MP1 EVR Presenter/dshowhelper community development (12 Viewers)

[robyf]

Hi Robyf, (I have filled in my system specs)

I didn't get any joy using ReClock but can you confirm :

I installed it and it seemed to prompt me and ask me as to whether it should work with Mediaportal - I answered Yes.

I have not explicitly set it as an audio codec in Mediaportal settings - should I?

For vsync - on the Video Tab, there are 3 tick boxes under Vsync - I take it I should untick all 3?

I don't understnad why I get considerable judder in Mediaportal but not in my tests in GraphStudio! - not complaining, but for me, the judder seems to have been a more recent occurence.

Thanks for your help/guidance.

Yes as said by cecet you have to set reclock as audio renderer in codecs configuration of mediaportal (do it only for videos, not tv). By the way have you also stuttering with live tv?

As for the codecs I suggest you to not use ffdshow for video.

Try this configuration:

TV
Mpeg2: Microsoft
H.264: Microsoft

Videos
Mpeg2: Microsoft
H.264: MPC-HC

Do not use ffdshow for spdif passthrough, it sometimes gives problem, prefer ac3filter instead.

 

[Jay_UK]

robyf,

Good heads up/info..... but you can't use passthru with reclock?

Thanks,

J.

 

[webs0r]

Hi jet70 if that is the case, then I suspect Graphstudio is rendering a different graph. Note that MP uses its own modified EVR renderer, so that is always going to be different.

Are you using that setting in MP to use directshow to automatically build the filter graph? In this case, the rest should be the same & it will just be the video renderer that is different (maybe the audio depending on ur settings but that doesn't seem to be your problem). What video renderer is graphstudio using? If it is using stock EVR, how does the vanilla MP EVR dll work for you?

To check if reclock is being used by Mediaportal, use the reclock hotkey(s) to test while playing video through MP. Its something like ctrl-alt-shift-f9 - check the config app. You'll get a tearing bar etc.

 

[doornjoostje]

I've tried all versions, to see if it works better than RC4.
I used it without reclock.
v2 and v4 for me are the best, way better picture quality then RC4.
I use v2 now.
The picture looks as good as version 1.0.2.

So Owlsroost, for this!

 

[robyf]

robyf,

Good heads up/info..... but you can't use passthru with reclock?

You can if you enable "Accept AC3 sources" but you will not resample audio so you have to use a refresh rate multiple of fps media for that. Or you can resample audio and reencode in AC3 but I will not suggest it (in my experience quality is not good) and you also loose DTS.

 

[edterbak]

Hi all,

I have been trying numerous combinations to get propper playback. I am (as it seems) unable to get smooth playback, but also after 10+ minutes with various mediafiles avi/mkv files. Especially with HDTV (1080i) I am having a lot of issues. First 4 seconds of HDTV are smooth and fine... then the judder starts.

I am running Win7 x64; GPU is a HD4350; CPU is AMD BE-2350 (2.1ghz dualcore). The HD4350 is capable of dxva 2 and deinterlacing up to adaptive (I can select vector adaptive and it works IMO, but some people doubt it will actually do it). At first I was steared to think the limit of the card is causing these issues. but the more I look at it, I am starting to believe it is not.

If I disable Aero I have a perfect picture, up untill some tearing starts to occur after 30-180 seconds. This is ofcourse due to the lack of aero v-sync correction. CPU load is 25-40% and GPU load is 55% with weave de-interlacing and 75-80% with vector-adaptive de-interlacing set in ATI CCC. If I press SHIFT+1 I see actual framerate (red line) of 50fps, like it should. No HW bottleneck IMO.

If I enable Aero I have 4-6 seconds of perfect picture. After that I see one or two constant framedrop(s) each second. CPU load is still 25-40% and GPU load is 55% with weave de-interlacing and 85% with vector-adaptive de-interlacing. There seems to be no significant load increase to cause judder. Especially if I look at the weave de-interlacing load on the GPU (55%) which should give a good picture. When pressing SHIFT+1 I always see a actual framerate (red line) which drifts between 25-45fps while it should be 50. windows framerate is 50.

