View Full Version : madVR - high quality video renderer (GPU assisted)
pirlouy
31st October 2011, 20:04
All post-processing is not something which should be done by LAV Filters.
Indeed post-processing often needs profile support, and Nevcairiel does not use that much post-processing himself.
It would be better if Nevcairiel was able to create a kind of plugin support (ala Firefox or Miranda IM) for post-processing, but I think it would be boring/hard for him to implement if he does not use it in the end (and he has a subtitles engine to develop first :D)...
For now, FFDshow post-processing is better, even if bugged and not that much developed.
Gser
31st October 2011, 20:10
For now, FFDshow post-processing is better, even if bugged and not that much developed.
I mostly use its avisynth function to post process. I have been using coreavc for h.264 and Dscaler for mpeg2 for a long while already hence I have no need for LAVF.
pankov
31st October 2011, 20:56
Come on guys/galls,
this thread is called "madVR" not ReClock or LAV Filters for a reason.
I do understand that some of the questions about ReClock could be related to madVR and but all LAV talk should be moved to the LAV Filters thread.
I had to check the page title in my browser a couple of times while reading this last page.
Let's spare madshi all this important but not related to madVR talk.
dansrfe
31st October 2011, 21:39
Come on guys/galls,
this thread is called "madVR" not ReClock or LAV Filters for a reason.
I do understand that some of the questions about ReClock could be related to madVR and but all LAV talk should be moved to the LAV Filters thread.
I had to check the page title in my browser a couple of times while reading this last page.
Let's spare madshi all this important but not related to madVR talk.
@pankov
Completely agree. The thread has like 10 pages of consecutive LAVSplitter/ReClock/calibration replies...it's hard for the newcomers to sift through for strictly madVR information.
@ madshi
I have now pinpointed the problem to files within the MKV container. After 5+ mins of playback the RENDER queue goes from 7-8/8 to 1-3/8 and then frames start to drop progressively.
My scaling settings: Chroma, Luma Up, Luma Down respectively -> SoftCubic 60, Spline @ 3 taps w/ no linear light scaling, Spline @ 3 taps w/ no linear light scaling.
My decoder settings: All internal off, ffdshow libavcodec video/audio external decoders.
My splitter settings: LAVSplitter.
At first I thought it was a splitter issue but after trying the internal splitter and witnessing the same situation I have come to the conclusion that either my madVR configuration is incorrect for my system or something else is wrong and perhaps there is a bug.
Portioli
1st November 2011, 00:36
Come on guys/galls,
this thread is called "madVR" not ReClock or LAV Filters for a reason.
I do understand that some of the questions about ReClock could be related to madVR and but all LAV talk should be moved to the LAV Filters thread.
I had to check the page title in my browser a couple of times while reading this last page.
Let's spare madshi all this important but not related to madVR talk.
i completely disagree.
the main goal is to have the best PQ & the smoothest playback combined with MadVR.
most of the times, these offtopic posts *like bitstreaming vs decoding or color correction using yCMS or calibration* are the most usefull posts.
At the End , only the final result matters
rahzel
1st November 2011, 02:04
I have to agree that LAV/Reclock discussion doesn't belong here. Even though most of us use both, there IS a dedicated LAV thread and the two are very loosely related - especially the audio discussion.
Gary.M
1st November 2011, 03:27
This topic is specifically to discuss and obtain advice on the use of MadVR. We are lucky enough to have the developer reading this and responding. This is the reason why the thread should be kept clean, he shouldn't have to wade through irrelevant discussions. I'd personally be happy if a moderator was able to delete off topic discussion.
i completely disagree.
the main goal is to have the best PQ & the smoothest playback combined with MadVR.
most of the times, these offtopic posts *like bitstreaming vs decoding or color correction using yCMS or calibration* are the most usefull posts.
At the End , only the final result matters
ikarad
1st November 2011, 11:55
All post-processing is not something which should be done by LAV Filters.
Indeed post-processing often needs profile support, and Nevcairiel does not use that much post-processing himself.
It would be better if Nevcairiel was able to create a kind of plugin support (ala Firefox or Miranda IM) for post-processing, but I think it would be boring/hard for him to implement if he does not use it in the end (and he has a subtitles engine to develop first :D)...
For now, FFDshow post-processing is better, even if bugged and not that much developed.
For example, the sub renderer the better for bluray movies is the sub renderer of ffdshow (the sub renderer of mpc-hc is buggued and nobody works on these problems since many years).
