madVR - high quality video renderer (GPU assisted)

[leeperry]

I think back when you tested the high refresh rate model, you were using different settings for many other tweak options. Maybe you could switch to high refresh rate model for the next 2 movies, just to see if it really is worse? Maybe it's just as good as the normal timing model now?
I actually ran some tests only changing the timings model, but I'll try again...just don't make the "normal" timings option vanish for now please :p

and is it really necessary to reseek after going windowed>FS? cannot hurt I guess?

[yesgrey]

I PM'ed you those samples madshi, so they won't be found on the forums.

Sent you a PM iSunrise.
I also would like to give it a try - I have a GF 8600GT. Would you PM me the samples too, please?

[yesgrey]

madshi,
one thing I don't understand... What do you mean with the "? + xx ms" as rendering time?

[madshi]

I actually ran some tests only changing the timings model, but I'll try again...just don't make the "normal" timings option vanish for now please :p

and is it really necessary to reseek after going windowed>FS? cannot hurt I guess?
Don't worry, I won't remove the "normal" timing model, if your test shows that it's really better for you.

And I don't think that a reseek is necessary.

madshi,
one thing I don't understand... What do you mean with the "? + xx ms" as rendering time?
If you don't use "flush & wait" for "after last step" then the rendering time measurement shown in the OSD is too low, because madVR doesn't wait for the rendering to complete. So that "? + xx ms" means that you need to add an unknown amount of time to the measured numbers.

[Mark_A_W]

I've got the Zoomplayer mouse/control bar missing issue too madshi.


Oh, and I wonder what the secret is that caused the OSD to be rendered opaque?? ;) I hope it will make Zoomplayer usable in Exclusive mode.



Oh, and the left mouse button mouse doesn't work at all in Zoomplayer during playback, but the right click does.

[Mark_A_W]

Oh, and madshi, I'm getting some super-smooth results with 0.22.

But it's not consistent. Sometimes there are periods with delayed frames, then sometimes the phases align and it's hyper smooth. The reclock tearing bar has perfect motion, the reclock v-sync marker barely moves and the delayed frames don't change at all, for ever after it comes good (I think).

The problems is getting it to come good. I'm using Upload in Render = On. High Refresh Rate, and 3 buffers. Flushes at stock settings.


Also, I fiddled my timings a bit, back to 95.904, instead of 95.905, it seems to help (it's difficult with an interlaced res, I actually enter 47.952 in Powerstrip and it multiples by 2, with a rounding error..it's hard to get it bang on - I have 95.90467hz now.).

Cheers

Mark

[yesgrey]

it's hard to get it bang on - I have 95.09467hz now.).
What's your graphic card, working resolution and display type (is crt)?

[nlnl]

Thanks madshi for 0.22!
Tried to play 1080i50 H264 film based content using CoreAVC (set to No interlacing (weave))+ MADVR. Display is set to 1080p50.
PQ is very good. No deinterlacing artfacts.
Does that mean that CoreAVC+Madvr restore original 25p film source from 1080i50 'container'?
And we do not need advance hardware vector adaptive deinterlacing for films in 1080i50 format?

[iSunrise]

Since you are on Vista/Win7, please disable Aero, enable Flush and Wait for 'After Last Step' (so you can see the actual render times), set Chroma resampling to Spline64 (to make rendering time changes more obvious), and run through that section again watching for a render time spike when those orange spinning things appear on screen.
Ok, thanks for the detailed explanation.

I just went back again and just did exactly what you described, I:

1) disabled Aero on Vista x64
2) set luma up-, down- and chroma resampling to Spline64, instead of my default settings (luma upscaling Bicubic75, luma downscaling Spline64 and chroma resampling Bicubic75)
3) changed my windowed mode tweaks from "don´t flush" on everything to "flush, flush & wait (sleep) / flush & wait (loop), don´t flush, don´t flush
4) and played back your sample with your 3DLUT in windowed mode and fullscreen

What I saw is this:

1) There´s a (looks quite horrible) tearing line at the bottom when Aero is off. (That´s one of the reasons why I have always used Aero on Vista x64, because it works flawless.)