I do see some improvements if I switch dshowhelper.dll.
Original: good playback on normal mediafiles; HDTV judder like crazy
V1+2+4: good playback on normal mediafiles (a rare judder); HDTV judder slightly less but still very unacceptable.
v10+11: sometimes judders on normal mediafiles; HDTV judder even slightly less compared to v1/2/4, can apear smooth at times, but it is still not accepable / watchable.

What have I tried:
-Inside ATI CCC; V-sync always on/off, force smooth playback, de-interlacing weave/bob/>/vectoradaptive.
-Codecs; like SAF4/5, Powerdvd 9/10 with DXVA, FFDShow video DXVA, MPC-HC, ATI AVIVO...
-Tried ATI catalyst 10.3/10.4/10.5 and now I use 10.6beta2.
-I have played with reclock as well. The audio output I use is analogue, with ASUS Xonar to convert on the fly to DTS Live.
-Inside MP; dont drop late frames, different timeshifting modes, different timeshifting locations (now on SSD)
Best playback results I get with Powerdvd 9/10 with the use of dxva. ATI catalyst 10.6beta2 seems to work best in Win7/MP, but no change on judder.

I have a recording of a channel (Nederland 1 HD - MPEG2 1080i) and it plays fine outside MP in Graphstudio/MPC-HC. Inside MP not.
Some other HD channels show a LOT less judder than Ned1 HD. Some are even smooth for 99% of the time, but these might be 480/720p or something (didnt check).

So I am inclined to believe there is an other issue at hand here than a hardware limitation. Or atleast, I dont see a CPU/GPU bottleneck during playback especially forcing weave de-interlacing.
I would like to know what your thoughts are about this. Is it dshowhelper related? Settings ATI CCC related.
Also, If you like to see more detailed logs can you tell me what actions you'd like me to show. I would like to squash this thing which is bugging me for a while now.
How do I go about this best? Please advice.

Kind regards,

 

[Owlsroost]

It's that time again....

New dll version - v0012 - and source code patch file attached to first post

Tony

 

[red5goahead]

It's that time again....

New dll version - v0012 - and source code patch file attached to first post

Tony

This Dll is very good for me because on my system in some tests few weeks ago there was a problem as I wrote when video frames and video refresh rate match 1:3 ratio so 25 fps on a 75 Hz. monitor (stuttering, shift-1 info windows with a huge numebr of red line peaks). Now with this dll no problem at all. I remember to you my test is with 24 fps mkv h264 ac3 speedup-ed with reclock to 50-75 hz monitor. my plasm work at 50 HZ. no problem at all also with this dll and 50HZ refresh rate (1:2 ratio).

 

[tourettes]

and the sample buffers have been increased from 5 to 10 - plus a few other minor changes. Note - early feedback is showing problems on low-end GPUs at least...

1920x1080x32*10 / 1024 / 8 which is approx 80MB. On top of that MP textures and video decoding memory requirement + AERO. I'm quite sure that the GPU memory usage is causing at least some of the trouble low end GPU users are seeing. It would be better to base the amount of frames to the available / free GPU memory. Some care needs to be taken since it could be that Vista / Windows 7 report the virtual GPU memory as well (that is something we really don't want to use with video playback ).

 

[Owlsroost]

and the sample buffers have been increased from 5 to 10 - plus a few other minor changes. Note - early feedback is showing problems on low-end GPUs at least...

1920x1080x32*10 / 1024 / 8 which is approx 80MB. On top of that MP textures and video decoding memory requirement + AERO. I'm quite sure that the GPU memory usage is causing at least some of the trouble low end GPU users are seeing. It would be better to base the amount of frames to the available / free GPU memory. Some care needs to be taken since it could be that Vista / Windows 7 report the virtual GPU memory as well (that is something we really don't want to use with video playback ).

I've added a version (v0013) to the first post which uses 5 sample buffers (same as the standard dll) instead of 10. It's otherwise identical to v0012.

(The problems one early user reported with v0010 were actually caused by ReClock vsync correction being enabled.....)

Tony