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 12th September 2019, 02:39   #601  |  Link
Calvi
Registered User
 
Join Date: Jul 2019
Posts: 38
elFilou: Thanks, you are right, that's the one setting I hadn't tried yet. Assumed it wouldn't be relevant for 25fps content. Should never assume anything.

If I disable IVTC then DXVA2 CB does handle the rolling balls correctly. There is still a difference to CUVID though. There is shimmering (as you mentioned) most noticeable in the broadcast logo, especially if you pause or resume, or frame by frame. Its there at playback speed though as well.
This does not happen with CUVID with IVTC ON or OFF so I'm still getting a different (and better) result with CUVID.

You are right that I don't have a lot of need for IVTC being in Australia as most of our content is PAL and some derivative of 25fps. Of course there may be cases where I get a video that is converted from NTSC but this in unlikely to be something I am too fussed about picture quality.

Your observations with Radeon are identical to mine. It's a bit late to detect the first ball and the same on the last (would be because of the delay for the last ball to appear).

huhn: Possibly IVTC is broken in CUVID as turning it On or OFF makes no difference to the result (on this source).
I would need to try a source that requires IVTC to verify this.

IF DVXA2 CB needs IVTC off to work properly and CUVID can't do IVTC (not confirmed) but de-interlaces better than DXVA2 CB then I am still staying with CUVID for my sources.

My whole reason for starting this discussion is that I had read that CUVID and DXVA2 CB use identical de-interlacing firmware and that CUVID was no longer recommended so I tried converting over to DXVA2 CB but was not getting the same results.

Perhaps it would be good for Nev to check this out and I am all for the switch to NVDec if it gives the best of both worlds.

I will add though the power saving of DXVA2 over CUVID is overstated if using MadVR to near max the card as this will end up using the same power anyway. If not using MadVR then DXVA2 will be a fair bit more efficient as it throttles the clock much more aggressively.

On the flip side CUVID changes clock frequencies in huge steps and stays stable with fluctuating loads so the need to change the driver from optimal to power-adaptive etc is also not required.
Calvi is offline   Reply With Quote
Old 12th September 2019, 02:55   #602  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
NVDEC is just a repack CUVID.

nvidia can't IVTC(or noone has an implementation of it) it is just there name it will not recreate the original frame rate.
huhn is offline   Reply With Quote
Old 12th September 2019, 19:53   #603  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
You mean it doesn't decimate?
I feel like we've had this same discussion before, but with 25 PsF content even if it doesn't decimate I definitely see a difference when I check the 'use ivtc' box or not (I remember watching a HD stand-up show with colored metallic ribbons in the background, and the shimmering was horrible with ivtc off), so it's at least detecting the 2:2 cadence. I don't know for the others.
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 12th September 2019, 20:50   #604  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
i didn't say it doesn't do anything. 60i in gives 60p out with that setting or without.
huhn is offline   Reply With Quote
Old 13th September 2019, 07:59   #605  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
AMD released new Radeon Driver 19.9.2
Klaus1189 is online now   Reply With Quote
Old 16th September 2019, 10:19   #606  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
I have found some typos in Radeon driver and already submitted them to AMD, but it isn't fixed yet. Here you can see screenshots of these two typos:
"% 2" has to be "%2", so there is one space, where it shouldn't be and the wildcard doesn't work
"Zurücksetzten" has to be "Zurücksetzen", so there is one "t" too much

What else can I do to reach the real people who are programming or translating that?
Klaus1189 is online now   Reply With Quote
Old 16th September 2019, 11:15   #607  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
AMD Forums?
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 20th September 2019, 21:17   #608  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
I have some kind of tinted banding with Win10 18362.356 and Geforce 436.30 while having the Calibration Loader disabled in Task Scheduler.
Everything is fine as long as I keep the Desktop in SDR mode but some elusive magenta tinted posterization shows up on certain scenes if I switch to HDR mode which persists until a reboot.
Using madVR with NVAPI HDR doesn't trigger this bug, only Win10 HDR does (and no other software is needed, simply flipping the HDR switch does the trick).
I only notice this magenta thingy on HDR content but I guess it's always there (just even less visible). But strangely enough it's not obvious to spot. W,R,G,B,C,M,Y gradient ramps look mostly fine (nothing is ever completely *perfect* but none of these look obviously broken). Most of the scenes in any random movie look fine until some problematic shades bring this out (but then it's clearly noticeable because grayish shades tend to turn into a distinctly magenta shade...).
I am not sure when this started because I didn't use the OS HDR mode. I recently started using the Netflix app from the Windows Store and that requires the OS HDR mode for HDR content. That's when the problems started (and it was hard to figure out where it comes from).
janos666 is offline   Reply With Quote
Old 21st September 2019, 08:40   #609  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
Did it came with Windows update? Are/Were previous drivers also affected?
Klaus1189 is online now   Reply With Quote
Old 21st September 2019, 11:32   #610  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
Quote:
Originally Posted by Klaus1189 View Post
Did it came with Windows update? Are/Were previous drivers also affected?
Unfortunately, I can't tell. I preferred to use madVR with NVAPI HDR for the sheer convenience of seamless auto-switching (these days the transition happens without any flashing to black or noisy screen, it's really neat).
The sole reason I started flipping that Win10 HDR switch was the Netflix app. (I used to go for the TV's built-in app but that's harcdoded to use DolbyVision for HDR and the TV's latest firmware has some tone response curve issues in DolbyVision mode, so I tried to get around that with the Win10 app which is HDR10-only...).
The only other time I flipped this switch (or more precisely: had it toggled automatically for me by a user software) was when I played an HDR video game (Anthem doesn't support NVAPI HDR, or even DolbyVision for that matter). But I didn't play that game since late July and this is so subtle and elusive that I doubt I could even spot this in a colorful sci-fi game (it could easily look like a poorly compressed texture or strange design choice of colors in that context).

