View Full Version : madVR - high quality video renderer (GPU assisted)
dansrfe
23rd March 2013, 03:53
My last comment on this topic, promise. :)
"Most"? I don't think so.
Some that claim to only support 60 Hz can be made to accept other refresh rates using special software but certainly not all. Even my high-end 24WMGX3 only supports 50 Hz via HDMI and even then it clearly isn't natively displayed because there is tearing.
With custom resolutions or Custom Resolution Utility, I've been able to use 48Hz and 50Hz on all my displays excluding my Panasonic plasma which accepts 24hz and, according to the settings, repeats each frame 3 times to achieve 96Hz on its own.
turbojet
23rd March 2013, 06:42
Yes, for wasapi output. But also because there is a small difference between 24hz and 23.9 meaning in order not to see a frame added every so often you want to alter the audio clock instead(essentially what reclock does)
I don't use wasapi because windows 7 normalization doesn't work with it. The other reason you mention changed my mind on uninstalling. Another thing is slowing 25 to 24 to reduce refresh rate changes. I tried but can't get it working right with madvr, has anyone else?
madvr 25 as 24 checked, reclock slowdown checked = 24 at 50hz, bad
madvr checked, reclock unchecked = 25 at 50hz, good but expect 24 at 48
madvr unchecked, reclock checked = 24 at 50hz, bad
both checked = 24 at 50hz, bad
Perhaps it's a madvr bug that can be worked around by using reclock runevent.vbs and disabling madvr switching.
dansrfe
23rd March 2013, 06:59
I don't use wasapi because windows 7 normalization doesn't work with it. The other reason you mention changed my mind on uninstalling. Another thing is slowing 25 to 24 to reduce refresh rate changes. I tried but can't get it working right with madvr, has anyone else?
madvr 25 as 24 checked, reclock slowdown checked = 24 at 50hz, bad
madvr checked, reclock unchecked = 25 at 50hz, good but expect 24 at 48
madvr unchecked, reclock checked = 24 at 50hz, bad
both checked = 24 at 50hz, bad
Perhaps it's a madvr bug that can be worked around by using reclock runevent.vbs and disabling madvr switching.
Well, the more important point of PAL slowdown is to watch the video at the speed it was filmed at. madVR's job is to make the output as smooth as possible to the best of its ability.
Users who usually check the "treat 25 as 24" are those that have a 24Hz multiple refresh rate in the list of valid refresh rates that madVR can choose from. That setting is moot if you don't have a 24Hz refresh rate and causes more harm than good in such a situation.
nevcairiel
23rd March 2013, 07:01
With custom resolutions or Custom Resolution Utility, I've been able to use 48Hz and 50Hz on all my displays excluding my Panasonic plasma which accepts 24hz and, according to the settings, repeats each frame 3 times to achieve 96Hz on its own.
Just because you can send that to your TV doesn't mean it runs the panel natively at that rate. I would be very careful with the non-standard rates, and do thorough checks to ensure it doesnt internally mess everything up, like most TVs do.
turbojet
23rd March 2013, 07:26
Well, the more important point of PAL slowdown is to watch the video at the speed it was filmed at. madVR's job is to make the output as smooth as possible to the best of its ability.
Users who usually check the "treat 25 as 24" are those that have a 24Hz multiple refresh rate in the list of valid refresh rates that madVR can choose from. That setting is moot if you don't have a 24Hz refresh rate and causes more harm than good in such a situation.
There's 48hz in the switch line but it's switching to 50hz with 25fps source even when reclock is slowing down. Removing 50hz from the line makes it change to native 60hz with 25fps.
Just because you can send that to your TV doesn't mean it runs the panel natively at that rate. I would be very careful with the non-standard rates, and do thorough checks to ensure it doesnt internally mess everything up, like most TVs do.
Are smoothmotion testpatterns sufficient? This displays specs are 56-82hz but 48 and 50hz show no judder in testpatterns.
DragonQ
23rd March 2013, 11:35
My last comment on this topic, promise. :)
With custom resolutions or Custom Resolution Utility, I've been able to use 48Hz and 50Hz on all my displays excluding my Panasonic plasma which accepts 24hz and, according to the settings, repeats each frame 3 times to achieve 96Hz on its own.
