View Full Version : madVR - high quality video renderer (GPU assisted)
VBB
11th October 2019, 18:14
Out of curiosity, how does these two pictures look on an LG OLED in PC Mode when displaying it at 100% size (1:1 Pixel)?
http://www.ozone3d.net/public/jegx/201412/chroma-subsampling-test-4k-tv.png
http://madshi.net/madVR/ChromaRes.png
Some Makro photos of the screen using a camera would be interesting.
Only LG OLEDs are affected or also LCD based TVs from LG? Is it year depending?
How about Panasonic OLED or LCD TVs?
Klaus, those two chroma test patterns look just fine in PC mode, since PC mode does enable 4:4:4. The issue is that PC mode also switches the TV into 8-bit mode (this hasn't been proven 100%, but it is highly likely), which in and by itself would be OK, if it wasn't for the terrible banding it results in. You can see this in both SDR and HDR, although it's more prominent in HDR.
No clue if it also affects LG's LCDs or any of the other brands that use their panels.
thighhighs
11th October 2019, 18:37
Read the first part of this post (https://forum.doom9.org/showthread.php?p=1871836#post1871836) and read the topic from here (https://forum.doom9.org/showthread.php?p=1882835#post1882835) couple of pages.Ok, 444@23... may be very useful (https://www.rtings.com/tv/learn/chroma-subsampling) :) But it’s very bad that it’s not "free". As well as any 4:4:4 modes on this TV, if i understand properly.
Klaus1189
11th October 2019, 18:52
Klaus, those two chroma test patterns look just fine in PC mode, since PC mode does enable 4:4:4. The issue is that PC mode also switches the TV into 8-bit mode (this hasn't been proven 100%, but it is highly likely), which in and by itself would be OK, if it wasn't for the terrible banding it results in. You can see this in both SDR and HDR, although it's more prominent in HDR.
No clue if it also affects LG's LCDs or any of the other brands that use their panels.
Thanks for the info. ;)
I thought of an 4:4:4 issue (4:4:4 and RGB ... :rolleyes:, RGB is always 4:4:4 eqivalent ;)) because of that info:
@jk82 is right, LG does some crazy things, e.g. there's no proper chroma 4:4:4 in PC mode at 4k resolution.
Chros has a B8, you have a C7, so probably it differs from series and year?
Only LG OLEDs are affected or also LCD based TVs from LG?
If someones knows that, please let me know :)
But it’s very bad that it’s not "free". As well as any 4:4:4 modes on this TV, if i understand properly.
What is not free?
thighhighs
11th October 2019, 19:17
Klaus1189
PC mode allows 4:4:4 chroma, but has some problems... See chros notes (https://forum.doom9.org/showthread.php?p=1871836#post1871836) about PC/non-PC modes from the link. That i mean not "free".
VBB
11th October 2019, 23:01
Thanks for the info. ;)
I thought of an 4:4:4 issue (4:4:4 and RGB ... :rolleyes:, RGB is always 4:4:4 eqivalent ;)) because of that info:
Chros has a B8, you have a C7, so probably it differs from series and year? If someones knows that, please let me know :)
There is no difference between models and years, as far as we can tell. What @chros was referring to was a statement made by Ted over at AVS, claiming that the LGs don't do a perfect 4:4:4, but very close. Nevertheless, chroma patterns are displayed without (noticeable) sub-sampling.
mclingo
11th October 2019, 23:09
Hi, couple of issues with AMD 19.9.2 ive found, I've moved back to 19.8.1, firstly the vocal volumes were off for me, much quieter then they should be, thought it was my setup but issue was resolved by moving back, to double check i installed again, same issue again, so went back again.
Second issue is corruption when skipping, this is something AMD cards have had on and off for years, it sorts itself out after a few seconds of playback usually so no big deal.
Not a MADVR issue I know but thought i should say.
ryrynz
12th October 2019, 01:21
I find that de-banding on low/low doesn't hurt picture quality at all, and is unfortunately necessary with most content, regardless of the display. It works really well, though.The low settings were designed to avoid any noticeable image degradation and it works quite well on content.. I doubt that it would affect the panels colors enough to be considered an improvement TBH.
garson
12th October 2019, 03:35
So suggested setting for de-banding is low/low? I had it on medium/medium so far, 1080p Samsung LED TV.
VBB
12th October 2019, 06:54
So suggested setting for de-banding is low/low? I had it on medium/medium so far, 1080p Samsung LED TV.
Depends on how heavy the banding in your content is. Low/low doesn't cause any noticeable harm and gets rid of most banding. I leave it on low/low for everything.
el Filou
12th October 2019, 14:38
As always with compression artifacts, it depends on the source.
