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
Register FAQ Calendar Today's Posts Search

Reply
 
Thread Tools Search this Thread Display Modes
Old 26th September 2019, 19:40   #621  |  Link
GTPVHD
Registered User
 
Join Date: Mar 2008
Posts: 175
https://www.intel.com/content/www/us...s-drivers.html

I don't know if this is enough to help you, but you can try reading it.

The version number will increase to 27.20.100.7xxx in the future because the next 1909 release of Win10 will be WDDM 2.7.
GTPVHD is offline   Reply With Quote
Old 26th September 2019, 21:48   #622  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
as far as i know a new WDDM version isn't planned.
the next update that is planned for this month is not like the others.

they are slowing it a bit down with the live releases and i hope do more quality control which i think is the right direction as long as new insider builds are provided.
huhn is offline   Reply With Quote
Old 27th September 2019, 00:53   #623  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
An updated 1903 is already 1909 with a few (unintresting) features disabled (a ~20kb update can enable those). 20H1 is also available for testing (with WDDM 2.7) but it seems to be the same as far as these graphics quality problems go (although I didn't see any WDDM 2.7 drivers from AMD yet).
janos666 is offline   Reply With Quote
Old 28th September 2019, 18:11   #624  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
Quote:
But in driver 19.9.2 the pixelformat is "RGB 4:4:4 Full RGB" ("4:4:4" for those who don't know that RGB is always without chroma subsampling.)
But madVR and VMR 9 windowed is fine. Strange, at least for my understanding.
not related to it. the GPU driver always assumes the windows desktop is full range (and that the video that is rendered there is too). so it only means full range RGB do "nothing" and limited range do a full range to limited range conversation.

this is an example for nvidia which effects EVR: https://abload.de/img/videorangeplkyj.png
huhn is offline   Reply With Quote
Old 28th September 2019, 18:15   #625  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
OK, thanks for the info. But I nevertheless don't understand why only EVR and EVR custom presenter is affected.
Klaus1189 is offline   Reply With Quote
Old 28th September 2019, 18:19   #626  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
they use DXVA processing so the GPU driver is in charge of YCbCr -> RGB conversation and if that it done to limited range and not fixed later it will be washed out.

if you have the issue with nvidia you can change the showed setting and force it to full range.

you can try the mpc renderer settings output and change the range to full range.
huhn is offline   Reply With Quote
Old 28th September 2019, 18:22   #627  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
Quote:
Originally Posted by Klaus1189 View Post
OK, thanks for the info. But I nevertheless don't understand why only EVR and EVR custom presenter is affected.
madVR is also affected if you enable the "trust DXVA color & levels conversion" performance option (and I guess this obviously needs DXVA2 decoding, but I am not sure if it applies to native only or copy-back as well, I also never tested this with D3D11 decoding) but it used to cause issues with HDR10 videos (the last time around I tried it, roughly 1-2 years ago).
janos666 is offline   Reply With Quote
Old 28th September 2019, 18:23   #628  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
If you are referring to the videorender page in options in both MPC-BE and MPC-HC, it is already set to 0-255, I already tried to set it to 16-235, if it changes anything, but it did nothing.
Klaus1189 is offline   Reply With Quote
Old 28th September 2019, 18:26   #629  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
Quote:
Originally Posted by janos666 View Post
madVR is also affected if you enable the "trust DXVA color & levels conversion"
Thanks, I tried it but madVR is still fine.
Klaus1189 is offline   Reply With Quote
Old 28th September 2019, 18:26   #630  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
Quote:
Originally Posted by Klaus1189 View Post
If you are referring to the videorender page in options in both MPC-BE and MPC-HC, it is already set to 0-255, I already tried to set it to 16-235, if it changes anything, but it did nothing.
No, this one (-> see the attached image below).

Quote:
Originally Posted by Klaus1189 View Post
Thanks, I tried it but madVR is still fine.
Did you try both NVCP choices while having this madVR option enabled and the decoder set to DXVA2 Native? (Things might have changed over the last year...)
Attached Images
 
janos666 is offline   Reply With Quote
Old 28th September 2019, 18:30   #631  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
Sorry post #640 was referring to huhn.
@janos666: I found it in madVR -> rendering -> trade quality for performance -> 2nd last feature. (I can not open your pic, because: "Attachments Pending Approval")
But as in #641 already posted, it is still fine, which I also don't understand.
Klaus1189 is offline   Reply With Quote
Old 28th September 2019, 21:22   #632  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
with my RX 5700 XT i'm limited to limited range too.