If by "displays" you mean TVs then sure. PC monitors do not, by and large, support anything other than 60 Hz natively (apart from the newer 120 Hz ones of course).
If it's the case most displays support the change, why is it that they offer technology that is suppose to eliminate judder when fed a 24Hz signal when the same displays supposedly natively display 24 Hz and why now do we have 120 Hz+ tv's??
That is different. Even when watching 24p at a native refresh rate (e.g. 96 Hz), there will be "judder" during panning shots just because the frame rate is low (which gives the "film effect"). Interpolating new frames between the existing ones smooths things out - an effect that some like and some don't. This is not related to the judder you get when frames are not displayed for an equal amount of time (e.g. watching 24p at 60 Hz).
I'm welcome to be proved wrong, but I'm quite sure the ability to accept 24Hz does not equal the actual television refreshing at that rate, but instead it being "converted" to the native refresh rate.
This was true in the early days of flat panels, it is not now. Any review of an HDTV will tell you whether it natively displays 24p/23.976p content properly. And again, pretty much all TVs sold in Europe can do 50 Hz and 60 Hz natively, even the old CRTs that were made long before HD or 24p BDs.
leeperry
23rd March 2013, 12:43
Even when watching 24p at a native refresh rate (e.g. 96 Hz), there will be "judder" during panning shots just because the frame rate is low (which gives the "film effect").
But that would also depend on what actually happens between the actual source frames, madshi discussed this matter at http://www.avsforum.com/t/802850/
I just got ahold of a high-end flat screen CRT and 24p@96Hz looks so amazingly smooth! Yes, it's darn small but black is as deep as on a TOTL Plasma(0.03 cd/mē) and movies look as smooth as they will ever get IMO :devil:
BTW, I recently tried a bunch of LCD TV's, they would all support 48Hz but that would hiccup very badly.....hell, even 24Hz hiccuped on one of them, hah!
nevcairiel
23rd March 2013, 12:55
they would all support 48Hz but that would hiccup very badly
Which would mean they don't actually support 48Hz, and just fake it.
Dodgexander
23rd March 2013, 15:27
Well I own a Samsung M86 series (LN61?? in North America).
It was a tv advertised with no "official" 24p support, but it does accept a 24hz signal when forced, its just not in its EDID.
So this leads me to believe that 24hz mode on my tv and probably many others, either isn't implemented well, or converts the refresh rate itself and I can't believe they would invent technology that equals a multiple of this refresh rate, if displays already displayed 24hz ok.
How exactly do you know that each display is infact showing 24hz? I can't think 120mhz+ tvs were designed just for frame interpolation.
DragonQ
23rd March 2013, 17:30
Well I own a Samsung M86 series (LN61?? in North America).
It was a tv advertised with no "official" 24p support, but it does accept a 24hz signal when forced, its just not in its EDID.
So this leads me to believe that 24hz mode on my tv and probably many others, either isn't implemented well, or converts the refresh rate itself and I can't believe they would invent technology that equals a multiple of this refresh rate, if displays already displayed 24hz ok.
Ah so because your TV doesn't support it properly, others obviously don't either. :rolleyes:
How exactly do you know that each display is infact showing 24hz?
It is incredibly obvious when the refresh rate doesn't match the video frame rate if you know what you're looking for. I am watching 23.976p on an HDTV right now that has no interpolation features at all. I would know if it was displaying it at 50 or 60 Hz. Seriously, you can Google this stuff. It's like saying "how exactly do you know that this colour TV is colour?" There were lists made a few years back showing which TVs did and didn't support true 24p playback.
Not really related but also note that TVs sold in the USA (possibly all of North America) are intentionally crippled to lack 50 Hz support.
vivan
23rd March 2013, 20:56
There's something wrong with new madTestPatternSource.
With new version image is broken with any renderer (except madVR), pic (http://firepic.org/images/2013-03/23/abkbo7314h1q.png). With old one everything is ok.
leeperry
23rd March 2013, 22:35
Which would mean they don't actually support 48Hz, and just fake it.