I have some MPEG2 SD TV channels that are so bitrate-starved that they need deband on high. The good thing is I find when you have horrible source, that deband on high also has the side-effect of smoothing out other compression artifacts without needing to enable other algos. Image looses some detail but I definitely prefer it to the bands-and-blocks fest that is the alternative.
garson
12th October 2019, 17:05
Depends on how heavy the banding in your content is. Low/low doesn't cause any noticeable harm and gets rid of most banding. I leave it on low/low for everything.
I mostly watch 1080p movies, ~10GB in size.
Ok, so low/low should be fine.
:thanks:
Bo.
16th October 2019, 09:14
Hi everyone.
When downscaling UHD->FHD (projector FHD) on test patterns S&M (Chroma Multiburst and Chroma Zone Plate), the Lanczos (tape 4) chroma upscaling looks much better than everyone else (with the exception of NGU AA, I just don't have enough power for it). super-xbr: the test circle looks awful because of the moire throughout the area. Jinc - too very dark edges in the corners. Spline looks a bit darker around the edges than Lanczos. Bilateral is generally terrible. Lanczos visually very similar to NGU AA on all test patterns. I use the SSIM 1D + LL + AR + AB downscaling algorithm.
How to choose the right algorithm? According to test patterns, as I did, I chose the closest to NGU (Lanczos for me), which looks great. Or put the most powerful (super-xbr) and not look at the test patterns?
huhn
16th October 2019, 09:59
test pattern can be like pixel art a place where jinc does bad.
i personally ignore test pattern fro chroma and select the chroma scale on real world image where it matters.
BTW you could avoid the chroma scaler in your case and use the option under trade for performance "scale chroma separately, if...".
NGU AA and super xBR AR are considered to be better choices for chroma.
Bo.
16th October 2019, 10:12
Thanks! If I cancel chroma scaling, then there is a clear difference on the real content. I can’t say for better or worse, it's just different. Without chroma scaling bright white-red areas look more blurry. With scaling more compiled. What would you advise to choose:
1) chroma: super-xbr + AR; luma: SSIM 1D + LL + AR + AB or 2) without chroma scaling; luma: SSIM 2D + LL + AR + AB
huhn
16th October 2019, 10:45
i'm not down scaling so i have no real experience here not touching the chroma should be relative save and cheaper on the GPU.
okc_smoker
16th October 2019, 14:00
Are there any examples out there of what the chroma layer actually looks like without the luma layer (and vice-versa)? Even better if there are examples of a chroma/luma layer upscaled with different algorithms.
Sent from my SM-G965U using Tapatalk
RXP
16th October 2019, 15:40
Worst comes to worst ... can tell MadVR to output in Top/Bottom format ... then can tell TV to output 3D in Top/Bottom
if you use 4K LG OLED ... this should not result in any resolution loss
downside ... cannot watch 3d rips in 2d (as switching tv to 2d will show you top/bottom picture) ... solution - create symbolic link to your 3d rip ... name one mkv another mk3d ... then use madvr profiles to output one in top/bottom and another in 2d
plus side to this ... you using madvr to do chroma upscaling and horizontal stetching in 4k for 3d ... and for 2d you also using madvr for upscaling to 4k
Just wanted to say thanks for this tip. I just setup my 1660TI today and it works perfect :)
Cowboydude99
17th October 2019, 01:27
Hey there
Can someone help me figure this out!
Why is my video concerting to a different resolution?
I don’t see any setting with that resolution...
The movie won’t really play. A few frames a minute.
I posted this same question earlier, but in the wrong place. Posting here for assistance.
https://forum.doom9.org/attachment.php?attachmentid=17009&d=1571271534
mclingo
17th October 2019, 11:05
I doubt anyone here has the crystal ball required to help you with this, you've given us nothing to go on really, post your MADVR control+J screen and your system and gfx card specs + your video file specs.
Cowboydude99
17th October 2019, 17:34
Yes, I supposed having information would help! :)
I attached screenshots of what options are configured.
If more images are needed, please let me know.
