AMD, Intel and Nvidia driver issues and last recommended version

[janos666]

not this again...
every* commercial file you play that is about to be converted to RGB is 32 bit float this counts for DVDs, BD, UHD BD. if a file is encoded in 8, 10 or 12 bit has nothing todo with this.

or with other words you could never check if a file gets out without truncation rounding or dithering because it's not possible.

*add unimportant number of exclusion here
I am not entirely sure what you mean. I know video files are YCC and that's always converted to RGB on PC. But seriously... Open a 10bit gradient pattern, disable dithering and then check if you see any difference between 8bit and 10bit output modes (on a >=10bit display).
I obviously meant something like "this gradient now visually like 10bit equivalent to my naked eyes within the margin of error" or "this gradient now looks more like 8bit equivalent because there are a lot less visible steps and more bands when compared to the previous test".
I didn't know I have to add so much footnotes to get the message through instead of getting hung up on every technical details.

[huhn]

I obviously wish to output 10bit if both the source and the display are native 10bit.
this is the core issue is just not how this works.
there is a reason i use a 16 bit RGB gradation test.

and for some reason if someone sees banding it is always set with a mindset of no dithering.

[janos666]

this is the core issue is just not how this works.
there is a reason i use a 16 bit RGB gradation test.

and for some reason if someone sees banding it is always set with a mindset of no dithering.

Again. I suspect the 10bit output of madVR (or anything else) is converted (in bad quality, so not by dithering but truncation/rounding) to 8bit by the WDM or the driver before the GPU outputs 10bit to the display.
How else should I test this other than disabling dithering (for testing purposes), using >=10bit test images and switching between 8 and >=10bit output formats (either or both in madVR settings and GPU settings).
I can't visually see too much difference if everything is dithered for 8bit anyways. That masks the suspected error.

That does not sound right. Are you using a Windows calibration or something? Overlay looks correct on my system.

No. My secondary display (cheap old LCD) has an ICM profile. Windows handles the LUT loading (DisplayCAL's LUT Loader caused issues with DolbyVision games launched on the primary display). The main display (UHD HDR TV) runs with the system default sRGB profile.
But now that you say... I guess it's possible the overlay mode applies the calibration LUT of the secondary display to the output on the main display. The secondary display LUT is very far from neutral (the display is set to do the least amount of processing on it's own because it's processor is very stupid).
However, this overlay is not very useful to me nowadays either ways because HDR10 doesn't work with it.

[huhn]

if you want to test if something is rounding to 8 bit you should leave dithering >on< if the image still get's rounding based 8 bit banding which is very very easy to see the dithering will be undone and should looks the same as madVR 8 bit output without dithering everything inbetween should be an issue on the TV side which is pretty much confirm on all TV which rtings limited banding test.

you can easily test banding with an 8 bit yCbCr source file that is dithered to 10 bit the same way.

so using dithering isn't hiding other parts that are not dithering.

[ryrynz]

How else should I test this other than disabling dithering (for testing purposes), using >=10bit test images and switching between 8 and >=10bit output formats (either or both in madVR settings and GPU settings).


That's how I did it and Asmodian told me off [emoji28]
I think the takeaway here is that 8bit with dithering is actually very good and is in a similar level if not actually sometimes better than 10bit with dithering. I can't see a wrong choice here if you see little to no difference between the two so I wouldn't get hung up on it.

[janos666]

That's how I did it and Asmodian told me off [emoji28]
I think the takeaway here is that 8bit with dithering is actually very good and is in a similar level if not actually sometimes better than 10bit with dithering. I can't see a wrong choice here if you see little to no difference between the two so I wouldn't get hung up on it.

Alas, I suspect this entire bitdepth conversation also lost it's context of suspected nVidia WDDM 2.6 drivers / Windows 1903 issues.
I wasn't searching for the optimal settings on my set of hardware in general. This conversation started when some of us updated to the latest Windows 10 build and accompanying GPU drivers. Most of what I tried to explain were attempts to diagnose that problem, not to fine-tune a system.

[huhn]

i'm mostly pointing at the flawed test to for this.

and do not forget we are now from d3d9 overlay does the same to not working to fixing it.

[oldpainlesskodi]

did you make sure that stuff like d3d11 and FSE are disabled while testing with overlay?

Tried again, and yep, as a test, it fixed the issue for a wddm 2.6 driver.

So what have they broke?