2) The max stats (5s) are very stable throughout playing that sample, there´s no spikes at all. Not even for a millisecond. No dropped or delayed frames, either. This is the same as with Aero on in Vista, however, with Aero off my max present times are a bit higher. The rendering times are very low and always very stable, too.

I´m sorry, but I still can´t reproduce the spikes you are getting. I even tried a short sample of a 42mbps broadcast with your settings and 3DLUT, but I cannot see any abnormal behaviour on that, either (apart from the tearing).

[Razoola]

@cyberbeing.

Seeing iSunrise cannot seem to reproduce your situation I'm wondering if your PC has an issue with system latency due to an incorrectly set MB bios setting or mobo drivers. Give http://www.thesycon.de/deu/latency_check.shtml a try (freeware) and see if your getting some spikes there.

[ryrynz]

Revision 3508 of ffdshow now adds adds madVR to list of compatible renderers for dynamic format changes.

What this means is that the option allow output format changes during playback no longer needs to be ticked and can instead stay greyed out which is the default setting.

Xhmikosr's latest build can be found here.

http://www.mediafire.com/?sharekey=3f33c77c2cf9ce25ab1eab3e9fa335ca95e79d7d0540e1e1

Regards,
Ryan.

[Mark_A_W]

What's your graphic card, working resolution and display type (is crt)?


Ati 4850, 512mb.

1920 x 1080 interlaced at 95.90467hz (according to madVR).

In Powerstrip or the Nvidia drivers this would be referred to as 1080i 47.952hz.

And yes, I have CRTs. I have a pair of 24" Sony Trinitron 16:10 monitors on the desktop. The secondary is fed through a video splitter which also sends the signal through to the CRT projector in the next room.

Running a CRT is good for this testing, because I know that I am seeing the true output of the video card - the CRT monitors can't "fiddle" with the refresh rate, like some digitals can.

(But this will be my last CRT projector, I saw a friend's Sony HW15 last night....whoa. It spanks the CRT in every way except black level - amazing at the price.)

[ajp_anton]

Feature request:

If the video is going to be upscaled by a factor or number of pixels less than [set by user], do not resize at all.
For example, when playing a 1916x800 video on a 1920x1080 screen, leave 2 pixels wide black bars on each side of the video instead of upscaling.

[ramicio]

This latest video I'm encoding isn't working with madVR. It just draws a black screen, no video at all. It gets the input right but it doesn't draw it. It says it's YV12 too. I am encoding with x264. Here is the command:

x264 --pass 2 --bitrate 10700 --stats "file.stats" --threads 8 --level 4.1 --bframes 3 --ref 4 --slices 4 --aud --nal-hrd vbr --b-pyramid strict --keyint 24 --min-keyint 2 --vbv-maxrate 14000 --vbv-bufsize 14500 --weightp 0 --colorprim "bt709" --transfer "bt709" --sar 1:1 --colormatrix "bt709" --output "out" "in"

I don't know what half of that means. If there is any problem in that code, please help! I just copied the command from megui. Video is 1920x1040, 24000/1001 fps. Thanks!

[andybkma]

Hi, just want to report a problem with the newest version .22. Using XP SP3 and Zoom Player, mouse functions don't work inside the viewing area whether in window mode or fullscreen. For instance, single click to pause/unpause the video does not work. This is a new problem inherent with .22, going back to .21 does not have this issue.

ANd thanks for making/developing this great renderer...

[Trigunflame]

Madshi, while doing some random testing (unrelated to AR), I've came across a regression of sorts.

To reproduce on my end:
I play any h264 video, so that CoreAVC2 is utilized in MPC-HC (r2116).
While the video is playing, I open the CoreAVC2 Filter configuration, modify any setting (say, input level for example) & hit apply.
After doing so, the video will freeze, meanwhile the audio continues in the background.
I can stop/start/pause & seek around the video normally, however, the "Video Image" itself will no longer function.
Therefore, the player must be restarted in order to continue watching.