Ffdshow is also the only sotware where I can use inverse telecine filter for interlaced movies.
For that, I must use ffdshow.
THX-UltraII
1st November 2011, 13:45
I m still wondering which in madVR gives me the correct information. Here are my results:
NVIDIA 1080p custom 23Hz @23,976
- madVR reports 23,97495 in the top of the OSD (so a deviation of 0.00105).
- one frame dropped every 6 minutes
- 22 dropped frames after 120 minutes of playback
The above all seems 'correct', but now
NVIDIA 1080p custom 23Hz @23,978
- madVR reports 23,97795 in the top of the OSD (so a deviation of 0.00195).
- one frame dropped every 18 minutes
- 2 frames dropped after 120 minutes of playback
The above just is not right. With this I mean that some of the information is incorrect. First off all the deviation is BIGGER then with the 23,976nvidia setting but the 'one frame dropped every' information is 'better' then the 23,976nvidia setting.
And the this that is the most strange to me is that I only got 2! dropped frames so this assumes it is almost perfect which does not correspond with the other information.
no ideas what could be going on here anyone? Seems like madVR is reporting non-correct information but which of the information is correct and which is not is the question that remains.
Razoola
1st November 2011, 16:27
From what I have seen the madVR OSD is always re calculating and updating the estimated drame drop info. Maybe thats because I use reclock though but I think the info is just an estimate and not a garantee thats how often frames will drop.
Andy o
1st November 2011, 19:43
You have to wait for it to settle.
THX-UltraII
1st November 2011, 21:02
You have to wait for it to settle. I know and it settles after 10 minutes or so at 18 minutes but after 150 minutes of playback I got 3 dropped frames so one of the is is totally wrong.
cyberbeing
1st November 2011, 22:18
The real question is do you notice little stutters which make it seem like it's dropping more frames than madVR is reporting?
What happens if you use another audio renderer?
Qaq
2nd November 2011, 00:06
What happens if you use another audio renderer?
default reference clock is in fact synced to a hardware clock somewhere in your sound card. Video frames just follow this clock, making jerkiness inevitable
I wonder what happen if we output sound to ...Null? Is there "clock" in Null too?
pacemaker1000
2nd November 2011, 01:25
an update on the 23.976 issue and getting dropped frames in MadVR
I returned the Nvidia GT430 and went back to ATI with a 6570
Instant success!
default setting shows as 23.9767 in MadVR with NO dropped frames, Glitches etc and all this without Reclock. same goes for 24 and 50 stuff
very pleased but also a little dissapointed as i wanted to use cuvid but for me Nvidia was a definate no go
also as a bonus there are a lot of features new to the 6xxx series that i didnt have in my 4xxx model. plus latest test shows the 6xxx to have the highest ever HQV results ever
http://www.tomshardware.co.uk/radeon-hd-6570-radeon-hd-6670-turks,review-32169-14.html
look forward to trying out its post processing features too
Andy o
2nd November 2011, 01:58
Those benchmarks usually aren't valid with madVR. The post-processing options aren't either. Right now AMD simply can't compete with NV for madVR use if you're doing deinterlacing. Nothing I would like more, but it's just not happening (yet?).
madshi
2nd November 2011, 09:15
(yet?)
Hehe... :D
cyberbeing
2nd November 2011, 11:38
madshi, here is the link to the issue tracking the implementation of some sort of VSFilter<->madVR interaction which somebody opened:
http://code.google.com/p/xy-vsfilter/issues/detail?id=40
No rush if you're busy, since he has plenty to keep him occupied for couple months to finish up his other planned changes and optimizations for xy-VSFilter. I asked him to register on Doom9, but if you have any further thoughts or ultimately decide against doing it, you could post updates on that GoogleCode link.
Nevilne
2nd November 2011, 12:19
I've been using xy-vsfilter lately, and even yv12 colours are good. Plus it supports hi10p and runs infamous Air clip flawlessly :o
TheShadowRunner
2nd November 2011, 15:45
Hi madshi,
about this ZP + madVR bug that triggers freeze on seek after frequency switch, I believe I found a more robust method to reproduce it.
Here it is:
With the latest ZP (http://t.inmatrix.com/zplayer800prefinal28max.exe), latest madVR, on XP SP3.
1. In ZP, set the Control Bar as follows (http://videoff7.free.fr/zpbarset.png), and the Playback Video as follows (http://videoff7.free.fr/zpvideoset.png).