looks like AMD dropped the ball again.
the rendering output range is ignored.
huhn is offline   Reply With Quote
Old 28th September 2019, 23:27   #633  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
So, it turns out the heavy banding on the AMD Vega based notebook was also caused by the DXVA2 Native setting in LAV (I still can't remember why I set both to DXVA2, though I didn't know it was known to be fundamentally broken quality wise). The video image is clean with D3D11.
However, the color space metadata bug is unrelated and random. I found the signal info banner on the TV OSD (click anywhere once and then click on the HDMI icon on the top-left corner for LG) and "rec2020" disappears from the list when I see unsaturated colors. It seems to happen with both the AMD private API (madVR) and the Win10 HDR desktop modes (same movie file, the madVR OSD shows clearly recognized Rec2020 but even the converted Windows desktop GUI colors look off, so...). The latter is 50/50 (sometimes it's fine, other times it's off) but I don't remember seeing correct colors with the former (could be a coincidence or the private API is still useless, it was also obviously very dark with older drivers all the time).
janos666 is offline   Reply With Quote
Old 29th September 2019, 09:08   #634  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
Quote:
Originally Posted by huhn View Post
with my RX 5700 XT i'm limited to limited range too.

looks like AMD dropped the ball again.
the rendering output range is ignored.
I going to make a thread at AMD forum, but I am not sure how to make it 100% clear for the devs what exact range in the driver is affected. Can you help me finding the right descripion for this issue?

What I got now is that the "video output range setting" outputs 16-235 regardless what is selected 0-255 or 16-235. Note: It is not the output setting of the pixelformat.
Klaus1189 is offline   Reply With Quote
Old 29th September 2019, 09:45   #635  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
words are not my strong point.

the pixel format should be ignored it should always been outputted as full range from the video renderer even if the pixel format is set to limited or it will not match.

the problem doesn't trigger with the windows movie app so that hints that it is a directshow issue because it looks like it works with media foundation.

thanks to mpcVR(i use an old version) we get another hint that it is coupled with DXVA2 processing because it works fine with D3D11 processing.

the next problem is that this is a video issue and these companies usually don't really care about this as long as there is an image. the 5700 series is a just a driver nightmare i would say they have more important things to do.

so i would use something like:
videos are outputted in limited range in a directshow video player when DXVA2 processing is used.

for example play a video using MPC-HC using the default enhanced video renderer (custom presenter) the output range will be limited.

driver version card.
and add an dxdiag they love these like candy.
finding the driver where they broke this wouldn't be bad too.
huhn is offline   Reply With Quote
Old 29th September 2019, 09:45   #636  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
I just tested MPC Video Renderer with MPC-BE and it is also affected in standard settings:
Code:
DirectX 9
Graphics adapter: AMD Radeon RX 5700 XT (1002:731F)
VideoProcessor  : DXVA2 ProgressiveDevice
DeinterlaceTechnology: none
Display Mode    : 3840 x 2160, 59 Hz
But when checking Use Direct3D it looks fine to me:
Code:
DirectX 11
Graphics adapter: AMD Radeon RX 5700 XT (1002:731F)
VideoProcessor  : D3D11
Klaus1189 is offline   Reply With Quote
Old 29th September 2019, 09:47   #637  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
Quote:
Originally Posted by huhn View Post
thanks to mpcVR(i use an old version) we get another hint that it is coupled with DXVA2 processing because it works fine with D3D11 processing.
Yes
Klaus1189 is offline   Reply With Quote
Old 30th September 2019, 09:56   #638  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
just to add more to the fire they broke yet again deinterlancing on my system too.

so we currently have wrong ranges in DXVA2 "processing" NN as an "de"interlacer truncated 10 bit to 8 bit DXVA2 native decoding.
huhn is offline   Reply With Quote
Old 30th September 2019, 16:00   #639  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,667
Quote:
Originally Posted by huhn View Post
"de"interlacer truncated 10 bit to 8 bit DXVA2 native decoding.
Sorry, but I don't get that, can you please explain it so I can decribe it deeply on the AMD forum?



I just browsed on YouTube and came across the following videos:
https://www.youtube.com/watch?v=_rxFxdvO3fQ
https://www.youtube.com/watch?v=TY4s35uULg4

Could this method help Nvidia users to get more accurate 23 Hz timings out of the box?
And perhaps can then the madVR image processing be done in the Nvidia card and Framepacked 3D be sent over an AMD card or perhaps an Intel iGPU which lots of users also have already in their HTPC.
Just an idea, but perhaps it is helpful ...

Last edited by Klaus1189; 30th September 2019 at 16:02.
Klaus1189 is offline   Reply With Quote
Old 30th September 2019, 18:47   #640  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,925
i already talked and tested this.

yes it can be done when windows allows you to and it is dodge as hell.

about the decoding issue with native will try to make screens and such but this is a very old issue pretty much ignored because it is DXVA2 native which has issues with madVR in general.
huhn is offline   Reply With Quote
Reply


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 04:06.


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