Also something interesting to note:
I can prevent this "Freeze" from occurring by making use of DirectVobSub.
I'm assuming this is because it places itself in-between MadVR & CoreAVC in the filter chain?.

Lastly, I can confirm that this particular issue did not occur in v0.21; so it apparently has something to do with changes made in v0.22

Logfile: http://www.multiupload.com/42WQ958GBA
Demonstration: http://www.youtube.com/watch?v=2gQuIWm-x9s

[DigitalLF]

Trigunflame i got the same problem... it just happens randomly for me i do not need to change settings...

[leeperry]

Revision 3508 of ffdshow now adds adds madVR to list of compatible renderers for dynamic format changes.
What does that tranlate to IRL? It'd improve the transition between different AR's in seamless playback mode(KMP/PotP) I presume? mVR only does YV12 so I guess we're not talking about colorspace formats.
Don't worry, I won't remove the "normal" timing model, if your test shows that it's really better for you.

And I don't think that a reseek is necessary.
OK, I ran some tests yesterday but the movies I watched had hiccuping telecines huh...need more time.

Oh, I thought a reseek would be necessary to make Reclock happy and allow mVR to catch the VSYNC fliptime properly(XP's main weakness as I understand it)....I'll keep doing it for the next tests, and later on I won't anymore in order to see whether that'd make any difference.
I prefer ATI right now because:

(1) The "power per watt" ratio is clearly better, which is very important for me (because HTPCs are usually thermally limited).
Well, would that still be the case if you took CoreAVC CUDA in account? I can easily add heavy Avisynth post-processing on high bitrate 1080p AVC...it takes 35% of my 96SP G92 VP2(monitored in GPU-Z), the CPU is hands-free. Actually, I've got a wattmeter so I could easily double-check. And mVR doesn't allow DXVA, so we're talking about 100% GPU Vs 100% CPU decoding. I think some tests were showing that VP2 HA hardly increases the GPU temperature/power consumption.

EDIT: after more testing, it would appear that I really do need the "normal" timing model in 96Hz...the "high refresh rate" model gives random delayed frames from time to time..

[Razoola]

For those people using madVR with reclock note that there is a new reclock version (1.8.7.0). In the release notes it mentions improoved refresh rate detection so it may help some people who use older versions and have stutters. On the other hand the new version may introduce new issues. Its something to be aware of.

[leeperry]

there is a new reclock version (1.8.7.0)
dead smooth for 100 mins here w/ those settings on XPSP3/CoreAVC CUDA in 96.000Hz: http://thumbnails27.imagebam.com/8829/73dd5488285677.jpg (http://www.imagebam.com/image/73dd5488285677) and w/ this kind of stats in the end: http://thumbnails28.imagebam.com/8844/a1b5f788438648.jpg (http://www.imagebam.com/image/a1b5f788438648)

I can't believe how smooth the telecine of the US "Assault on Precinct 13" BD is, and it's got a lot of very slow pannings too...really great stuff :cool:

[Frank K Abbott]

madshi,
Can you make the renderer downsize anything greater than or equal to the screen resolution to like 1/2 the original size? I like the way Haali Renderer does that. If you can add an option for that or change that then I'm switching to madVR forever because the smoothness is amazing :D I hope it just gets better and better :)

[pankov]

madshi,
last night I upgraded my video card to ATI 5750 and I was pretty happy with the results until suddenly an hour or so into a recorded basketball game (720p@59.940) it started dropping frames.
Everything was butter smooth before this.
I tried everything - changing settings in madVR, changing back to v0.21, removing ReClock because I've upgraded it to 1.8.7.0 the same evening, changing refresh rates and going back to 59.970 but sadly nothing helped.
Since the problem occurred it was present on both monitors (2x 1920x1080 @ 59.940).
I've managed to make a log file for you
http://www.mediafire.com/?iiymnwmzmht
As far as I noticed the problem was in the high rendereing times (max (5s)) but even after I lowered the settings to Nearest Neighbor the dropping continued.
I checked with GPU-Z and the GPU load was ~35% with Spline64 at all scaling options ... and there was no fluctuations in the GPU/Mem frequencies.
The only thing that helped was a restart of the computer.
So the question now is - is my new video card somehow defective or it's something else? But what?
(I've check the CPU usage and the list of processes and there wasn't anything out of the ordinary)

[namaiki]

pankov, did you try 'clean-up video clock timings database' in ReClock Configuration?