[huhn]

the fact that overlay is perfect as always means nvidias driver rendering path which should always be the same is still working fine and something at the wddm level goes totally wrong. so this whole issue could be totally on microsoft.
we just need a clear conformation from an AMD card the earliest i will get an new AMD card is still far over a month so... someone else should do that.

a new intel iGPU may be better for testing this overlay works there too.
looks like my hd 4400 still get's driver if it has wddm 2.6 support now i will test with that GPU later.
edit: i don't have the new windows version on this system and i'm not eager to get it on this system so i'm not going to test this anytime soon.

[janos666]

we just need a clear conformation from an AMD card the earliest i will get an new AMD card is still far over a month so... someone else should do that.


My mobile Vega produces banding with both 19.5.2 (generic AMD driver from amd.com, WDDM 2.6) and 17.7 (HP manufacturer packaged version from Windows Update, WDDM 2.3). It looks like only a small number of WDDM 2.5 drivers might work correctly with Win10 1903 (an old WDDM 2.1 nVidia driver I tried also produces banding).

[oldpainlesskodi]

Ok, good to know it's not just me then...as I mentioned, on my Nvidia setup, with the latest version of windows, wddm 2.5 is ok, 2.6 is not (excluding Hunn's work around) .

[TK423]

Nvidia user here. Haven't updated my drivers since 417.71.

In order to fix the HDR issues with the 417.71 drivers I...


Used 7-Zip to extract the installer files (417.71-desktop-win10-64bit-international-whql.exe) to a folder
Deleted the HDAudio folder from the extracted 417.71 installer files
Replaced the deleted HDAudio folder with the one extracted from the 388.59 installer files.

This allowed me to install the 417.71 drivers while retaining HDR ability with MPC-HC and MadVR.

It's been about five months since I've updated my GPU drivers and since I also use my PC for gaming, I'd like to update them.

Does anyone know if the aforementioned fix will work with the newest 430.86 drivers - deleting the HDAudio folder and replacing it with the one from the 388.59 drivers?

[hotripper]

Why wouldnt it?

But may be unnecessary as I believe the HDR nit/color issue was confirmed as fixed a couple of drivers ago.

[SoySauceWDF]

Still getting stuttering on an RTX 2080ti, latest drivers, applied the nvidia profiles. Audio seems to be out of sync now as well

[el Filou]

(I know this thread is about GPUs and madVR, but as the graphics drivers also include the audio drivers and a lot of HTPC users have AVRs I figured it's still on-topic. If not, I apologize)

Has anyone here using an NVIDIA and HDMI out to an AVR noticed that it does a 'pop' sound when changing refresh rates?
I am on Windows 1809 (but the problem was already there on 1803), 1050 Ti with driver 430.64, standard driver timings no custom mode, the AVR is 10-year old and its firmware and config haven't changed in ages so I'm pretty sure it's not the cause.
Ever since I bought the card two years ago I never heard anything when changing refresh rates, but for a few weeks now I've noticed that I hear a pop sound from the AVR, like the sound old analogue PC sound cards made when initializing their driver at boot.

At first I thought it was due to an NVIDIA audio driver problem so I tried older drivers but it didn't fix it. I tried rolling back Windows updates but it didn't fix it either. Now after having search all over the web and not finding any forum thread on this issue I'm out of ideas.

This is getting on my nerves because I frequently change display refresh rate between 50 Hz and 24 Hz and I can't stand hearing that pop sound which wasn't there before.
Something weird I've verified: it doesn't happen if the sound format set in Windows control panel (for shared mode) is higher than 48 kHz. Anything 48 kHz and below, be it 16 or 24 bit, gives a pop sound on refresh rate change, and anything higher than 48 kHz doesn't.
I don't want to use 96 kHz because it gives lower quality with my hardware and is more processor-intensive with ReClock.

Anyone has any idea what could be causing this, or has noticed the same thing? If yes, was it just after a driver update?
Thanks.

[Manni]

Did you try bitstreaming? In my setup I don't get any pop.

[hotripper]

Anyone has any idea what could be causing this, or has noticed the same thing? If yes, was it just after a driver update?
Thanks.

The only thing that ever gives me a pop is Dirt Rally and DR2.

[hotripper]

Still getting stuttering on an RTX 2080ti, latest drivers, applied the nvidia profiles. Audio seems to be out of sync now as well

Stuttering in what? Madvr? MPC? Games? Everything video?

Welcome to RTX. Youre not alone. MAdvr in particular and MPC-HC give many of us heartaches.

