View Full Version : madVR - high quality video renderer (GPU assisted)
jmonier
13th March 2015, 13:24
That's basically copy pasted from what I linked. Didn't answer my question if it is better than a global calibration following the dispcal guide.
As nevcairiel says, the avsforum thread is the original. dispcalgui has changed quite a bit since the last time the avsforum instructions were updated so you have to take that into account.
Whenever the instructions are brought up to date, it will be in the avsforum thread.
n3w813
13th March 2015, 17:56
It's been 2 years since I last created the 3DLUTs for my TVs. I think it is a good time to update them and see if the new software versions can improve the picture quality. I will update the AVSForum thread with updated steps soon. :)
Asmodian
13th March 2015, 18:16
That's basically copy pasted from what I linked. Didn't answer my question if it is better than a global calibration following the dispcal guide.
Usually, yes. A global calibration is 3 1DLUTs that runs on your video card. All that is calibrated (in Windows) with a monitor profile (global calibration) is the white point, gray scale, and gamma.
A 3DLUT can also change the gamut and is generally a much more complete and accurate calibration.
Arm3nian
13th March 2015, 21:10
Actually your link is copy-pasted from the forums. Thats not an official or endorsed website for madVR.
As nevcairiel says, the avsforum thread is the original. dispcalgui has changed quite a bit since the last time the avsforum instructions were updated so you have to take that into account.
Whenever the instructions are brought up to date, it will be in the avsforum thread.
I know, looking at the dates, the avs thread was created more than a year prior to the post on the fake madvr site. The OP isn't updated with the info from the following 96 pages so it's the same thing as I linked.
Usually, yes. A global calibration is 3 1DLUTs that runs on your video card. All that is calibrated (in Windows) with a monitor profile (global calibration) is the white point, gray scale, and gamma.
A 3DLUT can also change the gamut and is generally a much more complete and accurate calibration.
So windows can't actually use all of the information provided from the calibration process. Pressing profile and calibrate in dispcal just creates an icc profile and loads it. The calibration does however give you the ability to create a 3dlut file that has the original values to use with other programs.
My original question was if there is a difference from creating the 3dlut file from my settings (without using madtpg) compared to the settings provided on the guide. My settings are optimized for my monitor/viewing conditions but the guide's settings are optimized for madvr/video.
Asmodian
13th March 2015, 22:09
My original question was if there is a difference from creating the 3dlut file from my settings (without using madtpg) compared to the settings provided on the guide. My settings are optimized for my monitor/viewing conditions but the guide's settings are optimized for madvr/video.
Since we have no idea what your settings are we cannot answer that. Specific calibration questions should probably not be in this thread though, it is too easy to add a lot of detail that isn't specific to madVR. :o
I do get the same results using madTPG or Argyll's pattern generator using the same settings.
baii
13th March 2015, 22:45
I find that 3d out created from "global profile" works fine as long as the profile is "good enough" ( enough test patches ") . you can always make the 3dlut then verify it using madtpg and said 3dlut enabled.
As previous quoted, calibration (video card one, not osd/hardware) may not be needed. I recently did experiment using such method and verification result come back perfect.
Sent from my 306SH
huhn
14th March 2015, 13:27
It's been 2 years since I last created the 3DLUTs for my TVs. I think it is a good time to update them and see if the new software versions can improve the picture quality. I will update the AVSForum thread with updated steps soon. :)
i guess you wait for the new dispcalGUI version now?
garson
18th March 2015, 01:31
Hi.
I have nvidia GT730, XP 32bit and latest driver (347.88). MPC-HC with madVR gives black screen. Audio is present, playback starts, and ends in 5-10 seconds.
With system default renderer playback works fine.
Before GT730 I had GT520 and this worked. So I've changed only graphics card.
Any suggestion? How to debug this?
leeperry
18th March 2015, 02:29
I have nvidia GT730, XP 32bit and latest driver (347.88). MPC-HC with madVR gives black screen. Audio is present, playback starts, and ends in 5-10 seconds.