[makakam]

pankov, I had sth similar. I installed the newest mpc hc, matroska splitter and started using coreavc as a decoder. All this with reclock and everything is just fine.

[pankov]

namaiki,
when I was looking for solution I even stopped using ReClock - I went back to DirectSound Device but the problem was still present.

makakam,
I'm not using mpc-hc, I use ZoomPlayer and for this recording (but it wasn't only it) I can use CoreAVC because the video stream is MPEG2.

The strange part is that the same file even the same parts of the file where working absolutely fine before the problem started about an hour into the game. And only restarting the PC cured it :(
Tonight I'll try and check if it will happen again, because I have a few more games to watch

[leeperry]

I never really paid attention(due to the previous OSD color being too dark) to the amount of jitter you get when switching resolutions..yay! it usually starts at 96.00033Hz for me...then hiccups very quickly between ~33/~25, and then stabilizes after a few mins and slowly decrements to reach 96.00015Hz 90 mins later.

I wonder if someday there'll be a manufacturer interested in making a rock-stable card for HTPC use, using a 1ppm/low jitter clock for instance...they've done it for audiophile gear, it's about time we get a "videophile" graphic card :devil:

They could keep the cheapo 27Mhz PLL, and simply add a clock conditioner/cleaner to get 75ps jitter: http://www.cirrus.com/en/press/releases/P543.html

it'd only cost one dollar or so...and this would provide dead stable refresh rates :cool:

PS: we could always use a 5ps 27Mhz Burson clock, hah: http://bursonaudio.com/burson_clock.htm

[pankov]

leeperry, Mark_W,
a while ago you were discussing about which is better 23.976 vs. 24.000.
I know that the original framerate on the disc is 23.976 but in the cinema it's 24.000 and if I don't mind audio resampling with Reclock (currently I can't use HD audio bitstreaming because my receiver doesn't support it) won't it be better (closer to cinema experience) to use 24.000?
The other reason I'm asking this is because yesterday I upgraded my video card to ATI HD5750 and now I lost the 23.976 (23Hz in CCC) option for my primary monitor. It can't work on such low frequency but nevertheless I used it (and turn the monitor off) because this way I was able to have both my monitors at 23.976 which allowed the composition rate to go down to 23.976 and have stutter free playback with Aero enabled (so no tearing). Sadly now I only get 24.000 (24 in CCC) for the primary and 23.976 and 24.000 for the secondary (EPSON FullHD projector) and if select 24 for monitor and 23 for projector I get stutter, because the composition rate (Aero on) stays at 24.000.

[namaiki]

Sadly now I only get 24.000 for the primary

Sounds like you can only use 24Hz/fps, unless you want to shut off your other screen.

I'm editing this post because I'm not really contributing, but since you don't care if the audio is resampled, I would think that 24.000 is the better option. In this way, ReClock will be very convenient, not to mention that you probably won't notice the difference.

[pankov]

yes currently this is the case, but I just want to clarify if it's worth the effort to try getting back 23.976 ... or it's better to leave it at 24.000

[leeperry]

I just want to clarify if it's worth the effort to try getting back 23.976 ... or it's better to leave it at 24.000
movies are shot at 24.000fps, on the UK DVD of Revolver there's a making of where you can see a monitoring CRT screen used during shooting, it clearly says "24.000 fps" in big...I will try to find the pic again, I posted it on the Reclock forum a while back.

another quote from there:
Many 24p productions, especially those that are made only for TV and video distribution, actually have a frame rate of 24 * 29.97 / 30 frame/s, or 23.976frame/s (24/100.1% to be exact). Many use the term "24p" as a shorthand for this frame rate, since "23.976" does not roll off the tongue as easily. This is because the "30frame/s" framerate of NTSC is actually 30/100.1%, also referred to as 29.97frame/s – this framerate is matched when video at 23.976frame/s has a 3:2 pulldown applied. Similarly, 60i is shorthand for 60/100.1% fields per second.