2. Set madVR to change the frequency as follows (http://videoff7.free.fr/madswitchset.png).
3. Open ZP, put it in fullscreen. Have the control bar appear at the bottom of the screen. If you copied the settings above, it should stay there always.
4. Now open a PAL media (exemple 25fps xvid avi) while your current resolution is 720p60.
5. madVR switches to the resolution instructed in 2.
6. The bug can be seen: ZP interface is now slow (http://videoff7.free.fr/zpsluggish.avi), the windows "paint" function for ZP works much slower. If you scroll the control bar for say 30 seconds when ZP is in this state, it will freeze ZP entirely.
Before the freeze occurs, if you reopen the same media from there, it solves the issue instantly (ZP becomes normal/fast again) and no more freeze on seek due..
Please let me know if you can reproduce.
It doesn't _always_ happen, so please try a couple of times.
Thanks,
TSR
Addendum: The issue never happens with any other renderer + ReClock switching.
pacemaker1000
2nd November 2011, 19:55
Hehe... :D
The suspense is killing me :D
Plutotype
2nd November 2011, 20:50
Hi all,
Since 2 weeks I experience dropped frames. They occur either at the beginning of a movie or later, or never. It is very simillar to dansrfe´s post above. The movie sometimes plays without single frame drop for 30minutes and then starts to drop frames progressively. Sometimes the same movie starts and starts to drop frames immediately, sometimes 5minutes later, sometimes 50minutes later. But I had a movie without a single dropped frame. Everytime dropouts happen, the dropped frames are provided with decrease in presentation queue, which drops to 0-8/8 and also all 3 queues ( decode, upload, render ) go down and back. I tried windowed or exclusive mode, both gave me dropped frames. Tried different older versions of LAVfilters, madVR, Catalyst, MPC-HC, with or without Reclock, 60p / 24p. Nothing helped. Has anybody AMD card and was able to watch framedrop-free movies in the last 2 weeks? My setup is MPC-HC 3801, madVR 0.77, LAVfilters 0.38, Reclock 1.8.7.7, Catalyst 11.10 WHQL, hardware specs please in my signature.
Madshi, this is the log from today ( compressed 117MB, after decompression 4GB ):
http://www.mediafire.com/?cg1x2rgd5br7cq4
Thanks
Pluto
Gary.M
2nd November 2011, 23:38
Dropped frames, or presentation glitches?
... Since 2 weeks I experience dropped frames.... Everytime dropouts happen, the dropped frames are provided with decrease in presentation queue, which drops to 0-8/8
Plutotype
3rd November 2011, 00:20
Dropped frames, or presentation glitches?
Dropped frames, it happens with this video for example:
http://www.mediafire.com/?1rja6beiptppzq8
After some time of playback, frames start to drop. 20 minutes of playback gave me 100 dropped frames.
dansrfe
3rd November 2011, 03:59
The dropped frames problem happens so randomly for me now that I don't know what type of file, resolution, scaling algorithm or anything for that matter to trace the problem back to.
6233638
3rd November 2011, 16:49
If you are getting dropped frames, have you tried changing your GPU power saving options?
I have some files that are HD, but low bitrate, and this can cause dropped frames as it changes between power states when the GPU load drops low enough.
dansrfe
3rd November 2011, 18:46
Actually I think the problem was my GPU temps....Apparently the thermal pads (I hate thermal pads) that were being used on my GPU and CPU combined heatsink was melted and gross looking. All my temps were marginally higher than normal and after cleaning out and replacing the crappy pads with fresh AC5 thermal paste all the temps are nice and low and madVR seems much more stable than before as is my system overall.
Xaurus
3rd November 2011, 19:02
Will a GTS 450 run madvr sufficiently? 1080p display, mostly hd material and using spline 3, spline 3, mitchell. I am thinking of getting a dedicated htpc box (fanless GPU card).
This with a Core 2 Q6600 CPU.
whitestar999
3rd November 2011, 20:14
GTS450 is sufficient for all video files except maybe some high fps & high bit rate interlaced videos but not sure.
nevcairiel
3rd November 2011, 20:17
The GTS450 is plenty fast for any kind of video.
Bitrate/complexity actually doesn't make that big difference to hardware decoders, because they are fixed function hardware, optimized to do one task, and one task alone. There is a world of difference between that and a general purpose CPU.