With system default renderer playback works fine.
Before GT730 I had GT520 and this worked. So I've changed only graphics card.
Newest nvidia XP drivers are completely broken with mVR and their newest chips, old boards usually work fine as drivers were not updated. Your best option is to bite the W7/W8 bullet or put back the old board IMO.
garson
18th March 2015, 10:17
Newest nvidia XP drivers are completely broken with mVR and their newest chips, old boards usually work fine as drivers were not updated. Your best option is to bite the W7/W8 bullet or put back the old board IMO.
I've tried oldest driver for GT730 (a year old) and problem was still there. :(
This is my secondary PC (used just as HTPC), it's not that easy to switch to W7/8. :(
leeperry
18th March 2015, 14:23
I've tried oldest driver for GT730 (a year old) and problem was still there
This problem has been there for a few years now, they have broken the VSYNC detection and they do no care.......can't really blame them, XP is utterly obsolete and starting from Vista you'll benefit from a much tighter timer source.
mVR always forces the highest timer resolution and will force 0.5ms granularity when XP is stuck at 9.7ms(and a lot more if you don't tweak your boot.ini), Windows has never been a realtime OS but XP was really not meant for A/V. Exclusive mode WASAPI and 0.5ms timer are totally worth the hassle to upgrade IME :)
kalston
19th March 2015, 10:35
This problem has been there for a few years now, they have broken the VSYNC detection and they do no care.......can't really blame them, XP is utterly obsolete and starting from Vista you'll benefit from a much tighter timer source.
mVR always forces the highest timer resolution and will force 0.5ms granularity when XP is stuck at 9.7ms(and a lot more if you don't tweak your boot.ini), Windows has never been a realtime OS but XP was really not meant for A/V. Exclusive mode WASAPI and 0.5ms timer are totally worth the hassle to upgrade IME :)
Uh? My old XP laptop runs with a 1.25ms clock just fine, not 0.5 sure but still good enough. And that's a laptop with shitty hardware I'm sure XP can do 0.5 with proper hardware (I think I can even push 0.95 or something on that specific machine with the timertool.exe). And XP has Kernel Streaming and ASIO and they do the job as far as I'm concerned (not that I use XP a lot these days anyway)
TheDarkTemplar
20th March 2015, 00:26
Question: Why is there a Winamp logo in the bottom left of my MPC-HC whenever I run a file and is this bad?
http://i.imgur.com/ftTJfu1h.jpg
It says 'software decoding' when I hover over it. Should I be using some sort of hardware decoding?
vivan
20th March 2015, 00:52
This is not related to madVR, but to MPC-HC (or even Windows). MPC-HC displays icon of the program associated with this filetype.
XMonarchY
21st March 2015, 01:59
Does anyone know of advantages or improvements of ArgyllCMS 1.7.0 Beta over the older ArgyllCMS version 1.6.3? Does it result in more accurate 1DLUT's or 3DLUT's for madVR? 1.7.0 Beta is the latest release, is it not?
Ver Greeneyes
21st March 2015, 03:01
There are lots of small changes, and a few flags that might make it easier to achieve a good black level for calibration. Nothing too astronomical though.