Film productions may be shot at exactly 24.000 frame/s. This can be a source of confusion and technical difficulties if material is treated as normal video, since the slightly differing framerates can be problematic for video and audio sync. However, this is not a problem if the video material is merely treated as a carrier for material which is known by the editing system to be "true" 24frame/s, and audio is recorded separately from moving images, as is normal film practice.
The nvidia drivers don't allow to make custom timings for both frequencies...so I went for 24.000, as I never watch TV junk :p

Movies are slown down to 23.976 due to NTSC-Film compatibility, but more and more BD's show up at the unmolested 24.000 speed.

[noee]

yes currently this is the case, but I just want to clarify if it's worth the effort to try getting back 23.976 ... or it's better to leave it at 24.000

fwiw, I am lucky enough to have a playback monitor (LG HDTV) that supports 23.976, 24, 59.86 and 60Hz, all at 1080p (and it's S-IPS, fwiw). Coupled with CCC (HDTV options) and the HD2600XT, I am able to test at all of these settings for my given source material (23.976fps and 29.970fps).

My CPU is an AMD X4 oc'ed to 3.5Ghz, so resampling multichannel (good or best) with Reclock is child's play.

Just viewing some testing sequences, I can't tell the difference using Reclock @ 24Hz or not using it at 23.976Hz, except the Reclock resampling does have an impact on CPU usage. madVR gives ultra-smooth playback in both instances.

Running a complete movie is the same. I see at most 2 dropped frames (probably at the beginning) and then 0,0 all the way through with madVR.

I guess I'm just suggesting that it might not matter and that it comes down to what is easier for you to accomplish a consistent experience in everyday use.

HD2600XT, CCC10.3, Win7 x64, Aero OFF, dual-mon (primary 60Hz, sec 24Hz, both 1920x1080)
HMS/FFDshow/Reclock/madVR/MPC-HC
3Dlut OFF|Spline64|Spline64|SoftCubic80
flush|flush and wait(sleep)|don't|don't

[leeperry]

Just viewing some testing sequences, I can't tell the difference using Reclock @ 24Hz or not using it at 23.976Hz, except the Reclock resampling does have an impact on CPU usage. madVR gives ultra-smooth playback in both instances.
If you watch a 24fps movie at 23.976Hz, the sound will be 0.001% too slow...hence the music will be out of tune. You can try to focus and switch back and forth, the sound is audibly too slow in 23.976 :p

It would be like listening to 44.1kHz music at 44.05kHz. This might not matter if you don't use a high resolution/low jitter transport...as your average computer transport has very poor clocking to begin w/: http://hifiduino.wordpress.com/2010/04/08/reading-sample-rate/

m2tech tell more about this "out of tune" problem in that white paper(the Hiface uses two 2.5ppm discrete clocks for 44.1/48kHz multiples...the maximum error will be 0.5Hz at 192kHz!): http://www.m2tech.biz/public/pdf/White%20Paper%20on%20hiFace.pdf

[noee]

@leeperry:
Yes, I agree. I should've been more precise. I watch true 24p material only at 24Hz. My observation was about 23.976fps material @ 24Hz.

[leeperry]

Yes, I agree. I should've been more precise.
No worries, I didn't try to sound patronizing or anything...English isn't my native tongue, so stuff gets lost in translation from time to time. Anyway, most bit-perfect HD movies are out-of-tune :devil:

The only way to play them at the right 24fps pitch w/o resampling is to have one of those grossly overpriced/overkill professional soundcards(RME/Lynx), but then no more Reclock realtime adaptation...so "catch-up" glitches could and will occur as the video clock is very jittery.

BTW, the "Firestone Bravo"(that I own) provides 0ppm error/50ps jitter clocking: http://focus.ti.com/lit/ds/symlink/pll1707.pdf

[DigitalLF]

leeperry then i got a question for you.... should i slave my gfx to my rme hammerfall 9632 audiocard in reclock?

i do have very to little knowledge about the technical stuff around it all... but i really want to learn...