In any case, the last benchmarks i did showed quite alot of headroom on a GTS450 after deinterlacing and madVR.
gommorah
4th November 2011, 07:12
Hi madshi,
When switching between multiple subtitle streams in Haali while the video is paused, madVR seems to cause the whole player to deadlock. Doesn't happen with EVR as far as I can tell. You can use this sample (http://warpsharp.info/random/lucky-star-vsfiltermt-crash-001.mkv) to test. I've tested this with the latest version of madVR and CCCP 7-30-2011. Let me know if you need more info.
cyberbeing
4th November 2011, 11:33
I've run into the issue gommorah just mentioned many times, but its occurrence has always been somewhat random. The majority of the time it does hang, but occasionally I'll have long stretches where it doesn't, for reasons unknown.
SamuelMaki
4th November 2011, 16:57
I do not know if this is potplayer issue, but it may be madVR as well...
So, when using potplayer subtitle renderer, I cannot use exclusive mode (it switches to windowed every time subtitles is displayed). Is this a global issue, or have I done something wrong?
fairchild
4th November 2011, 17:30
I do not know if this is potplayer issue, but it may be madVR as well...
So, when using potplayer subtitle renderer, I cannot use exclusive mode (it switches to windowed every time subtitles is displayed). Is this a global issue, or have I done something wrong?
This doesn't happen to me. I use Potplayer with Lav filters + madVR and I use Image Overlay on screen skin option which uses Direct3D 9. Not sure if that has something to do with me not having the issue.
SamuelMaki
4th November 2011, 19:27
This doesn't happen to me. I use Potplayer with Lav filters + madVR and I use Image Overlay on screen skin option which uses Direct3D 9. Not sure if that has something to do with me not having the issue.
OK... Exclusive mode fails only if I use madVR own h264-decoder... LAV and ffdshow works well. So, any possible fixes?
golagoda
5th November 2011, 00:46
Hi madshi,
When switching between multiple subtitle streams in Haali while the video is paused, madVR seems to cause the whole player to deadlock. Doesn't happen with EVR as far as I can tell. You can use this sample (http://warpsharp.info/random/lucky-star-vsfiltermt-crash-001.mkv) to test. I've tested this with the latest version of madVR and CCCP 7-30-2011. Let me know if you need more info.
I sometimes get this problem too but it only happens when I use haali as my splitter for whatever reason.
ryrynz
5th November 2011, 09:34
Hard to get glitch free playback with an Nvidia card, got a Ti 550 I've tried Nev's config, 6233638's and a number of variations.. still haven't found something that's problem free yet.
golagoda
5th November 2011, 09:51
I'll post my settings later when I get the time and I'm on windows, I just about never get frame drops or presentation glitches and I especially don't notice bad or glitchy playback when I actually watch videos regardless of whatever statistics.
ryrynz
5th November 2011, 13:14
I found some settings that are doing quite well, I have zero glitches after 30 mins. I'll continue to refine them then post them up tomorrow, thanks for the offer.
Andy o
5th November 2011, 15:53
I do not know if it was already discussed, but if i start a playlist with my mpc hc (build r3788 and lower) with more than one video file (e. g. mkv) with mpc hc in fullscreen, i get problems with it. After the first video was ended and will go to the next video, windows and mpc hc hangs completly. It's completly stuck and i only can kill the task via my taskmanager. This happens only if mpc hc are in fullscreen and i have a playlist with some videos. Maybe a MPC HC problem or madVR? Or some settings wrong here. I'm under Windows 7 x64 SP1 and nvidia graphics (latest drivers).
If anyone else is having problems like this, do you by any chance have Firefox or IE9 running in the background? If so, could you close them and see if it still happens? I've had similar problems but this is hard to troubleshoot cause it doesn't happen always. It seems to happen more when the first file has been running for some time, but that could be just me. As far as I can tell it doesn't matter if you're switching from SD to HD files or anything like that. It also doesn't only happen with playlists, just skipping to the next file in the folder will do it too.
With madVR I only get a crash like described in the quote, but with the EVR renderers I get a crash dialog. Apparently with Haali VR it doesn't happen. It might be related to HW acceleration on Firefox and maybe IE. So far (knock on wood) it's not happened if I don't have FF opened, which I used to have all the time.
I'm using LAV filters .38 with CUVID enabled, but IIRC it also happened when I was using my 5770. Been having the problem for quite some time, only that it didn't bother me as much and it didn't happen as often for me to wanna troubleshoot it.