Edit: Well... It's worse than I thought. I decided to try taking some photos (NV / OS comparison) to show how it looks like but the magenta tint was already there with NVAPI! I am sure nothing triggered the Win10 HDR mode since the last reboot. I didn't mess with anything in the Win10 Settings. The only change was the switch from 60 Hz 8bit to 23 Hz 12bit in NVCP. Yesterday I kept rebooting with 23Hz 12bit.

Anyways, here are some bad quality photos (-> see the attachments below). It looks worse in real view. Look at the outlines of the white clouds. I saw something similar in a different movie on white shirts in dark rooms.
Even now, I don't see any anomalies on the SDR desktop (I am using this same TV as a PC monitor to type this) and I don't see anything similar on gray ramp test patterns either (neither in SDR, nor in HDR mode). I think it affects a limited amount of shades close to white (but not quite exactly pure white).

Edit2: Ridiculous.
- I didn't even reboot, just switched back to 60Hz 8bit, then launched MPC-HC again with the same movie. madVR automatically switched to 23Hz and the bit depth automatically flipped to 12bit (this is the expected behavior if that resolution+refresh mode was used with 12bit earlier) and there is no magenta tint on the clouds now.
- Then I switched the Windows Desktop to HDR (while in 60Hz 8bit) and there is still no tinting (madVR yet again switched to 23Hz and 12bit came along with it).
- Finally, I switched the display from 60Hz 8bit to 23Hz 12bit manually from NVCP (with Win10 HDR mode ON), launched MPC-HC (obviously there was no display mode switching by madVR now) and the clouds show tinting again. The same thing happens if I switch to 23Hz from Window's advanced display settings (so it's not NVCP but the NV driver or the OS).

Workaround (a stupid one): set the desktop to 60Hz 8bit, toggle Win10 HDR, start MPC-HC to get madVR switching to 23Hz (and 12bit along with it), start the Netflix app and watch HDR content. (LOL)
Attached Images
  

Last edited by janos666; 21st September 2019 at 21:01.
janos666 is offline   Reply With Quote
Old 22nd September 2019, 11:45   #611  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
Quote:
Originally Posted by el Filou View Post
AMD Forums?
Thanks, I registered and posted it there:
https://community.amd.com/thread/243762
Klaus1189 is online now   Reply With Quote
Old 22nd September 2019, 17:57   #612  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
I checked my AMD Vega based notebook (2500U with 19.9.2) with the same TV and it's worse than the nVidia Pascal (GTX1070) based desktop PC. The AMD one doesn't show distinct magenta tinting but the banding is much worse (at least with >8bit) and something is wrong with the HDR10 color gamut (wrong metadata?). The colors are much less saturated than they should be (faces look pale, etc).

By the way, the VGA calibration LUT is always set to something "fuzzy" after any changes are made to the Windows display settings (with the CalibrationLoader disabled in Task Scheduler) on both systems.
When it's a custom non-neutral but smooth curve then the graph plot of DisplayCAL starts showing a lot of aliasing (it gets distorted each time I apply a change in Windows display settings but smooths out again if I reload the curve with DisplayCAL).
When it's neutral (no custom ICM profile) the plot looks mostly fine, except it's slightly drifted: 255 maps to 254.1, etc. (This is not so crazy but still clearly distorted.)
But this disturbed LUT is not the only issue. The banding remains after I manually clear the LUT with DisplayCAL. This is just yet another bug (although they might be related).