[leeperry]

should i slave my gfx to my rme hammerfall 9632 audiocard in reclock?
slaving to the soundcard is not advisable AFAIK, but you may want to try disabling resampling in Reclock and change the tune in your RME drivers w/ the SteadyClock feature to play 23.976fps audio @24fps speed *w/o* resampling..but I'm quite sure that you'll end up w/ "catch up" glitches once in a while, as the graphic card clock is too jittery. Also, the lipsync won't really be spot-on...there's no free lunch, Reclock addresses all those issues at the price of a resampling pass(which sounds very good upsampled to 88.2/96kHz).

Don't hesitate to check the Reclock forum and ask questions there, as this might not be the right thread to discuss Reclock matters ;)

[Mark_A_W]

You are kidding leeperry.

I can't hear the difference between 24fps and 25fps without an A-B test. That's 4% speedup.


And you are saying you can hear 0.001% ??!! I don't believe you.


24.000 vs 23.976: I don't think it actually matters with Reclock - both are close enough.

[yesgrey]

should i slave my gfx to my rme hammerfall 9632 audiocard in reclock?
Yes. I do it with my FF400 and it works great! Go to RME forum here (http://www.rme-audio.de/forum/viewtopic.php?id=5256), I've put there instructions on how to.

[namaiki]

I can't hear the difference between 24fps and 25fps without an A-B test. That's 4% speedup.

I can tell the difference, but that's only when the video features a song I know.

[nevcairiel]

Hearing the PAL Speedup is possible, especially when you know the original. However most people don't, so they won't ever notice. We haven't had a big public outcry about PAL, did we :D

Hearing the 0.001% NTSC slow-down .. i doubt it. Seriously. Thats just impossible.

[leeperry]

Hearing the 0.001% NTSC slow-down .. i doubt it.
http://www.fact-index.com/p/ps/psychoacoustics.html
changes in pitch larger than 2 Hz can be perceived. However, even smaller pitch differences can be perceived through other means.
playing 48kHz 24fps audio at 23.976fps ends up being played at 47952Hz hah! you have to be tone-deaf to some degree to not hear the difference tbh...Why do you think ppl see the "m2tech Hiface" as the holy grail? because it has 2.5ppm clocks, and plays your music at pretty much the right pitch. Newer interfaces are 0ppm so you do get *spot-on* sample rates. Music is out of tune in 23.976 movies, it just is.

Here's a great test to improve your tone discerning skills, because all the human brain requires is training: http://www.nidcd.nih.gov/tunetest/
We haven't had a big public outcry about PAL, did we :D
US ppl always sound like Donal Duck in PAL.

[Mark_A_W]

There's nothing on that link that is relevant to this discussion.

2hz at what frequency (or octave if you like)? 30hz? Yep, I think I could pick 32hz from 30hz.


2hz at 48khz? I can't hear 48khz. Neither can you.

Even at 1khz, I don't think I could pick 1002hz.


You are not making any sense whatsoever. Please keep it real.

[nevcairiel]

So with a 96kHz Sampling rate the change is even more extrem, yes? :P

The sampling rate doesn't map to the pitch like that. Just because both use "Hz" as their unit doesn't mean they are the same.. :)

But hey, i play my 23.976 Movies at 24Hz these days anyway, Reclock to the rescue! :)

[leeperry]

well, let's take a 15kHz sound...it'll be 14985Hz in 23.976fps, duh. a 15Hz pitch-down is very audible if you A/B them(the human threshold is more or less 2Hz). Anyway, this is OT...we're on a videophile thread where we nitpick until the end of days about video, not quite audiophile :p

"if it sounds good to you, then it's good"...some ppl watch movies in 25fps, some in 23.976..whatever floats your boat.

and even 24/48.000Hz on a PC doesn't mean much, as this is what DirectX sees...but it's prolly off by ±0.002/3Hz or so anyway, that's what some real world experiments have tended to prove.