Media File Info:ID : 1
Format : HEVC
Format/Info : High Efficiency Video Coding
Commercial name : HDR10
Format profile : Main 10@L5.1@High
Codec ID : V_MPEGH/ISO/HEVC
Duration : 2 h 32 min
Bit rate : 47.2 Mb/s
Width : 3 840 pixels
Height : 2 160 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 23.976 (24000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0 (Type 2)
Bit depth : 10 bits
Bits/(Pixel*Frame) : 0.237
Stream size : 50.1 GiB (93%)
Default : Yes
Forced : No
Color range : Limited
Color primaries : BT.2020
Transfer characteristics : PQ
Matrix coefficients : BT.2020 non-constant
Mastering display color primaries : BT.709
Mastering display luminance : min: 0.0050 cd/m2, max: 4000 cd/m2
Maximum Content Light Level : 500 cd/m2
Maximum Frame-Average Light Level : 200 cd/m2
Klaus1189
17th October 2019, 17:56
We cannot watch your attachments, see here:
https://drive.google.com/file/d/1ETmRdWJCDxp2GBEyxoo5yncjF-Tq_eMR/view?usp=sharing
Please upload your files to Google drive or whatever.
Cowboydude99
17th October 2019, 18:03
Sorry, I'm kind of new here!
https://drive.google.com/drive/folders/1yBTzBQgznJITOSNBxQLUpymz6FxWA1wG?usp=sharing
Asmodian
17th October 2019, 19:26
Your profile rules are probably bad
Use this instead:
if (srcHeight <= 720) "720p"
else if (srcHeight <= 1080) "1080p"
else if (fps > 31) "4K60"
else "4K"
Edit:
I posted this same question earlier, but in the wrong place. Posting here for assistance.
That was a perfectly fine place too. :)
Can you post a screenshot/photo of the madVR OSD (crtl-J)? What GPU do you have?
Cowboydude99
17th October 2019, 20:08
Thanks! I added the image. GPU is GTX1080
I can update the rules, but why would the resolution be changed to what is in the newly added image?
Asmodian
17th October 2019, 20:18
Sorry, your rules weren't this issue.. but they probably weren't doing what you wanted them to either. :)
It seems your issue is with the player, it is asking madVR to 'touch window from inside, +1%'. I have no idea why it would do that. I suggest using Zoom Player. ;)
This is not a madVR issue or option, it is simply doing what the player asked of it.
Cowboydude99
17th October 2019, 20:26
Thanks for the rule. I realized I had Width 720, instead of height, so that's good!
This is the only video the player has an issue with. No other video shows that touch from inside +1%...
Klaus1189
17th October 2019, 20:28
Thanks for the rule. I realized I had Width 720, instead of height, so that's good!
This is the only video the player has an issue with. No other video shows that touch from inside +1%...
What player are you using?
Cowboydude99
17th October 2019, 20:41
Jriver media center
Asmodian
17th October 2019, 20:44
It is definitely doing something different... that target resolution can only be set by the player.
Cowboydude99
17th October 2019, 20:50
I could see that, but what would do that!? ha
If it plays every other 4K video properly, it seems that either there is a problem with the file itself, or maybe an incorrect tag or something in the software player?
Asmodian
17th October 2019, 20:52
Try MPC-HC or Zoom Player... or something else? If it is reproducible there are many here who could investigate the technical details. :)
Klaus1189
17th October 2019, 20:56
Try MPC-BE ;)
https://disk.yandex.ru/d/r7JYQ2NjHvE4lA/Beta/1.5.4
Cowboydude99
17th October 2019, 21:06
Thanks! I'll give it a shot.
Blight
18th October 2019, 13:04
Has anyone using Windows 7 recently noticed MadVR leaking GPU memory?
For some reason, after playing some media, I see Zoom Player's GPU memory usage go up to 1.8gb (out of 2gb, with wdm/csrss taking the rest) and eventually MadVR will not be able to create the Direct3D device or just show an empty image.
Even after closing the graph, I can see that Zoom Player uses GPU memory (which it doesn't natively, only the filters allocate GPU memory), which leads me to suspect it's a leak.
I haven't been able to find that culprit, the only major piece of software I installed was office 365 and windows update patches.
Tried updating to the latest NVidia drivers, didn't make a difference.
Windows 7 64bit fully patch.
NVIDIA GT1030 latest drivers.
LAV Decoding (DXVA2 native).
MadVR set to use:
1. windowed overlay rendering.
2. 6cpu/6gpu queue size.
3. presenting (windowed mode) 6 frames in advance.
4. chroma upscaling : lanczos 3 taps
5. image downscaling : lanczos 3 taps
6. image upscaling : lanzos 3 taps
7. use DXVA chroma upscaling when doing (both).
Cowboydude99
18th October 2019, 13:58
It is definitely doing something different... that target resolution can only be set by the player.
Well you were right. I looked at the tags for the file within the software. It had some odd playback options set. Once I cleared them, it worked as expected.
Thanks so much. I didn't even realize the player had playback options per file.