6233638
5th November 2011, 16:25
Hard to get glitch free playback with an Nvidia card, got a Ti 550 I've tried Nev's config, 6233638's and a number of variations.. still haven't found something that's problem free yet.Make sure you're running the latest WHQL Nvidia drivers. (285.62) Stay away from Betas, as they always seem to introduce problems.
derp
5th November 2011, 16:55
When the "disappearing subs when paused" problem will be fixed? It's a pretty damn important issue.
sneaker_ger
5th November 2011, 17:08
When the "disappearing subs when paused" problem will be fixed? It's a pretty damn important issue.
Which problem are you referring to exactly? Which madvr version are you using?
A problem with disappearing subs was fixed in version 0.71.
madshi
5th November 2011, 20:38
If you've had a change of heart and are willing to make the 8-bit textures setting usable (with OSD stats), I'd be interested to see if there was any significant performance gain to be had with heavy duty processing.
Why not make it simpler and just add a drop-down box to choose between 10-bit and 8-bit buffers for those two settings? It doesn't make any sense to have a temporary warning, since it could lead people to believe 10-bit textures are actually working when they're not.
I've simply removed the warning for now, I've no fun/time doing anything more complicated with this atm. I've added to my long term to do list to maybe reintroduce a better warning or something else.
On another note, ever since you added those extra stats to the OSD, it's now 4 lines too tall to display at 360p.
Honestly, who cares whether the OSD fits at 360p? I do not really care at this point in time. Maybe you can mention this again when madVR has reached v1.0. But right now I've literally hundreds of more important things to do than polishing a debug OSD to fit into ultra low resolution movies.
i tried MPC-HC_r3792_x86, and using MadVr 0.77, with Nvidia notebook driver 280.26.
anisotropic and anti-aliasing is disabled, and not using Reclock.
The freeze still occurs, i think it is not caused the movie file, as the point where it freezes, might not freeze again when reopen the video file.
Same problem that jovovick has (video freezes after some time (about 60-90 seconds) when using nVidia card on Optimus system).
Updated to 285.62 nVidia notebook driver... Problem still occurs. Plus now it's impossible to use discrete graphics (choice is disabled) for mpc-hc.exe, so I had to rename exe file in order to use it.
Seems that problem is in Optimus...
This is really sad. To be honest, right now I don't know what to do or what to suggest. You could try updating your BIOS. Or maybe you could make and upload a madVR debug log (try to keep it as short as possible).
10-bit 4:2:2 sample:
BBB422p10-rr3.rar (http://www.mediafire.com/file/3n64vyvlrqduhg1/BBB422p10-rr3.rar)
Thanks. This will be fixed in v0.78.
The issue with my dropped frames is not related only to the video I posted. Today I watched other videos, there were 10-15 dropped frames. Again I have everything up to date: MPC-HC 3794, LAVfilters 0.38RC, Reclock 1.8.7.7. and madVR 0.77.
Here you can find a log for the same video I posted here.
I installed the same SW setup on different computer with HD6570 (fresh win7 install ), also 11.9 Catalyst and the result was the same also there. Even under 24p setup.
The issue is the same in both cases. According to the log, playback is running just fine. Then all of a sudden the decoder stops sending frames to madVR for a short while. Of course audio continues to play in the meanwhile. So once the decoder catches up again, madVR has no other choice than to drop a couple of frames to make sure that audio and video stay in sync. Now what you need to find out is why the decoder sometimes freezes for a short while. This is outside of madVR's control, as far as I can say. It could be caused by a bug in the decoder, or by your PC being very busy for a short while, so that no CPU time is left for the decoder at that time. My first suggestion would be to try a different decoder, just to make sure it's not the decoder. If it's not the decoder (and I guess it's not) then you should run some profiling tool which records CPU consumption during video playback. Then when you get these frame drops, stop playback and check whether the profiling tool has recorded 100% CPU consumption shortly before.
HD6570 ( different computer ), log with 2 dropped frames in approximately 10 minutes of video
This is a different issue. It seems that the refresh rate is very slightly too low (or the movie playing very slightly too fast). So madVR has no other choice than to drop a frame once in a while.
I'm having a similar issue to Plutotype yet I can't figure out why. 8 frames queue limit on decoder, upload, render, and backbuffer. About 2-3 frames drop every 10-15 seconds. Some frames are also delayed. I'm not sure if it is for this specific file I;m playing or it happens on other files as well.