aracnoz
21st March 2015, 12:35
hi
I'm trying to integrate madVR in kodi, i made some important progress
I'm stuck on a stupid thing, the mouse events don't pass from madvr to kodi
this was my code
if (m_pVideoWindow)
{
HRESULT hr;
m_pVideoWindow->put_Owner((OAHWND) g_hWnd);
m_pVideoWindow->put_WindowStyle(WS_CHILD | WS_CLIPSIBLINGS | WS_CLIPCHILDREN);
m_pVideoWindow->put_MessageDrain((OAHWND) g_hWnd);
}
in this way, the kodi window is inaccessible, I can not even do a resize or move it ... obviously not pass keyboard events and mouse
then i modified by adding
HWND hwnd = GetWindow (g_hWnd, GW_CHILD);
EnableWindow (hwnd, false);
after that i have disabled the madvr window, kodi can receive the keyboard messages and I can move and make a resize, but remain blocked mouse events
I made some video
in the first video madVR it's a child of kodi window with put_owner and I disabled the madVR window
http://youtu.be/RDDPydqN7QE
in the second madVR it's out from kodi, so no put_Owner, I analyzed the window of madVR with Winspector, when I pass the mouse over the window does not receive events, and the mouse cursor has the animation of the loading i don't know if it matters
http://youtu.be/oV2yDTpZw7g
my question is for the developers of player that maybe know about this behavior with madvr
put_MessageDrain seems to no produce any effect in any case
it's a wrong approach disable the madvr child window?
thanks
edit:
ok solved... it was a problem with handling messages in the thread where madvr was created (thx madshi)
Warner306
21st March 2015, 20:52
I am excited to try out the performance of madVR when the integration into Kodi DSPlayer is complete.
Mish12
22nd March 2015, 01:01
http://bugs.madshi.net/view.php?id=215&PHPSESSID=8dde4e94680d9e0e5bfc98594d441189
madshi told to write here. Could anybody help?
Driver version: 347.52. madvr version: 0.87.14. Video doesn't play if system color depth is 32 bit.
huhn
22nd March 2015, 01:36
what type of error do you get ? the bug reports talks about issue at playback but playback works in general.
Mish12
22nd March 2015, 14:41
madVR doesn't report about any error. Videoplayer hungs.
I have just found second workaround! Now I can use madVR with 32 bit system color depth. I run another videoplayer (for example, windows media player) with not-madVR renderer, pause video. Then I run videoplayer (MPC-BE) with madVR renderer and it works perfectly.
huhn
22nd March 2015, 19:04
so it's kinda like this issue?
http://forum.doom9.org/showpost.php?p=1713600&postcount=28466
do directx games work in window mode?
this simply sounds like a driver issue limited to windows XP.
Mish12
22nd March 2015, 19:59
so it's kinda like this issue?
http://forum.doom9.org/showpost.php?p=1713600&postcount=28466Yes.
do directx games work in window mode?I have tested New Supaplex (OS with DirectX 8.1 or higher in system requirements) and Luxor 5 (DirectX 9.0 in system requirements). It work in window mode.
THEAST
23rd March 2015, 14:13
Is anybody experiencing duplicate frames after "excessive jumping" using MPC-HC and madVR? Whenever I jump back and forth for a few times in a video using MPC-HC's shortcut keys and then stop, after a couple of seconds I get a bunch of duplicate frames on the screen from previous scenes that only lasts (much) less than a second and may repeat one or two times and then go away but sometimes the issue keeps repeating until I restart the player. At first I thought this might have something to do with my VGA (HD7950) or driver (Catalyst 14.12) but I'm also experiencing the same issue on my laptop with GTX 850M (renamed MPC-HC's executable to be able to force high performance mode in Nvidia's driver).
michkrol
23rd March 2015, 16:14
Is anybody experiencing duplicate frames after "excessive jumping" using MPC-HC and madVR? Whenever I jump back and forth for a few times in a video using MPC-HC's shortcut keys and then stop, after a couple of seconds I get a bunch of duplicate frames on the screen from previous scenes that only lasts (much) less than a second and may repeat one or two times and then go away but sometimes the issue keeps repeating until I restart the player. At first I thought this might have something to do with my VGA (HD7950) or driver (Catalyst 14.12) but I'm also experiencing the same issue on my laptop with GTX 850M (renamed MPC-HC's executable to be able to force high performance mode in Nvidia's driver).
I've never experienced anything like that. Just tested it to make sure.
How many seeks ("jumps") is a few? 5? 10? 25?