Well, you don't get the laconic "out of range" error message but still, it's not 1:1 so that's useless for video use.
Anyway, I see that there are patents explaining how to mimick the way a CRT works: http://www.google.com/patents/US7324933
I'm sure madshi already read those pages anyway, and I guess that me complaining about how his frame blending looks like a ghost story on a CRT is a gross misjudgment as CRT's work very differently from the rest anyway. I've ordered yet another LCD TV(not an edge-led model this time, though :D), so I'll try my luck again ^^
And going back to Sammy TV's, I find the DNIE effect really obvious on the BBC screencap used in this review: http://www.hdtvtest.co.uk/news/samsung-ue55f8000-201303212755.htm?page=Performance
I'm just a little confused when they write:
As with previous Samsung LED LCDs, the Samsung UE55F8000 can show all 1080 lines in a scrolling resolution test chart, provided that the [Motion Plus] system is engaged. With this feature off, you’ll see the usual LCD motion performance, which delivers just 300 lines worth of clear details. That’s enough for 24fps movies to appear without much in the way of discernable blur, but fast, high motion content like televised sports will still appear with some.
Thankfully, Samsung’s [Motion Plus] system has individually controllable deblur and dejudder parameters. That means that we can benefit from the increased motion resolution, but still watch movies with their original, characteristic, filmic motion, without the TV generating new interpolated inbetween frames and converting them to high-motion (the so-called “soap opera effect”).
300 lines worth of clear details :confused:
scrolling resolution test chart :confused:
DragonQ
24th March 2013, 00:28
Perhaps they are talking about the fact that fast motion in video results in a lot of detail loss. I remember reading a study that showed very high refresh rates (200+ Hz) hugely increased perceived detail during motion, even more so than increasing spatial resolution (e.g. 4K vs 2K). UHD is meant to support 100/120 Hz progressively scanned video so hopefully that is what future broadcasts will use.
Dodgexander
24th March 2013, 03:58
Ah so because your TV doesn't support it properly, others obviously don't either. :rolleyes:
It is incredibly obvious when the refresh rate doesn't match the video frame rate if you know what you're looking for. I am watching 23.976p on an HDTV right now that has no interpolation features at all. I would know if it was displaying it at 50 or 60 Hz. Seriously, you can Google this stuff. It's like saying "how exactly do you know that this colour TV is colour?" There were lists made a few years back showing which TVs did and didn't support true 24p playback.
Not really related but also note that TVs sold in the USA (possibly all of North America) are intentionally crippled to lack 50 Hz support.
Well even though my display shows on info 24p, there is judder every so often. I suspect because it wasn't implicated yet, so it doesn't know how to deal with 23.976>24Hz correctly.
You mention the difference to be "obvious" but there is still no definitive way to test, unlike using a test pattern for colour!
Sent from my Blade S using Tapatalk 2
DragonQ
24th March 2013, 11:42
You mention the difference to be "obvious" but there is still no definitive way to test, unlike using a test pattern for colour!
Do you need a test pattern to see whether a TV is colour? I'm done with this conversation, if you really can't use Google to look this up then I can't help you.
secvensor
24th March 2013, 11:55
Can it be used directly in MPC/POTp?
I do not know, I not the expert in it.:)
Owyn
24th March 2013, 12:19
madshi, could you also write date of release near version "madVR v0.86.1" here ? ^.^
Dodgexander
24th March 2013, 13:10
Do you need a test pattern to see whether a TV is colour? I'm done with this conversation, if you really can't use Google to look this up then I can't help you.
Not to see if it displays colour, to see if it displays colour correctly.
Comparing this to whether a tv is colour or not is like comparing whether a TV refreshes or not, not whether it refreshes correctly.
Sorry to antagonise you, not really my intention.
pirlouy
24th March 2013, 16:44
madshi, could you also write date of release near version "madVR v0.86.1" here ? ^.^
+ Version in title, it would save useless clicks.
dansrfe
24th March 2013, 18:22
Version of release needs to be in the OSD IMHO.