Your log shows that the frames are coming from the decoder with swapped timestamps. Meaning that probably the splitter and video decoder don't like each other much. I see from the log that you're using LAV Splitter -> ffdshow Video Decoder. LAV Splitter should have automatic detection of the decoder and behave in such a way as to make the decoder happy. Not sure, maybe you disabled the functionality? Or maybe something else is configured badly. Anyway, if you fix this timestamp order problem, the frame drops should go away. Alternatively, cut down the number of prepresented frames, that will give madVR some more time and space to reorder the frames for you. Well, this may not be necessary with madVR v0.78. But anyway, you should really try sorting out the swapped timestamp problem.
I was reading an updated version of one of my favorite HTPC configuration GUIDES (http://www.homecinema-hd.com/ffdshow-video-avisynth_en.html)
and it is now updated to suggest to upscale 720P material in aviSynth (spline64resize function) instead of in madVR so that the aviSynth function LimitedSharpenFaster can correctly perform the sharpening on the upscaled material. I completely understand the reasoning behind this; however, what is your opinion on the benefits of this vs. the quality of the madVR upscaler (which I still find to be unsurpassed, including aviSynth-scripts...)
If you like the results LimitedSharpenFaster produces then the suggestion of that configuration guide is not bad at all. Of course madVR's scaling quality itself is higher than aviSynth's spline64resize, due to higher bitdepth, but madVR does not have any sharpening built in and LimitedSharpenFaster is a rather good algorithm. So I would say: Try it both and let your eyes decide.
I've got a weird bug with madVR, when I turn off/on(shift+X by default) audio processing filter in Daum Potplayer, video hangs or player crashes. First I thought that this problem occurred due to some kind of splitter incompatibility, I tried various combinations of splitter/decoder pairs, haali/lav + CoreAVC/LAVcuvid/LavVido with cuvid implementation/Built-in Decoder in renderless mode, and got same result, then I tried to change render to EVRcp, and it helped. Is it possible to fix it?
If I can reproduce it. I've just tried and turning audio processing on/off in PotPlayer does not crash here with my latest v0.78 sources. You can try again with v0.78 once it's released. If you still have the crash with v0.78, let me know, but then I'll need to be able to reproduce the crash on my PC.
I'm using mpc-hc with lav filters, madvr, ffdshow raw video filter.
Is there a way to check my frame rate? because i'm not sure if anything happens with LAV video decoders deinterlacing 25p/30p or 50p/60p. Nothing seems to change when i switch between those two options.
Then if i use ffdshow with yadif's deinterlacing double framerate, i can hear fan picking up speed immediately and if i use zoom players playing information, i can see actual frame rate doubleing.
i tried ctrl+j in mpc-hc but can't see it telling me if framerate is doubeling; nothin changes if i switch between 25p/30p or 50p/60p in LAV video decoder.
madVR does not currently show how many fps the decoder sends. madVR does show in the debug OSD (Ctrl+J) which framerate the decoder reports (e.g. "movie 29.970 fps"), but that's just one bit of information sent by the decoder which could be wrong. Maybe I'll add more information in a future madVR version. If you want to know right now, maybe you should ask nevcairiel in the LAV thread.
EDIT: Scale in linear light is the cause of dropped frames in exclusive mode is enabled in Luma Upscaling for those with CPU's that aren't powerful enough.
madVR's scaling in linear light is performed by the GPU, not by the CPU.
madshi can you look into playing files with firefox open in the background. Apparently that seems to be the problem with the dropped and delayed frames due to the queues not filling up.
Which queues are not filling up exactly?
madshi,
Please take a look at this log:
http://www.sendspace.com/file/xu41u8
Scenario:
I start playing a 1080p file from my external hard drive and then I go to exclusive mode. After playing for more than (10 mins?) All the queues drop and frames start dropping. I'm going crazy trying to figure out why this is happening when I haven't changed anything from the start to end of playback.
No, not all queues drop. *Please* guys, you need to be exact and correct in what you're reporting. According to the log, the decoder queue is always pretty much full. It's the upload and render queue which drop down, which means that your GPU isn't running fast enough for whatever reason. Maybe it has clocked down?
So I was going through my old DVDs and came across my Digital Video Essentials disc.
I find it to be mostly useless these days, but some of the multiburst patterns were interesting, especially when evaluating upscaling algorithms for sharpness vs ringing. (has me re-evaluating my choices, though I still prefer a lack of ringing over sharpness)
But it really highlights what seem to be some big errors when using linear light scaling. (upscaling at least, downscaling with Bicubic 75 hasn't caused me too many problems)
This is at high magnifications, and using Lanczos 8 which shows things more clearly.
