Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 1st June 2013, 01:10   #18921  |  Link
karamancho
black stain
 
karamancho's Avatar
 
Join Date: Apr 2013
Posts: 46
avg. 100 on SD, avg. 120 on 720p
karamancho is offline   Reply With Quote
Old 1st June 2013, 01:21   #18922  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,406
Well that makes sense, even just one frame of 8bit RGB 720x480 video is 8.3MB so you don't have enough video RAM for the present queue.
Asmodian is online now   Reply With Quote
Old 1st June 2013, 07:26   #18923  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by sgraves66 View Post
Completely understand. I doubt that mine will ever be seen outside of my home, anyways. It's taken years to develop and is constantly in motion. I'd need more developers to make it publicly available.

put_MessageDrain didn't seem to improve anything. madVR does work fine in most cases, but the mouse wheel seems to give me the most issues, plus some overlapping keyboard shortcuts between the frontend and madVR. The later is easily solved and there may be an issue within SDL 2.0 causing the intermittent loss of the mouse wheel. Disabling the window just seems to care of all these issues in a simple manner.

The fronted is primary controlled by bluetooth remote and/or joystick, anyways, so it's rare that a mouse is actually used.
Well, I guess you could create a thread which disables and de-focuses your window once every second. Wouldn't that workaround the problem?

Quote:
Originally Posted by leeperry View Post
I only used ffdshow to check the levels, the very same problem occurs with LAV alone.....apparently it's due to my gamut mapping PS scripts(used as post-scaling in PotP).

Could it be that your BTB/WTW expanding "trick" gives the aforementioned issue if there originally was data in those areas somehow? Or maybe it's just not compatible with what this PS script expects? How could I fix it then please?

If I don't enable this PS script, then indeed PC is lighter than TV on this sample and this PS script works fine on 16-235 content.
Don't know why the script would make problems. Do the same problems occur when using the same script in MPC-HC?

Quote:
Originally Posted by nekromantik View Post
Yes I tried that.
That breaks playback, I get blank screen but with audio.

With windowed mode and AMD card I still get around 50 - 60ms rander times and its jittery and lots of frame drops.
The card is not been utilized properly, gaming works fine. I think I need to stick with HD4000.
That sounds weird to me. So it seems your AMD card is slower than the HD4000 when using madVR, is that correct? But is it (much) faster than the HD4000 for gaming? I don't really understand why it would be slower for madVR, but faster for gaming, because madVR mostly works similar to a game...

Does your AMD card have on-board RAM for video? Or does it use a part of your system RAM? How much RAM does it have? Maybe it's running out of RAM, similar to problem karamancho seems to have?

Quote:
Originally Posted by Asmodian View Post
Well that makes sense, even just one frame of 8bit RGB 720x480 video is 8.3MB so you don't have enough video RAM for the present queue.
Yeah, that makes sense! @karamancho, try to decrease the GPU queue and present queue to lower values. E.g. try using 4 for both. Then check how much RAM is used.
madshi is offline   Reply With Quote
Old 1st June 2013, 07:26   #18924  |  Link
petran79
Registered User
 
Join Date: Aug 2007
Posts: 87
Quote:
Originally Posted by leeperry View Post
In order to revert to the old rendering path, you need to uncheck "present several frames in advance" in the /rendering/exclusive mode settings/ tab and it would also help to check "use managed upload textures (XP only)" in the /rendering/general settings/ tab as that latter option would smoothen windowed/FS changes IME. I've tried the new rendering path on several occasion but it would very often hiccup right after skipping when OTOH the old one has always worked perfectly for me.

you also seem to imply that CUVID hardware acceleration in LAV would be part of the issue, so how about software decoding then?

some additional feedback would be greatly appreciated please if you got time to run more tests.
Unfortunately video continues to show occasional frame drops, even after checking those options.