Dodgexander
24th March 2013, 18:33
Is it normal to have dropped frames using madVR with lowest quality settings and a 1080i 25fps input?
The card in question is a Radeon 5450.
ryrynz
24th March 2013, 23:02
Version of release needs to be in the OSD IMHO.
Madshi has always stated he wanted the OSD to be as spartan as possible, this info is easy to find as it is, he would have it there already if he thought it useful I think.
Qaq
25th March 2013, 07:15
Is it normal to have dropped frames using madVR with lowest quality settings and a 1080i 25fps input?
The card in question is a Radeon 5450.
It is. I tried de-interlacing with 5450 and guess what? I'm 7750 GDDR5 fanless user now.
Dodgexander
25th March 2013, 17:29
Hmm the other card I can try is a GeForce 610. Also very low end but maybe it will work better.
Sent from my Blade S using Tapatalk 2
Dodgexander
25th March 2013, 23:10
It is. I tried de-interlacing with 5450 and guess what? I'm 7750 GDDR5 fanless user now.
Did a bit of research and tweaking, with the cyberlink powerdvd decoder in HAM mode, I can watch broadcasts in 1080i 25fps with no dropped frames. This is using softcubic 100/dxva scaling/smooth motion off.
I had to tweak the backbuffer queue to only 1 for full screen exclusive mode to work without dropped frames though. Disabling the backbuffer queue altogether caused presentation glitches and setting it to 2 was too much.
Everything else was default settings, which I am pleased about.
In windowed mode, I can't seem to tweak anything to work, but that doesn't really matter as its just a HTPC.
Qaq
26th March 2013, 06:38
Did a bit of research and tweaking, with the cyberlink powerdvd decoder in HAM mode, I can watch broadcasts in 1080i 25fps with no dropped frames.
Yes, you can watch it, but you can't get hardware de-interlacing (vector/motion adaptive) with Cyberlink HAM. Try those 1080i25 with Cyberlink DXVA and plain EVR (not EVR CP) to check out the difference. But first make sure you have Vector Adaptive and Auto in CCC/video. You can try EVR CP though with half floating or full floating processing, but as far as I remember it was the same story as with madVR.
Dodgexander
26th March 2013, 11:48
Deinterlacing seems to work fine, just isn't reported correctly to madVR osd.
I don't know what kind of deinterlacing it is, but it works.
Sent from my Blade S using Tapatalk 2
renethx
27th March 2013, 08:05
I don't know what kind of deinterlacing it is, but it works.
With HAM,
"Auto-select" = "Force bob" = blending (not even bob)
"Force weave" = weave
HAM decoder sends a 25fps (not 50fps of course) progressive (deinterlaced this way) video stream to madVR, whence "deinterlacing off (says upstream)" in madVR OSD. So it works, but is not good in quality. This is true whatever renderer you use, VMR9, EVR, EVR CP, Haali, madVR etc.
If madVR receives an interlaced video stream, its OSD shows "deinterlacing on (says upstream)" and it leaves video to the driver's deinterlacing algorithm (perhaps vector adaptive; of course there are lots of dropped frames if GPU is not good enough to do both vector adaptive deinterlacing and madVR's scaling algorithms).
DragonQ
27th March 2013, 10:55
I would personally much rather high quality deinterlacing + EVR than pants deinterlacing + MadVR. Of course high quality deinterlacing + MadVR is the best. ;)
Dodgexander
27th March 2013, 12:50
I see!! What kind of software deinterlacing can I use with madVR then? I guess I'll have to fiddle around with more decoders. I actually prefer yadif to the hardware deinterlacing on my other pc, but this one's cpu can't cut it.
It's at times like this that I wish I forked out for a better card;)
DragonQ
28th March 2013, 19:08
Does anyone know if there's any media centre software that can use MadVR for everything including TV?
I'm worried MediaPortal's EVR might not be good enough on a good TV, mainly due to banding. I've been doing some testing on my PC (with a monitor capable of passing banding tests very well) and even if there is no expansion from 16-235 to 0-255, there's still a bit of banding on demanding gradients (mainly sky blue). Nowhere near as bad as when expansion to 0-255 is used but still worse than I get with MadVR (presumably due to high-bit-depth internal processing).
