: source black/white adjustments + gamma processing: wrong order
* fixed: rare shutdown crash when using native DXVA decoding
* fixed: OSD bug when playing BT.709 HDR content
Lots and lots of new stuff for you guys to play with. Some notes:
1) Many of the new features/options are considered experimental. I want/need your feedback, and your feedback will decide which options will stay, which will be always forced on, and which will be removed again (soon). So please do provide feedback, YOU decide what stays and what goes. For a list of important feedback areas see the bottom of this post.
2) The new anti-bloating tries to concentrate sharpening on higher frequencies and to remove lower frequencies. I like the look that produces, but your mileage may vary. The anti-bloating filter is only active for "sharpen edges", Luma Sharpen and Adaptive Sharpen (and SSIM and SuperRes), but it's not active for "crispen edges" and "thin edges", because those don't bloat by design.
3) super-xbr chroma upscaling previously had stronger anti-ringing with the "anti-ringing" option unchecked than with it checked. That was kinda confusing. So I've changed that now. You'll get some amount of anti-ringing with the option checked and unchecked, but the AR effect is now bigger with the option checked, which makes more sense, I suppose. The anti-ringing effect was also improved a bit.
4) If you didn't like "crispen edges" (formerly called "FineSharp") until now, please give it another try now. It doesn't amplify noise and grain as much as it did before. My personal opinion is still that it's the sharpener which produces the most natural and high-def look. Although the new anti-bloating filter might bring LumaSharpen and AdaptiveSharpen nearer to "crispen edges" now, in terms of naturalness and high-def look.
5) For downscaling, the anti-ringing filter now has 2 methods. The "relaxed" one allows a little bit of ringing to remain, which produces sharper results. The "strict (soft)" option removes all ringing, but the final result is quite soft (of course SSIM helps a lot there). One anti-ringing sub-option will stay, one will go. You decide.
-------
Here are the most important areas of FEEDBACK I need:
a) Do you consider the "anti-bloating" filter useful for sharpening (LumaSharpen + AdaptiveSharpen)? If so, which strength settings do you like?
b) Do you consider the anti-bloating and anti-ringing options useful for SuperRes? I think anti-bloating might not be beneficial here, but I'd like to hear your opinion. Do you see any benefit in the anti-ringing option? SuperRes already didn't ring that much, so I wonder if the separate anti-ringing post processing option is worthwhile in your opinion or not?
c) Do you consider the "anti-bloating" filter useful for SSIM downscaling? If so, which strength settings do you like?
d) Which of the two anti-ringing options do you prefer for downscaling? (Edit: FYI, these options do *not* affect SSIM downscaling)
Again: Your feedback is very important. Thank you!
Mano
23rd March 2016, 11:48
AFAIA they're all stable as long as you're not deinterlacing. I've seen most recommend 13.12 still, It takes a few minutes to test so why not just update to the latest crimson
and if there's any issues then just post about them? Can't have people asking about what works every time a new driver is released.
Because i am not sure a few minutes is sufficient to test the new driver and frankly it is kind of hard to search a few hundred pages manually (search function just doesn't work for posts inside thread).
sneaker_ger
23rd March 2016, 11:50
You can search the thread. Look for the button on the top right. (right below the page numbers)
ryrynz
23rd March 2016, 11:55
madVR v0.90.16 released
Quite the change log.
I've found that changing the anti-ringing filter type for super-xbr doesn't prompt me to Apply after I've already changed it once, switching between them Apply is still greyed out.
Also it's not remembering strict once I Apply and judging from my screenshots it's not using Strict either when it does apply.
Ver Greeneyes
23rd March 2016, 11:58
Wow, that's a lot of new stuff. I'll start off by reporting a bug (sorry): I can't set the anti-ringing filter to "strict (soft)" for chroma upscaling or image upscaling. If I do, the Apply button doesn't light up, and even if I force it to light up by changing something else (like toggling superres), the change to anti-ringing doesn't stick.
madshi
23rd March 2016, 12:08
Ok, the bug is that the anti-ringing type *should* only be available/visible for image downscaling. It's not supposed to be visible at all for chroma upscaling or image upscaling. So don't worry, this is just a cosmetical bug in the settings dialog. Just ignore that option for chroma/image upscaling. There's only one AR algo available for upscaling. Only downscaling knows 2 different types.
ryrynz
23rd March 2016, 12:12
Okay. FWIW, super-xbr now looks good with anti-ringing enabled, similar to what it did when it was off in .15.
However anti-ringing off now in .16 isn't looking as good as it was in .15 so looks like super-xbr for chroma needs anti-ringing enabled now for best results.
So it's exactly as you said in the above post, the difference between .15 off and .16 on is pretty minor but nice to have small improvements anyway.
Super-xbr 100 AR is my preferred chroma upscaler.
madshi
23rd March 2016, 12:18
Okay. FWIW, super-xbr now looks good with anti-ringing enabled, similar to what it did when it was off in .15. However anti-ringing off now in .16 isn't looking as good as it was in .15 so looks like super-xbr for chroma needs anti-ringing enabled now.
Yes, as I mentioned in the release notes, I've swapped the meaning of the anti-ringing switch for "super-xbr for chroma", because it makes more sense the way it is now. But, with the option checked it should be slightly better now than it was with the option unchecked in .15! Try the "speech bubble" (I think you know what I mean).
sheppaul
23rd March 2016, 12:46
It is really nice to see your active developments but madVR seems to need a setup wizard for end [normal] users before reaching v1.0.