By "and then stop" do you mean the playback or seeking in the movie?
Does this not happen with other renderers available in MPC-HC?
Does this happen on all of your files?
Is your file ripped directly (from TV broadcast, etc.) or is it properly into mkv or other container that supports keyframes properly? Try muxing you file with mkvmerge GUI (free tool) to make sure it's not your file(s) being broken.
Do you have any of the "use separate device for ..." options enabled? If yes, try disabling them.
What does the debug OSD (CTRL+J) say after seeking?
What decoders and other filters are you using? What Windows version are you on?
leeperry
23rd March 2015, 17:22
Correct. If you want this fixed, could you please create a bug tracker report? Personally, I don't find this very important because the 16ms timer is idiotic, anyway. Maybe it made sense 10 years ago. But with today's PCs 16ms timer resolution simply makes no sense, whatsoever. Of course that's only my personal opinion. I do consider it "not nice" that madVR doesn't restore the original timer resolution, though.
This very issue would appear to annoy quite a lot of people actually: Windows Timer Resolution: Megawatts Wasted (https://randomascii.wordpress.com/2013/07/08/windows-timer-resolution-megawatts-wasted/).
I never shutdown my PC, I guess it doesn't really need 0.5ms while I'm sleeping and if that can lower my yearly energy bill by a few bucks then Amen to that.
Qaq
23rd March 2015, 17:37
Windows 8.1, madVR 0.87.11. Just checked with Clockres:
before playback - 15
during playback - 0,5
after playback (player closed) - 15
tp4tissue
23rd March 2015, 18:14
Hey guys..
Got questions.
When coming from Lav-decoder then into Madvr..
Which produces the best picture
Dxva2, Quicksync, or None (software)
And also, In Madvr, If I only have HD3000 (2520m), I can't use the Error diffusion (DX11) right? because the HD3000 only supports (DX10.1)
huhn
23rd March 2015, 18:31
Hey guys..
Got questions.
When coming from Lav-decoder then into Madvr..
Which produces the best picture
Dxva2, Quicksync, or None (software)
And also, In Madvr, If I only have HD3000 (2520m), I can't use the Error diffusion (DX11) right? because the HD3000 only supports (DX10.1)
the output of decoder should be bit perfect. so there is no difference.
if your GPU doesn't support directcompute than nope you can't use error diffusion. would be pretty borderline for a HD 3000 anyway.
jkauff
23rd March 2015, 20:46
Hey guys..
Got questions.
When coming from Lav-decoder then into Madvr..
Which produces the best picture
Dxva2, Quicksync, or None (software)
nevcariel has always said that software decoding is best if you have a decent CPU, because the CPU can process many more frames per second and there are no GPU driver bugs to worry about. As huhn said, the output visual quality is identical.
THEAST
23rd March 2015, 22:11
How many seeks ("jumps") is a few? 5? 10? 25?
By "and then stop" do you mean the playback or seeking in the movie?
Does this not happen with other renderers available in MPC-HC?
Does this happen on all of your files?
Is your file ripped directly (from TV broadcast, etc.) or is it properly into mkv or other container that supports keyframes properly? Try muxing you file with mkvmerge GUI (free tool) to make sure it's not your file(s) being broken.
Do you have any of the "use separate device for ..." options enabled? If yes, try disabling them.
What does the debug OSD (CTRL+J) say after seeking?
What decoders and other filters are you using? What Windows version are you on?
1. 5-10 consecutive jumps.
2. Stop seeking.
3. Haven't tried other renderers.
4. I think it only happens with high-resolution (720p+) H264 files.
5. Files are properly muxed with MKVmerge, quite a few of them by myself.
6. I had it enabled on my desktop, but not on my laptop.
7. Strangely enough, the one time I remember that the issue didn't go away no matter what, I checked the statistics and no dropped frames were being reported whatsoever.