I initially thought the common denominator will either be <60Hz or >8bit (applied manually from Windows display settings) but it's not so easy. The AMD system doesn't seem to care about these (although the banding is much worse with 12bit than 8bit but that's probably just AMD's always-on dithering).
The color saturation bug (AMD only) seems to require some "trigger" to get stuck (but I didn't pinpoint this yet, it's probably a switch to HDR10 mode with either Win10's or AMD's private API, I am not sure yet).

This is incredible. Several months have passed and 1903 is still garbage. And they didn't even have a real 1909 to work on.
I decided to switch the notebook to the Fast ring and see how that goes. -> No luck, the current 20H1 build behaves the same way on the AMD system (I already rolled the update back).

Edit: This never ceases to amuse... TV nVidia PC lost the magenta tint (which only appeared in certain display modes) when I switched the LAV Video Decoder from DXVA2(native) to D3D11(native). But I didn't find any working option for the AMD system (that one has much more serious banding, that's something else).

Last edited by janos666; 23rd September 2019 at 00:16.
janos666 is offline   Reply With Quote
Old 23rd September 2019, 23:56   #613  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,406
Thanks for the diagnostics on these very odd and evil issues with 1903!
__________________
madVR options explained
Asmodian is offline   Reply With Quote
Old 24th September 2019, 10:06   #614  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by janos666 View Post
Edit: This never ceases to amuse... TV nVidia PC lost the magenta tint (which only appeared in certain display modes) when I switched the LAV Video Decoder from DXVA2(native) to D3D11(native).
That's interesting.
Anyway, just use D3D11(native) in LAV with madvr, it's the fastest mode.
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config
chros is offline   Reply With Quote
Old 24th September 2019, 11:08   #615  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
@janos666 which build of madVR are you using, and are you doing passthrough or processing?
I just tried DXVA2 native because I found it strange that it would give different rendering results (I never use it due to NVIDIA lossy issue), and noticed that at least with beta test build 86 it skips some HDR rendering steps (and gains alot of time in the process!).

With D3D11VA native it's (2160 to 1080 with scale chroma separately and full HDR processing including highlights recovery):

1. DXVA11 Interop
2. Image downscaling
3. HDR Blur Dif; Blur; Frequency Split
4. Chroma
5. rest of HDR processing
6. HDR Final
7. Final Step

With DXVA2 native, apart from step 1 that is obviously replaced by something DXVA2-specific, madVR completely skips steps 3 and 6 (which I think have to do with highlights recovery).
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 24th September 2019, 23:37   #616  |  Link
janos666
Registered User
 
Join Date: Jan 2010
Posts: 479
@el Filou - v0.92.17 (the public build from the neighbor topic), passthrough (metadata included), minimalist setup (no post-process filters or overly fancy resamplers). I am not sure why I had LAV set to DVXV2, I guess I kept flipping it around while fighting the old Win10 1903 banding issue (with the Calibration Loader) and forgot about it.
NVIDIA lossy issue - Hmm? Is DX11 supposed to be of higher quality? I didn't use the DXVA2 resampler (in madVR's settings) if that's what you mean, I am aware that's broken (results in visible color luminance errors).

Last edited by janos666; 24th September 2019 at 23:40.
janos666 is offline   Reply With Quote
Old 25th September 2019, 00:23   #617  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
DXVA2 native and only native lowers image quality the rest is bit perfect.
huhn is offline   Reply With Quote
Old 25th September 2019, 16:01   #618  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
AMD Radeon driver 19.9.2 is available as Recommended (WHQL)
@RX 5700 (XT) card owners: Does this driver run without issues like BSOD for RX 5700 (XT) cards?
Klaus1189 is online now   Reply With Quote
Old 25th September 2019, 16:27   #619  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
the main BSOD problem was the hardware acceleration one that's "fixed".
now the driver crashes instead and usually recovers.

this is still in the known "issues":
"Discord™ may experience an application hang on Radeon RX 5700 series graphics products when HW acceleration is enabled."

things that some people help are.
don't use PCIe 4.0.
don't use 240hz and at best only 60hz.
use abba if available why this should have anything todo with the GPU i don't even wanna known.
use 1903 or newer see just a couple of post a top.

there are system that in general run that's usually a single 60 hz system.
that's the same driver with WHQL which is not worth much these days.
huhn is offline   Reply With Quote
Old 26th September 2019, 17:41   #620  |  Link
Klaus1189
Registered User
 
Join Date: Feb 2015
Location: Bavaria
Posts: 1,666
@GTPVHD: I want to create the Intel list but I must admit that I don't get the version numbering sheme of Intel and what chip is supported in each of the long version numbers. If you want you can create a short list and point me in the right direction, for example all version xx.xx are only for this chip / generation of intel processors.
Klaus1189 is online now   Reply With Quote
Reply

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 12:13.


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