Is it possible that there's a bug in the linear light scaling implementation?
(P.S. I wish there was some kind of no-ring Lanczos8 filter, aside from the ringing, it's definitely better than any of the other choices in madVR)
Can you upload a sample with that specific test pattern? I don't think that there's a bug in the linear light scaling algorithm, but I'm not 100% sure. Can't hurt to analyze.
I m still wondering which in madVR gives me the correct information. Here are my results:
NVIDIA 1080p custom 23Hz @23,976
- madVR reports 23,97495 in the top of the OSD (so a deviation of 0.00105).
- one frame dropped every 6 minutes
- 22 dropped frames after 120 minutes of playback
The above all seems 'correct', but now
NVIDIA 1080p custom 23Hz @23,978
- madVR reports 23,97795 in the top of the OSD (so a deviation of 0.00195).
- one frame dropped every 18 minutes
- 2 frames dropped after 120 minutes of playback
The above just is not right. With this I mean that some of the information is incorrect. First off all the deviation is BIGGER then with the 23,976nvidia setting but the 'one frame dropped every' information is 'better' then the 23,976nvidia setting.
And the this that is the most strange to me is that I only got 2! dropped frames so this assumes it is almost perfect which does not correspond with the other information.
The OSD information about how many frame drops/repeats to expect per second/minute/hour/etc depends on the audio clock deviation measurement, and that is very hard to get right because the audio hardware clock is extremely jittery. So you cannot fully trust this information. The number of actually dropped frames is much more reliable.
Hi madshi,
about this ZP + madVR bug that triggers freeze on seek after frequency switch, I believe I found a more robust method to reproduce it.
Here it is:
With the latest ZP (http://t.inmatrix.com/zplayer800prefinal28max.exe), latest madVR, on XP SP3.
1. In ZP, set the Control Bar as follows (http://videoff7.free.fr/zpbarset.png), and the Playback Video as follows (http://videoff7.free.fr/zpvideoset.png).
2. Set madVR to change the frequency as follows (http://videoff7.free.fr/madswitchset.png).
3. Open ZP, put it in fullscreen. Have the control bar appear at the bottom of the screen. If you copied the settings above, it should stay there always.
4. Now open a PAL media (exemple 25fps xvid avi) while your current resolution is 720p60.
5. madVR switches to the resolution instructed in 2.
6. The bug can be seen: ZP interface is now slow (http://videoff7.free.fr/zpsluggish.avi), the windows "paint" function for ZP works much slower. If you scroll the control bar for say 30 seconds when ZP is in this state, it will freeze ZP entirely.
Before the freeze occurs, if you reopen the same media from there, it solves the issue instantly (ZP becomes normal/fast again) and no more freeze on seek due..
Please let me know if you can reproduce.
It doesn't _always_ happen, so please try a couple of times.
Thanks,
TSR
Addendum: The issue never happens with any other renderer + ReClock switching.
I'm sorry, but this really *REALLY* sounds like a ZoomPlayer bug to me. I can still try to reproduce it, but not today.
Madshi, this is the log from today ( compressed 117MB, after decompression 4GB )
I appreciate the upload etc. But are you aware that a 4GB madVR log file contains about 100 million lines of text? My text editor can't even open a text file this large. And even if it could, I think you're underestimating how much time it costs to read and understand log files.
OK... Exclusive mode fails only if I use madVR own h264-decoder... LAV and ffdshow works well. So, any possible fixes?
That's weird. There's no reason why the internal h264 decoder should result in exclusive mode stopping to work. I'd say that this is likely a PotPlayer configuration problem or maybe a PotPlayer bug. Have you tried contacting the PotPlayer developer about this?
When the "disappearing subs when paused" problem will be fixed? It's a pretty damn important issue.
What problem do you mean exactly?
When switching between multiple subtitle streams in Haali while the video is paused, madVR seems to cause the whole player to deadlock. Doesn't happen with EVR as far as I can tell. You can use this sample (http://warpsharp.info/random/lucky-star-vsfiltermt-crash-001.mkv) to test. I've tested this with the latest version of madVR and CCCP 7-30-2011. Let me know if you need more info.