I know you're doing your best to make it simple but..... :p
aufkrawall
23rd March 2016, 13:18
Funny, with .16 I'm experiencing the same problem as described here with .15 + Afterburner:
http://forum.doom9.org/showthread.php?p=1761687#post1761687
But this time, it also happens without Afterburner. Should I provide debug logs? I can't really use the new version with this behavior.
GCRaistlin
23rd March 2016, 13:32
* fixed: sometimes media player froze during DVD playback
Unfortunately not: http://pastebin.com/LRxRi0Mu
GCRaistlin
23rd March 2016, 13:55
What is the minimal hardware requirements to play 2160p downscaled to 1080p? On HD5770 with software decoding, I have slow motioned video while audio is being played normally (so audio runs far ahead). If "decoder queue" value is "1-3/16" does it mean that the problem is on decode (LAV Video) side rather that on madVR side?
Werewolfy
23rd March 2016, 14:07
Thanks for the new version! I will try to test the new features this week end.
Does the decoder queue fill with the same video if you disable deinterlacing completely?
Actually, no it's the same behavior BUT I found out if I desactivate Reclock's Pal speedown (24 fps) and play it at 50hz, the queues still don't fill but I have no frampes dropped. So is it only reclock's fault?
I also tested a Blu-Ray with interlaced content and it plays fine, the queues completely filled. So it seems limited to DVD...
Let's me summarize :
-DVD interlaced with Reclock's Pal speedown (24hz) : queues don't fill, frames dropped when render queue drops down.
-DVD interlaced without Reclock's Pal speedown (50hz) : queues don't fill, no frames dropped.
-Blu-Ray interlaced and progressive : queues fill, no frames dropped.
-Other videos : queues fill, no frames dropped.
Do you have an idea to improve the situation with DVD? I can do more tests and videos, juste tell me what to do.
sneaker_ger
23rd March 2016, 14:11
What is the minimal hardware requirements to play 2160p downscaled to 1080p? On HD5770 with software decoding, I have slow motioned video while audio is being played normally (so audio runs far ahead). If "decoder queue" value is "1-3/16" does it mean that the problem is on decode (LAV Video) side rather that on madVR side?
Yes, the decoding is too slow. (Try LAV 64 bit if you haven't already, it's much faster for HEVC software decoding than the 32 bit version.)
madshi
23rd March 2016, 14:13
So it's exactly as you said in the above post, the difference between .15 off and .16 on is pretty minor but nice to have small improvements anyway.
Super-xbr 100 AR is my preferred chroma upscaler.
This is all for chroma, though. For super-xbr image doubling the AR improvement should be bigger.
It is really nice to see your active developments but madVR seems to need a setup wizard for end [normal] users before reaching v1.0.
I know you're doing your best to make it simple but..... :p
All in good time.
Funny, with .16 I'm experiencing the same problem as described here with .15 + Afterburner:
http://forum.doom9.org/showthread.php?p=1761687#post1761687
But this time, it also happens without Afterburner. Should I provide debug logs? I can't really use the new version with this behavior.
This sounds like an issue which is out of my control. If the problem goes away as soon as an UI element is visible then the chance is rather low that it could be my fault, because (unless you're using FSE) madVR neither cares nor even knows whether an UI element is visible or not.
Do you have the same problem in FSE mode? Have you tried totally closing Afterburner (and any other GPU related tools)?
Unfortunately not: http://pastebin.com/LRxRi0Mu
Ok, at least I can see that my change is "effective", even if it doesn't seem to really help. So right now it seems that either the splitter, decoder or media player is involved in the freeze.
Could you please download the PDB debug symbols for LAV and MPC-HC and create another freeze report with those "installed" (just need to be in the right folder)? Please be extra careful to use the correct PDB symbols, because if they don't match, the freeze report will be useless. The easiest way would probably be to download the latest LAV and MPC-HC builds with matching PDB symbols. Thanks.
What is the minimal hardware requirements to play 2160p downscaled to 1080p? On HD5770 with software decoding, I have slow motioned video while audio is being played normally (so audio runs far ahead). If "decoder queue" value is "1-3/16" does it mean that the problem is on decode (LAV Video) side rather that on madVR side?
If the "decoder queue" is empty that means your decoder is not fast enough. I suppose this is HEVC content? That's very hard to decode in software. Try x64 instead of x86, that's your best bet because the x64 HEVC decoder is noticeably faster than the x86 decoder. If that doesn't help, you'll need a new GPU with HEVC hardware decoding support.
Actually, no it's the same behavior BUT I found out if I desactivate Reclock's Pal speedown (24 fps) and play it at 50hz, the queues still don't fill but I have no frampes dropped. So is it only reclock's fault?
I also tested a Blu-Ray with interlaced content and it plays fine, the queues completely filled. So it seems limited to DVD...
Let's me summarize :
-DVD interlaced with Reclock's Pal speedown (24hz) : queues don't fill, frames dropped when render queue drops down.
-DVD interlaced without Reclock's Pal speedown (50hz) : queues don't fill, no frames dropped.
-Blu-Ray interlaced and progressive : queues fill, no frames dropped.
-Other videos : queues fill, no frames dropped.
Do you have an idea to improve the situation with DVD? I can do more tests and videos, juste tell me what to do.
I don't really know why Reclock would effect frame drops, especially when using forced film mode. You could try create a debug log, then I can have a look for which exact reason the frames were dropped. Remind me: The OSD does list the frame drops, right? If it doesn't, the debug log won't help.
I don't know why the queues don't fill. That's a question you'd have to ask the decoder developer (not sure if it's nevcairiel in your case?). Maybe the DVD Navigator somehow limits this? I've no idea, I believe it's totally outside of my control.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.