Usually that happens a few seconds after loading the video, then video runs smooth, then another frame drop occurs after few seconds.

the odd thing is that it happens on particular video parts, usually in the part where bitrate peeks. Even if I start the video exactly from where the drop occured, it happens again.
When I set the slider to a part with lower bitrate, video plays smoothly.

Even with these options enabled or disabled, same thing happens again.
petran79 is offline   Reply With Quote
Old 1st June 2013, 07:30   #18925  |  Link
truexfan81
Registered User
 
truexfan81's Avatar
 
Join Date: Nov 2012
Posts: 138
madshi i am trying to respond to you on the bug tracker but its not working, i type a response, click the "add note" button and nothing happens

edit: i wonder if this is a google chrome bug, will try firefox, see if it works

Last edited by truexfan81; 1st June 2013 at 07:34.
truexfan81 is offline   Reply With Quote
Old 1st June 2013, 07:37   #18926  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by petran79 View Post
Unfortunately video continues to show occasional frame drops, even after checking those options.

Usually that happens a few seconds after loading the video, then video runs smooth, then another frame drop occurs after few seconds.

the odd thing is that it happens on particular video parts, usually in the part where bitrate peeks. Even if I start the video exactly from where the drop occured, it happens again.
When I set the slider to a part with lower bitrate, video plays smoothly.

Even with these options enabled or disabled, same thing happens again.
This sounds like the decoder (= your CPU?) is not fast enough. Enable the madVR OSD (Ctrl+J) and check the state of the decoder queue when the frame drop occurs. Is it nearly empty in that situation? Or is it near full?

Quote:
Originally Posted by truexfan81 View Post
madshi i am trying to respond to you on the bug tracker but its not working, i type a response, click the "add note" button and nothing happens

edit: i wonder if this is a google chrome bug, will try firefox, see if it works
Please make sure you're logged in. You can read the bug tracker without being logged in, but you have to be logged in to change anything.
madshi is offline   Reply With Quote
Old 1st June 2013, 07:40   #18927  |  Link
truexfan81
Registered User
 
truexfan81's Avatar
 
Join Date: Nov 2012
Posts: 138
Quote:
Originally Posted by madshi View Post
This sounds like the decoder (= your CPU?) is not fast enough. Enable the madVR OSD (Ctrl+J) and check the state of the decoder queue when the frame drop occurs. Is it nearly empty in that situation? Or is it near full?


Please make sure you're logged in. You can read the bug tracker without being logged in, but you have to be logged in to change anything.
i was logged in, it apparently does not like google chrome, i was able to comment using firefox
truexfan81 is offline   Reply With Quote
Old 1st June 2013, 07:45   #18928  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Using chrome myself. Works fine here.
madshi is offline   Reply With Quote
Old 1st June 2013, 07:48   #18929  |  Link
truexfan81
Registered User
 
truexfan81's Avatar
 
Join Date: Nov 2012
Posts: 138
oh madshi i have been having a weird issue lately, but i think maybe its an nvidia bug. every couple of days madvr reports "gpu/driver does not support overlays." only way to fix it is to reinstall the nvidia driver and then reinstall madvr (a reboot does not even make the error go away) this has happened with nvidia 320.14 and 320.18. so i'm tempted to revert back to 320.00 to see if the problem goes away.
truexfan81 is offline   Reply With Quote
Old 1st June 2013, 08:42   #18930  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Sounds like a driver problem to me. madVR asks Direct3D whether overlay is supported. Seemingly in your case sometimes Direct3D thinks/reports it's not supported, for whatever reason...
madshi is offline   Reply With Quote
Old 1st June 2013, 09:30   #18931  |  Link
pie1394
Registered User
 
Join Date: May 2009
Posts: 212
Quote:
Originally Posted by madshi View Post
That sounds weird to me. So it seems your AMD card is slower than the HD4000 when using madVR, is that correct? But is it (much) faster than the HD4000 for gaming? I don't really understand why it would be slower for madVR, but faster for gaming, because madVR mostly works similar to a game...
That indeed sounds weird..