I can reproduce the problem but I don't have the means at this point in time to figure out what's going on exactly. On a first glance it doesn't seem to be a bug in madVR, but I could be wrong on that. FWIW, the issue doesn't seem to occur if you switch subtitles via madVR's tray icon. Which is what I'm usually doing, anyway. Isn't madVR's tray icon much prettier than Haali's? :p
madshi
5th November 2011, 21:09
madVR v0.78 released
http://madshi.net/madVR.zip
* added support for DXVA2 deinterlacing
* added option to enable/disable automatic activation of deinterlacing
* added option to force off deinterlacing for 48i sources
* added option enable/disable performing deinterlacing in a separate thread
* added auto 4:2:2/4:4:4 to 4:2:0 conversion when using DXVA2 deinterlacing
* added OSD deinterlacing information
* Ctrl+Alt+Shift+D turns deinterlacing on/off
* Ctrl+Alt+Shift+F switches deinterlacing field order (auto, top, bottom first)
* changed gamma decrease key combo from Ctrl+Alt+Shift+F to Ctrl+Alt+Shift+B
* increased decoder queue size to 12 frames
* added support for h264 10bit 4:2:2 decoding
* improved frame stepping
* improved internal frame locking mechanism
* removed warning "creating 10bit render texture failed, used 8bit instead"
* updated libav/ffmpeg dlls
http://forum.madshi.net/images/smilies/icon_redBalloon.gifhttp://forum.madshi.net/images/smilies/icon_greenBalloon.gifhttp://forum.madshi.net/images/smilies/icon_blueBalloon.gif
added DXVA2 Deinterlacing
no support for DXVA2 Decoding, just to avoid any misunderstandings
Some things to note:
(1) Deinterlacing requires DXVA2 to be present. On XP you probably need to install .NET 3.0 (or higher) to get DXVA2.
(2) Using DXVA2 deinterlacing means that potentially all the nasty GPU algorithms like noise reduction and edge enhancement are introduced to madVR. So make sure you have all this stuff turned off in your GPU driver's control panel. Of course madVR only uses DXVA2 if deinterlacing is needed. So for progressive stuff (or if you deinterlace externally) nothing changes compared to older madVR versions.
(3) At this point in time madVR mostly does the same that VMR and EVR do. Which is not ideal in some cases. E.g. MPEG2 flags are trusted, so movie deinterlacing might not always be correct, especially with MPEG2 DVDs and broadcasts. I might improve that in a future version.
(4) madVR is less picky than other renderers about which color spaces you can send. madVR will happily deinterlace all 8bit YCbCr formats. Even 4:2:2 and 4:4:4, which madVR then converts down to 4:2:0 for DXVA2 deinterlacing.
(5) ATI users please activate the switch "Advanced Color -> Dynamic Range" and set it to "Limited (16 - 235)". This will make sure that BTB and WTW are not cut off by DXVA2 deinterlacing.
(6) I've removed the Intel software decoder DLL from the madVR zip in order to cut down the zip file size, and because most people probably use the libav decoders, anyway. If you want to use the Intel decoders, you can download the dll separately here (http://madshi.net/libmfxsw32.zip). Just copy it to the madVR folder and the Intel decoders should start working again.
(7) This is a first experimental version, bugs are to be expected.
FEEDBACK:
I need to know...
(a) ... how deinterlacing quality and performance compares to LAV CUVID, could you please check that? If LAV CUVID looks or performs better, maybe there's room for me to improve, too.
(b) ... whether the option "perform deinterlacing in separate thread" makes things better or worse for you. Ideally I hope it will be better or worse for everybody so that I can then hard code the "right" setting and remove the option. On my PC it seems that having this option off works better for windowed mode, but having it on works better for exclusive mode.
Thunderbolt8
5th November 2011, 21:28
omg this is going to be great :D
one question, seeing this limited (16 - 235) switch: is there going to be a problem or any interference when using 3dlut scripts to modify black & white levels in combination with this?
is dxva2 comparible to dxva? is my HD 4770 supported? if so, how to recognise whether its active or not?
noee
5th November 2011, 21:41
sweet, madshi, very nice.
b) At my viewing distance (about 10ft), the option ON looks better in FSE mode.
HD6570 secondary NTSC interlaced video.
Thunderbolt8
5th November 2011, 21:51
all the BBC documentaries (VC-1) I have dont work with the libav decoder. shall I make samples?
dansrfe
5th November 2011, 21:52
Well, my problems are fixed. Testing out DXVA2 now...
Thunderbolt8
5th November 2011, 22:01
also, cant find this in ati catalyst "Advanced Color -> Dynamic Range" and set it to "Limited (16 - 235)"
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.