All good now!
huhn
18th October 2019, 14:40
@Blight
i guess you are playing UHD with 2 GB vram?
i would not call it a leak but madVR can use more Vram from time to time. GPU-z and other tools Vram usages is sadly very inaccurate and
iknow that madVR uses more Vram when some settings setting are changed until it is restarted. the same can happening if a webbrowser in the back.
queues of 4 and a backbuffer of 4 gave me the best results on these systems with nvidia this was totally different with AMD.
DXVA2 native may cost more vram give copyback a try.
Blight
18th October 2019, 17:31
@huhn:
With regards to the leak, I reported it after MadVR's graph was released, in theory it shouldn't leave any vram allocated to zplayer.exe in process explorer.
Yes, it's really bad with 4K video, but even at 1080P eventually it will ran out of vram and report the Direct3D device can't be created (and sometimes even ZP's regular windows UI will appear as completely white since there is no vram to display it).
I'll try copy-back, initial testing shows it drops vram use to ~1GB but nearly doubles the GPU usage (from ~40% to ~77%).
Sadly, there's is only 1 passively cooled card with more than 2GB vram I could find, the palit kalmx 1050, but it's not really available locally.
Shame there isn't an 1030 with 4GB.
Klaus1189
18th October 2019, 17:52
Take a look at semi passively cooled cards. Fans are off when using it for lighter loads.
huhn
18th October 2019, 19:16
there is something not working as intended there. if you mean 1080p output the source resolution shouldn't affect it to much.
most performance cards can be changed in a way that they can run passive. it's just quite some work.
MadVR's graph
is that from a test build?
Blight
19th October 2019, 11:18
@hunn
No, the latest final release, but it may be a side-effect of running out of vram.
I'm aware that I can probably buy a water-cooling block for higher end cards, but it's too much of a hustle for me these days.
el Filou
19th October 2019, 16:15
I use a media center app that never exits even when putting the HTPC to sleep so it frequently runs for a month non-stop between Windows updates, and I've never had a problem with system or graphics memory leaks when using madVR with it. I have a very different software setup though.
@Blight Have you tested other players just to check if the same thing happens with them too or if it's just with ZoomPlayer?
mclingo
22nd October 2019, 10:07
I havent been following the other HDR thread, out of interest has there been any news on the next MADVR update, V1.0, I thought it was coming soon.
SamuriHL
22nd October 2019, 10:55
I very much doubt it'll be soon but they're still working on tone mapping. And making huge gains.
Sent from my SM-G975U using Tapatalk
BetA13
22nd October 2019, 12:57
@Blight
i also have Windows 7 (fully patched) and using an gtx670 with 2gb Vram and i didnt notice those issues..
Since we have both win7 and a 2gb GFX Card, im offering my help for any testing if needed..
Best Regards, BeTa
Siso
22nd October 2019, 14:53
Does anyone encountered an issue - madvr detects the monitor as new, after each restart. I have to point the 3dlut path everytime, and to choose the display as a digital monitor.
Manni
22nd October 2019, 15:29
Does anyone encountered an issue - madvr detects the monitor as new, after each restart. I have to point the 3dlut path everytime, and to choose the display as a digital monitor.
I had a similar issue with build 1806 that went away when I upgraded windows to build 1903. That's with the latest nVidia driver and the latest madVR test build. If you bring everything back in time it should work again :)
Siso
22nd October 2019, 15:35
I had a similar issue with build 1806 that went away when I upgraded windows to build 1903. That's with the latest nVidia driver and the latest madVR test build. If you bring everything back in time it should work again :)
I'm on windows 7 ultimate 64 bit. Video drivers 385.28.
Manni
22nd October 2019, 15:45
I'm on windows 7 ultimate 64 bit. Video drivers 385.28.
No idea then, unless you're using more than one monitor, which can cause similar issues.
mclingo
22nd October 2019, 15:48
I very much doubt it'll be soon but they're still working on tone mapping. And making huge gains.
Sent from my SM-G975U using Tapatalk
I was hoping MADHI would at some point soon start looking into to better compression / streaming artefact algos to help with poorly compressed streamed material and some poorly mastered blurays, there hasnt been a major dev on anything but tone mapping for ages. :(
huhn
22nd October 2019, 16:13
what's wrong with the current ones?
that's the usual way things happen here he finished something after month -years of worl and then it is like this for a long time.
the best way to change madshi mind is to proof there is something wrong with the current implementation that was done countless times in the past and is done right now with tone mapping.
tone mapping work will not change anytime soon a product he works on is supposed to be shipped in this year.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.