HD8730M is 384 cores @ 650 MHz 128 Bit @ 1000 MHz DDR3
It should be more than enough for Lanczos3 scaling if deinterlacing job is not required.


As I know, some NB vendors choose the cost-down solution to wire some particular output port(s) like LVDS / TMDS to only 1 GPU, not both via the switching IC. The usual design is always to use Intel GPU to send out all video signals while AMD/nVidia GPU is the active one. So extra framebuffer copy back from dGPU to iGPU is always required and it creates extra latency.

Maybe it would work if all FLUSH settings are turned OFF?
pie1394 is offline   Reply With Quote
Old 1st June 2013, 10:16   #18932  |  Link
nekromantik
Registered User
 
Join Date: Feb 2013
Posts: 96
Quote:
Originally Posted by pie1394 View Post
That indeed sounds weird..


HD8730M is 384 cores @ 650 MHz 128 Bit @ 1000 MHz DDR3
It should be more than enough for Lanczos3 scaling if deinterlacing job is not required.


As I know, some NB vendors choose the cost-down solution to wire some particular output port(s) like LVDS / TMDS to only 1 GPU, not both via the switching IC. The usual design is always to use Intel GPU to send out all video signals while AMD/nVidia GPU is the active one. So extra framebuffer copy back from dGPU to iGPU is always required and it creates extra latency.

Maybe it would work if all FLUSH settings are turned OFF?
Yes the actual output to the screen is via the Intel card, the AMD is only there as a processor type card. So all data goes through 2 cards to get outputed. Will try no flushing.
Do you think Jin3 would be too much for the card?
As with Intel I can do Jinc3 image and Softcubic no AR 100 on chroma.
Thanks

Last edited by nekromantik; 1st June 2013 at 10:32.
nekromantik is offline   Reply With Quote
Old 1st June 2013, 12:48   #18933  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by madshi View Post
Don't know why the script would make problems. Do the same problems occur when using the same script in MPC-HC?
The script works fine with that 0-255 sample in EVR using PotP 1.5.3776 and MPC-HC 1.6.7.7069 but it looks equally too bright with madVR in both players...strangely the problem doesn't occur on 16-235 source files, it really only happens when they are 0-255(even if it's just padding data and not actual video content)....which would explain why I've randomly encountered this problem

Hope you can look into it and lemme know if I should add it to the bug tracker or something,

Quote:
Originally Posted by petran79 View Post
Even with these options enabled or disabled, same thing happens again.
I've got a remuxed BD with high bitrate that would exhibit the problem you're describing but I can force software decoding with ffdshow(libavcodec) instead of LAV/CUVID and presto! all is well.....the GPU doesn't take care of the video decoding anymore so its bitrate doesn't matter, doesn't that fix the issue for you as well

Last edited by leeperry; 1st June 2013 at 13:19.
leeperry is offline   Reply With Quote
Old 1st June 2013, 14:46   #18934  |  Link
pie1394
Registered User
 
Join Date: May 2009
Posts: 212
Quote:
Originally Posted by nekromantik View Post
Yes the actual output to the screen is via the Intel card, the AMD is only there as a processor type card. So all data goes through 2 cards to get outputed. Will try no flushing.
Do you think Jin3 would be too much for the card?
As with Intel I can do Jinc3 image and Softcubic no AR 100 on chroma.
Thanks
Without deinterlacing job and anti-ring option, Jinc3 might be ok up to 30 fps on HD8730M.

I have a HD7970 which is 950 Mhz 2048 GCN core + 384-bit DDR5-5500. Its performance should be at least 7 times as fast as HD8730M's.

The estimated maximum performance for 1440x1080i content to 1920x1080p with Luma/Chroma Jinc3+AR scaling is about 120~150 fps and 60 interlaced frames' vector-adaptive deinterlacing job per second.