8. LAV as splitter, ffdshow as both video and audio decoder. Windows 7 x64.
Sorry that I can't give much useful info, the issue isn't always reproducible.
aufkrawall
23rd March 2015, 22:11
Decent fixed function units (as seen on the GTX 960 or Broadwell) should be much faster than CPU for decoding, especially with HEVC.
baii
23rd March 2015, 22:59
nevcariel has always said that software decoding is best if you have a decent CPU, because the CPU can process many more frames per second and there are no GPU driver bugs to worry about. As huhn said, the output visual quality is identical.
Pretty much this. However there are some content out there now that aren't possible to decode fast enough using just software decoder on a average CPU(typical quad core). (Hevc 4k and those samples).
Hardware decode also typically give off less heat, hence less noise.
So for typical 1080p content, it is up to user liking IMO. I personally stick with software.
If the CPU is not fast enough for your content, then there's no other option but to enable some kind of hardware acceleration/decode.
Sent from my 306SH
tp4tissue
24th March 2015, 01:08
Gotcha guys.. Thx for the infos.
I need dxva on my pathetic 2520m laptop. The cooler is so crummy that using software lav + gpu catmull scale will push 80C, whereas dxva lav+madvr-dxva scaling only pushes 60C
That's 20C difference.. and it's only a 1366x768 output..
bouwew
24th March 2015, 08:31
Anyone else try the AMD 15.3 beta driver yet?
A quick check with 720p material upscaled to 1080p (on an R9 270, using NNEDI) shows a performance on par with the 13.12 driver :)
Will test more later...
huhn
24th March 2015, 09:45
i got this:
http://abload.de/img/haswellr9270l9kca.png
defiantly not a interop fix.
maybe be abit faster maybe not.
using a haswell CPU and a r9 270
QBhd
24th March 2015, 13:38
I tested the 15.3 Beta... and yet again on my rig, with settings pushed to the max, madVR performance dropped. All I do to recover the performance is roll-back the driver to 13.251 (I don't do a catalyst re-install)
QB
michkrol
24th March 2015, 17:50
3. Haven't tried other renderers.
Do try it. I think it might be a problem with the decoder, since no one else seems to be having this issue. If it's not really linked to madVR, we should move to a separate thread.
4. I think it only happens with high-resolution (720p+) H264 files.
Do you have hardware decoding enabled (DVXA/QuickSync/CUDA)? Try disabling it.
Also check GPU memory usage (with GPU-Z (freeware) or similar) and try lowering the queues' sizes.
8. LAV as splitter, ffdshow as both video and audio decoder. Windows 7 x64.
I would strongly advise you to switch to LAVFilters completely to avoid compatibility issues and just use a decoder that is actively developed.
jkauff
24th March 2015, 20:42
Decent fixed function units (as seen on the GTX 960 or Broadwell) should be much faster than CPU for decoding, especially with HEVC.
Agreed, but that doesn't help the OP until he does a future upgrade.
ryrynz
26th March 2015, 02:15
Madshi, any ideas what's causing madVR to crash when using latest MPC-BE along with XYsubFilter?
huhn
26th March 2015, 15:15
should be this: http://forum.doom9.org/showpost.php?p=1714727&postcount=2770
it's out of the mpc-be thread.
harshal_rio91
26th March 2015, 21:37
Any update on the black screen issue on Nvidia Optimus 800 series laptops?
Asmodian
26th March 2015, 22:05
No, it is still the same.
JonnyRedHed
27th March 2015, 02:43
Few quick questions please.
Just tried error diffusion - option 2 (latest madvr - Zoomplayer - Lav). I can see the small description next to it saying its different opt 1, but was just wondering is opt 2 a better option. I notice its slightly more expensive to run.