SamKook
28th March 2013, 19:57
Does anyone know if there's any media centre software that can use MadVR for everything including TV?
I'm worried MediaPortal's EVR might not be good enough on a good TV, mainly due to banding. I've been doing some testing on my PC (with a monitor capable of passing banding tests very well) and even if there is no expansion from 16-235 to 0-255, there's still a bit of banding on demanding gradients (mainly sky blue). Nowhere near as bad as when expansion to 0-255 is used but still worse than I get with MadVR (presumably due to high-bit-depth internal processing).
I don't know if there are any that support it natively, but what I personally do is use XBMC which can launch any external player (http://wiki.xbmc.org/index.php?title=External_players) which supports madvr. I guess TV would be more complicated like that though.
DragonQ
28th March 2013, 20:11
I don't know if there are any that support it natively, but what I personally do is use XBMC which can launch any external player (http://wiki.xbmc.org/index.php?title=External_players) which supports madvr. I guess TV would be more complicated like that though.
Yeah MediaPortal can do this too but, like XBMC, it doesn't work for TV. I think the reasoning is that the whole TV/EPG/recordings/OSD interface is built in to MediaPortal and can't be generated using MadVR (right now at least).
huhn
28th March 2013, 20:14
jriver should do the trick. but is not free
Dodgexander
28th March 2013, 23:17
For live TV you can use MPC-HC, of course without PVR capability.
Next PVR works too, but it has a bug with program info always showing at the bottom, specifically with madVR , also sometimes it doesn't go full screen.
leeperry
29th March 2013, 15:15
@madshi: I was wondering, what does Jinc3AR need the most in your opinion? ROP/shader performance or memory bandwidth? It's all discussed in this review (http://www.anandtech.com/show/6838/nvidia-geforce-gtx-650-ti-boost-review-/7), and these days regular 650Ti's are pretty much given away(95€ shipped). Surely the 2GB 650Ti Boost(160€) kills the regular 1GB 650Ti but they didn't dare releasing its 1GB counterpart yet....so it's basically a slightly slower 660(170€) and resellers have to ditch their regular 650Ti's ASAP before they get obsolete.
In shader-heavy games, the vanilla 650Ti is a screaming bargain (http://www.anandtech.com/show/6838/nvidia-geforce-gtx-650-ti-boost-review-/8) I think....and I believe Jinc3AR is all about shaders :confused:
If all Jinc3AR cares for is memory bandwidth then it'll be a waste of money, though....tough call :o
Niyawa
29th March 2013, 19:27
I'm asking this here because I made a search but I never got any complete answer from it. What exactly the option "don't use linear light for smooth motion frame blending" do? What I know is that when I enable it, I get a lot of dropped frames. Does it increase quality of smooth motion or something along those lines?
aufkrawall
29th March 2013, 19:40
I think....and I believe Jinc3AR is all about shaders :confused:
I think so too, afair 1080p24@720p (Jinc3 AR chroma, C-R AR LL luma) even worked with my Intel HD 3000 (OCed to 2.2 Ghz).
Haswell IGP with OC could be nice.
huhn
29th March 2013, 22:13
I'm asking this here because I made a search but I never got any complete answer from it. What exactly the option "don't use linear light for smooth motion frame blending" do? What I know is that when I enable it, I get a lot of dropped frames. Does it increase quality of smooth motion or something along those lines?
doing it in rbg is simple the correct way to do it that's all so quality is a little bit better
Niyawa
30th March 2013, 01:07
doing it in rbg is simple the correct way to do it that's all so quality is a little bit better
I'm sorry... could you elaborate on that?
huhn
30th March 2013, 01:33
from madshi
One last hint: There's a new "trade quality for performance" option in the madVR settings which affects the quality of the FRC frame blending. By default frame blending is done in gamma corrected light, which is fast, but not mathematically correct. If your GPU is fast enough, it's highly recommended that you disable the "trade quality" option for highest image quality. There are no negative side effects to blending frames in linear light - except for higher GPU usage, of course.
before i even try to answer that
Niyawa
30th March 2013, 04:30
from madshi before i even try to answer that
That's exactly what I was asking but in a more in-depth manner. madshi says: "it's highly recommended that you disable the "trade quality" option for highest image quality" but that highest image quality alone didn't answer me. Does it increase sharpness? Or maybe it makes video playback even less jittery? Maybe it makes motion more fluid? I really didn't notice any difference so I want to know what I can expect from it.