[Mark_A_W]

So with a 96kHz Sampling rate the change is even more extrem, yes? :P

The sampling rate doesn't map to the pitch like that. Just because both use "Hz" as their unit doesn't mean they are the same.. :)

But hey, i play my 23.976 Movies at 24Hz these days anyway, Reclock to the rescue! :)


I know I was taking the piss.

I know about Nyquist ;)

[SundaY82]

Playing mkv files with 0.22 makes video freeze after a few seconds on my system. Video freeze but audio continues and I can access menus and close player without problem. Tested with both ffdshow and coreavc. 0.20 does not have this problem and avi(xvid) files play fine with 0.22 also.

Created a log file, you can get it here: http://www.multiupload.com/SAKF1MPXTP

System:
Windows 7 x64, aero on
Nvidia 240GT
Reclock, Zoom Player
If you want more info let me know.

[Kamus]

Problem since i switched to nvidia...

I've been using madvr for a long time now on my windows 7 64 bit machine + MPC-HC.
Ever since i switched to nvidia i haven't been able to use madvr at all.
I switched from a 4870x2 to a 265 GTX.
It was a loaner until the 480 GTX went on sale, which i've now used since it came out.
Basically what happens is MPC won't play anything if madVR is the renderer, it just hangs right after i open any video file, any other renderer seems to work fine, and it used to work just fine back when i had the 4870x2.

I've tried numerous nvidia drivers, none of which work. (right now i'm using the latest beta, but i've been trough at least 5-7 driver releases since i switched, no luck what so ever though)

If anyone has a hint, i'll take it.
Thanks in advance.

Also, i'm not 100% sure if it did work on the 265, to be honest i simply forgot, but actually i think this happened as soon as i got the 480.

[nevcairiel]

@Kamus:
I had the same problem once. Reset all your MPC-HC settings (uninstall, reinstall, make sure to deselect all "keep settings" checkboxes). It solved it for me, back then!

[madshi]

Tried to play 1080i50 H264 film based content using CoreAVC (set to No interlacing (weave))+ MADVR. Display is set to 1080p50.
PQ is very good. No deinterlacing artfacts.
Does that mean that CoreAVC+Madvr restore original 25p film source from 1080i50 'container'?
And we do not need advance hardware vector adaptive deinterlacing for films in 1080i50 format?
European broadcasters use 1080i50 for movies and it seems that the flags are always set correctly. which means that you can simply set the decoder to "weave" and you'll automatically get perfect 1080p25. I've been using this setting in CoreAVC for years. I'm playing back all my 1080i50 h264 movies this way. Works perfectly.

"Vector adaptive deinterlacing" is made for video content, not for movies. For movies you just need to find the right fields and weave them together. For 60i that is called IVTC (inverse telecine).

Revision 3508 of ffdshow now adds adds madVR to list of compatible renderers for dynamic format changes.
Nice!

Feature request:

If the video is going to be upscaled by a factor or number of pixels less than [set by user], do not resize at all.
For example, when playing a 1916x800 video on a 1920x1080 screen, leave 2 pixels wide black bars on each side of the video instead of upscaling.
I agree that this would be nice. However, ultimately the media player controls exactly how scaling is done. madVR just tells the media player which aspect ratio is "preferred". But then the media player tells madVR into which window size to scale exactly. So your feature request should be posted in the support thread of your favorite media player.

That said, madVR could eventually overwrite the media player's wishes. But that's a feature that is way down my priority list right now. So many more important things I need to implement first.

This latest video I'm encoding isn't working with madVR. It just draws a black screen, no video at all. It gets the input right but it doesn't draw it.
Have you tried different decoders? Does it work with all other renderers? If so, a sample would be helpful.

Hi, just want to report a problem with the newest version .22. Using XP SP3 and Zoom Player, mouse functions don't work inside the viewing area whether in window mode or fullscreen.
Will be fixed in v0.23.

Madshi, while doing some random testing (unrelated to AR), I've came across a regression of sorts.

To reproduce on my end:
I play any h264 video, so that CoreAVC2 is utilized in MPC-HC (r2116).
While the video is playing, I open the CoreAVC2 Filter configuration, modify any setting (say, input level for example) & hit apply.
After doing so, the video will freeze, meanwhile the audio continues in the background.
Will be fixed in v0.23.