2nd. I tried option 2 because all of a sudden option 1 was showing thin vertical black lines on the screen. Took me ages to narrow down what was causing it. First I thought zoomplayer, so reset and loaded settings backup, and then moved onto madvr doing the same. I then started to go thru all the options one by one in madvr settings window and got to changing from error diffusion opt1 (my default) to opt 2 made the lines go away once I pressed apply. Back to opt1 and apply and the lines come back. No idea why, it just started today. Anyone have any thoughts on this?
screengrab: http://i57.tinypic.com/2a7tpbb.jpg
NV driver: 347.88 WHQL > HDMI > 50inch plasma TV
3rd question. The crash I mentioned above didn't manage to send via the web. I've only ever seen it send maybe twice in all the time I've been using madvr (which is years), and its tried many times. Zoom player Max v10 (my player) has access to the web, so do I need to allow some madvr exe access also via my PC firewall (Eset Smart Security 8). I hadn't had a crash for ages so just forgot to follow up why madvr crash reporter never managed to connect.
Thanks in advance guys.
dansrfe
27th March 2015, 08:52
No, it is still the same.
Do you know if it's a problem with the player, madVR, the intel/nvidia drivers or all the above? I wish I could understand the root of the black screen and freeze problem on optimus laptops with nvidia selected.
huhn
27th March 2015, 11:30
Few quick questions please.
Just tried error diffusion - option 2 (latest madvr - Zoomplayer - Lav). I can see the small description next to it saying its different opt 1, but was just wondering is opt 2 a better option. I notice its slightly more expensive to run.
2nd. I tried option 2 because all of a sudden option 1 was showing thin vertical black lines on the screen. Took me ages to narrow down what was causing it. First I thought zoomplayer, so reset and loaded settings backup, and then moved onto madvr doing the same. I then started to go thru all the options one by one in madvr settings window and got to changing from error diffusion opt1 (my default) to opt 2 made the lines go away once I pressed apply. Back to opt1 and apply and the lines come back. No idea why, it just started today. Anyone have any thoughts on this?
screengrab: http://i57.tinypic.com/2a7tpbb.jpg
NV driver: 347.88 WHQL > HDMI > 50inch plasma TV
3rd question. The crash I mentioned above didn't manage to send via the web. I've only ever seen it send maybe twice in all the time I've been using madvr (which is years), and its tried many times. Zoom player Max v10 (my player) has access to the web, so do I need to allow some madvr exe access also via my PC firewall (Eset Smart Security 8). I hadn't had a crash for ages so just forgot to follow up why madvr crash reporter never managed to connect.
Thanks in advance guys.
know issue. since driver version 347.25
http://bugs.madshi.net/view.php?id=250
i don't think mode 2 is better it's just different. just choice what looks better to you, if you don't see a difference don't use it.
huhn
27th March 2015, 11:35
Do you know if it's a problem with the player, madVR, the intel/nvidia drivers or all the above? I wish I could understand the root of the black screen and freeze problem on optimus laptops with nvidia selected.
it's pretty clear this is an nvidia/intel driver issue.
clsid
27th March 2015, 13:55
Didn't MPDN have similar problems with Optimus that were later fixed? I do not follow that topic closely, but I vaguely remember a black screen issue with too large queue size. I am not sure if that was Optimus specific. But you could test with all queues at minimum size in madVR.
huhn
27th March 2015, 14:21
that's what i know:
http://forum.doom9.org/showthread.php?t=170937&highlight=optimus&page=3
maybe MPDN has a better chance in fixing this it is a hole new problem while mpc-hc and player like that are normal playback software that is usually set to low powerstate/iGPU to spare power.
people can try min queue size but some feature may totally fail with that.
JonnyRedHed
27th March 2015, 14:30
know issue. since driver version 347.25
http://bugs.madshi.net/view.php?id=250
i don't think mode 2 is better it's just different. just choice what looks better to you, if you don't see a difference don't use it.
Ah I see a known bug, thank you for pointing that out. Yes I was on a driver some 4 months old before updating yesterday. I even updated again today to 347.90 in the hope it would fix it, but alas it didn't.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.