Dodgexander
30th March 2013, 05:09
That's exactly what I was asking but in a more in-depth manner. madshi says: "it's highly recommended that you disable the "trade quality" option for highest image quality" but that highest image quality alone didn't answer me. Does it increase sharpness? Or maybe it makes video playback even less jittery? Maybe it makes motion more fluid? I really didn't notice any difference so I want to know what I can expect from it.
I think it's just the quality of the picture. When the scaling is done in linear instead of gamma light it produces better picture quality than in gamma. (I'm guessing)
Sent from my Blade S using Tapatalk 2
Dodgexander
30th March 2013, 05:22
I see!! What kind of software deinterlacing can I use with madVR then? I guess I'll have to fiddle around with more decoders. I actually prefer yadif to the hardware deinterlacing on my other pc, but this one's cpu can't cut it.
It's at times like this that I wish I forked out for a better card;)
Thought I'd update to let anyone know. I have been able to get my Radeon 5450 working full screen with hardware deinterlacing, this is with a 1080i 25/576i source. even with softcubic 100/lanzcos3. Obviously no smooth motion, but I got this working by enabling the top option in the trade quality for performance section.
This is while using lav software decoder.
Now it can handle anything I throw at it, given that the hardest load is upscaling and deinterlacing 576i to 1080p. With 1080i sources I can even enable jinc3 for image upscaling.
Windowed mode however won't stop dropping frames no matter what I try.
So a suggestion perhaps would be to have different scaling options depending on source resolution?
Also wanted to thank Madshi for including the performance options saving me a good bit of money in the process.
Sent from my Blade S using Tapatalk 2
huhn
30th March 2013, 05:26
this is not about scaling is about blending. scaling in linear light is not recommend at least not for upscaling.
Coldblackice
30th March 2013, 08:35
Are there any benefits to using DXVA vs no hardware decoder at all (provided you have a CPU that can handle the processing on its own)?
I'm running an i7 930 @ 4.2Ghz, which handles video pretty well. But I also have a 7950 3GB @ 1150Mhz/1700Mhz. When using DXVA, there's a bit of a lag/loading delay when skipping around video, which isn't present if hardware decoding isn't used. I'm just wondering if there are any benefits to using DXVA (or hardware decoding in general), provided your CPU can handle the task on its own.
nevcairiel
30th March 2013, 08:52
I'm just wondering if there are any benefits to using DXVA (or hardware decoding in general), provided your CPU can handle the task on its own.
Hardware decoding is sometimes more power efficient, but all in all, if your CPU can handle it, then stick to CPU decoding, its more flexible and less error-prone.
Qaq
30th March 2013, 09:13
With DXVA I can get more smoother playback. Funny thing that it seems a bit tricky: for example, I watched a movie and had to enable forced subs and after that I've got the smoothest motion I've ever experienced on my TV. This never happened with SW decoding. Dunno if its somehow related to subs or pause-unpause or something else... I wish I know.
iaTa
30th March 2013, 12:12
Using any Intel drivers from the recent 15.31.x releases (15.31.1.3006 to 15.31.2.3055) I have an issue with the info panels (volume, play, stop etc) being left onscreen in the black bar areas.
Disabling exclusive mode fixes the issue as does rolling back to any 15.28.x driver.
I'm running a Clevo P150EM laptop with Intel HD 4000 and Nvidia 680M.
If I can do anything to assist with troubleshooting let me know.
Weirdo
30th March 2013, 16:17
Many, many thanks for making my old 50-60Hz projector play everything smoothly, without having to mess with the audio! Really great work. Do you happen to know when it'll be integrated into Media Center?
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.