I prefer ATI right now because:

(1) The "power per watt" ratio is clearly better, which is very important for me (because HTPCs are usually thermally limited).
Well, would that still be the case if you took CoreAVC CUDA in account?
The NVidia GTX 460 has a TDP of 150 Watts. The Radeon 5750 has a bit less memory bandwidth, but seems to have a bit more shader power (at least more GFLOPs) and has a TDP of only 86 Watts. That is an extremely big difference IMHO. And I don't think that CUDA will make up for that. But that's just my personal opinion.

madshi,
Can you make the renderer downsize anything greater than or equal to the screen resolution to like 1/2 the original size? I like the way Haali Renderer does that.
I never really liked how Haali did that. Maybe I could add that as an option somewhen in the future, but not anytime soon. So much more important stuff on my to do list.

madshi,
last night I upgraded my video card to ATI 5750 and I was pretty happy with the results until suddenly an hour or so into a recorded basketball game (720p@59.940) it started dropping frames.
Everything was butter smooth before this.
I tried everything - changing settings in madVR, changing back to v0.21, removing ReClock because I've upgraded it to 1.8.7.0 the same evening, changing refresh rates and going back to 59.970 but sadly nothing helped.
Since the problem occurred it was present on both monitors (2x 1920x1080 @ 59.940).
I've managed to make a log file for you
http://www.mediafire.com/?iiymnwmzmht
As far as I noticed the problem was in the high rendereing times (max (5s)) but even after I lowered the settings to Nearest Neighbor the dropping continued.
I checked with GPU-Z and the GPU load was ~35% with Spline64 at all scaling options ... and there was no fluctuations in the GPU/Mem frequencies.
The only thing that helped was a restart of the computer.
So the question now is - is my new video card somehow defective or it's something else? But what?
(I've check the CPU usage and the list of processes and there wasn't anything out of the ordinary)
5 seconds, as in 5000 milliseconds? Are you serious???

If a reboot cured the problem, then I don't think it's a hardware problem. But I don't really know for sure. Can you reproduce it?

I never really paid attention(due to the previous OSD color being too dark) to the amount of jitter you get when switching resolutions..yay! it usually starts at 96.00033Hz for me...then hiccups very quickly between ~33/~25, and then stabilizes after a few mins and slowly decrements to reach 96.00015Hz 90 mins later.
The madVR refresh rate information is based on doing some clever math and a lot of educated guessing. The only reason why it stabilizes after a while is that madVR collects more data and can based on that do a more exact guess. I don't think this has much to do with jitter. The information madVR gets is simply not exact enough for madVR to guess exactly right on the first try.

Playing mkv files with 0.22 makes video freeze after a few seconds on my system. Video freeze but audio continues and I can access menus and close player without problem. Tested with both ffdshow and coreavc. 0.20 does not have this problem and avi(xvid) files play fine with 0.22 also.

Created a log file
Can't see what's wrong in the log. The decoder suddenly stops sending frames, but I don't know why. Please try again with v0.23. Can you still reproduce it with v0.23? If so, does it happen with all videos or only with some? If only with some, can you try to make a small sample with which I could reproduce the problem on my PC?

Problem since i switched to nvidia...

I've been using madvr for a long time now on my windows 7 64 bit machine + MPC-HC.
Ever since i switched to nvidia i haven't been able to use madvr at all.
I switched from a 4870x2 to a 265 GTX.
It was a loaner until the 480 GTX went on sale, which i've now used since it came out.
Basically what happens is MPC won't play anything if madVR is the renderer, it just hangs right after i open any video file, any other renderer seems to work fine, and it used to work just fine back when i had the 4870x2.
Need a log file.

[madshi]

madVR 0.23 released

http://madshi.net/madVR.zip

* fixed: some mouse events didn't work in ZoomPlayer, anymore
* fixed: when changing CoreAVC settings, madVR video stopped updating
* fixed: when ffdshow needed bigger buffers, there was a crash
* made OSD half transparent again (for now)
* removed "use managed 3dlut texture" option