In reality, there is no such content. Unless madshi designs some other useful features (any in-progress? ) which require more computation power than Jinc3+AR mode, the GPU with near 1000 GCN / Kepler CUDA cores, or 384 Fermi CUDA cores should be enough for real-world contents up to 4K 60 fps.

Last edited by pie1394; 1st June 2013 at 15:03.
pie1394 is offline   Reply With Quote
Old 1st June 2013, 15:49   #18935  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
For those having problems with newer Intel GPU drivers...

It seems that Intel drivers 9.18.10.3071 (and .3111) have a bug which occurs when using the madVR option "use a separate device for presentation". You have 2 choices: Either go back to older drivers. Or disable that madVR option. FWIW, I've notified my Intel contact about this, but I've no idea how fast they'll fix this.

Quote:
Originally Posted by leeperry View Post
The script works fine with that 0-255 sample in EVR using PotP 1.5.3776 and MPC-HC 1.6.7.7069 but it looks equally too bright with madVR in both players...strangely the problem doesn't occur on 16-235 source files, it really only happens when they are 0-255(even if it's just padding data and not actual video content)....which would explain why I've randomly encountered this problem

Hope you can look into it and lemme know if I should add it to the bug tracker or something
From what I can see madVR does its job correctly. Fullrange content is converted properly *before* the custom pixel shaders are executed. Do you have any "trade quality for performance" options activated? If so, try to check whether unchecking them helps.

You could also try whether replacing "float4 c0 = tex2D(s0, tex);" with "float4 c0 = tex2D(s0, tex); saturate(c0);" makes any difference. Probably not, but maybe worth a try...
madshi is offline   Reply With Quote
Old 1st June 2013, 16:02   #18936  |  Link
DragonQ
Registered User
 
Join Date: Mar 2007
Posts: 934
Most web pages relating to drivers for "3rd Generation Intel Core Processors with Intel HD 4000/2500 Graphics" are error pages anyway, for me at least. Their automatic updater never does anything either.

I'm sure I managed to find the download page for the 9.18.10.3071 drivers last week via Google but I didn't save it lol.
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7
DragonQ is offline   Reply With Quote
Old 1st June 2013, 16:10   #18937  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
3071 for Ivy Bridge is here:
x86: https://downloadcenter.intel.com/Det...&DwnldID=22610
x64: https://downloadcenter.intel.com/Det...&DwnldID=22605
__________________
LAV Filters - open source ffmpeg based media splitter and decoders

Last edited by nevcairiel; 1st June 2013 at 16:12.
nevcairiel is offline   Reply With Quote
Old 1st June 2013, 16:12   #18938  |  Link
DragonQ
Registered User
 
Join Date: Mar 2007
Posts: 934
So what's the difference between 9.17.10.3071 and 9.18.10.3071? Is one for desktops and one for laptops?
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7
DragonQ is offline   Reply With Quote
Old 1st June 2013, 17:24   #18939  |  Link
ajp2k11
Registered User
 
Join Date: Jul 2011
Posts: 57
Quote:
Originally Posted by DragonQ View Post
So what's the difference between 9.17.10.3071 and 9.18.10.3071? Is one for desktops and one for laptops?
Some unreleased Intel/AMD stuff here...

asder00.blogspot.se
ajp2k11 is offline   Reply With Quote
Old 1st June 2013, 19:35   #18940  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by madshi View Post
Do you have any "trade quality for performance" options activated? If so, try to check whether unchecking them helps.

You could also try whether replacing "float4 c0 = tex2D(s0, tex);" with "float4 c0 = tex2D(s0, tex); saturate(c0);" makes any difference.
OK, thanks for the help but I don't have any "trade quality for performance" option checked and changing those lines didn't help either

BTW, problem is identical whether it's used before or after scaling.

Last edited by leeperry; 1st June 2013 at 19:42.
leeperry is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 20:03.


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