Try CUVDA for MPC custom presenter, try copy-back for Madvr and set frames in advance to 1 in madvr settings. This has worked for me for 2 months now. Ive just stopped caring tbh

[SamuriHL]

Ok, well, I now know what a potential cause of the stuttering is for RTX users because after a clean install of windows (don't ask....seriously, I'm beyond livid right now with my new machine and lost everything due to a massive SSD crash) I now suffer from it, too. I'm fairly certain it comes down to DCH vs standard drivers. I will test this theory tomorrow and nuke the DCH driver and replace it with the standard and see if the stuttering issue goes away. Note that I configured everything the way I had it before the crash and am getting the stuttering issue. No reported drops in madvr and I'm jamming around 24ms for a UHD with HDR tone mapping on my 2070. Yea, this is fun. LOL

[ryrynz]

Well if it helps, if you do find the cause, you potentially help a lot of people.. Fingers crossed.

[SamuriHL]

That was my thought too. Believe me I'll be all over it today. I'll post what I find as soon as I find it.

Sent from my SM-G975U using Tapatalk

[el Filou]

Did you try bitstreaming? In my setup I don't get any pop.Bitstreaming or decoding doesn't change the issue because the popping sound happens on refresh rate change, even in Windows when I'm not playing any media. It's like the NVIDIA disables the audio output for a very short time when changing the display refresh rate and enables it again right after. I also didn't get any pop before and I can't pinpoint what exact change is causing this. :(

[huhn]

i have this with analogue audio too my guess is a new thing with newer windows versions.

[el Filou]

Oh my. Thanks for confirming though.
And this also happens if you change refresh rates in rapid succession, like less than 10 seconds?
I know that now the NVIDIA audio driver forcefully closes the HDMI audio stream after 10 seconds of no sound to save power but I've edited the registry to fix this, so it isn't the cause.

[SamuriHL]

Well it's not going to f***ing happen any time soon as my system crashed again overnight and won't boot anymore. The brand new 970 pro appears to have completely f***ing died and I'm so far beside myself given what I paid for it it's not even funny. All the work I spent last night rebuilding the goddamn thing is gone again. WTF Samsung!

[huhn]

Oh my. Thanks for confirming though.
And this also happens if you change refresh rates in rapid succession, like less than 10 seconds?
I know that now the NVIDIA audio driver forcefully closes the HDMI audio stream after 10 seconds of no sound to save power but I've edited the registry to fix this, so it isn't the cause.

i don't have the issue anymore i don't run new windows version at the moment.

i personally would just ignore this "issue" an audio pop with an resync is sadly relative normal with all kinds of devices.

i'm using an bad expensive PCIe sound card not an good realtek onboard soundcard and this can easily the reason for me. can't wait for my ryzen with proper onboard soundcard...

[gonca]

Well it's not going to f***ing happen any time soon as my system crashed again overnight and won't boot anymore. The brand new 970 pro appears to have completely f***ing died and I'm so far beside myself given what I paid for it it's not even funny. All the work I spent last night rebuilding the goddamn thing is gone again. WTF Samsung!

What MoBO do you have?
Never had a Samsung SSD (or NVME) just die
Did you try unplugging the power supply from the wall for a minute?
Power supply still good?

[SamuriHL]

Yea, it's not the SSD. It's a much more complicated problem, unfortunately, which may end up ultimately coming down to my RTX 2070. I spent too much time this morning debugging the issue already so I have to set it aside til tonight. The issue is more bizarre than anything I've ever seen on a PC, though, which troubles me as everything is brand new within the last 45-60 days. The SSD, however, is not the issue and has been ruled out.

[hotripper]

Oh my. Thanks for confirming though.
And this also happens if you change refresh rates in rapid succession, like less than 10 seconds?
I know that now the NVIDIA audio driver forcefully closes the HDMI audio stream after 10 seconds of no sound to save power but I've edited the registry to fix this, so it isn't the cause.

pLE4ASE tell me wherein regedit and how u did this I am interested!

[el Filou]

@hotripper

1. Device Manager: open the sound controller device properties => Details => copy the 'Driver key' property
2. Regedit: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\<********paste Driver key here********>\PowerSettings => change the PerformanceIdleTime key to 00 00 00 00

(there's also a ConservationIdleTime key if you want a different timeout for when a laptop is on battery)

source: https://docs.microsoft.com/en-us/windows-hardware/drivers/audio/portcls-registry-power-settings

Disclaimer: this may make your AVR consume more power when no audio is played, depending on its design

[gonca]

Yea, it's not the SSD. It's a much more complicated problem, unfortunately, which may end up ultimately coming down to my RTX 2070. I spent too much time this morning debugging the issue already so I have to set it aside til tonight. The issue is more bizarre than anything I've ever seen on a PC, though, which troubles me as everything is brand new within the last 45-60 days. The SSD, however, is not the issue and has been ruled out.

Have you ruled out the Ram sticks or the sockets?

[SamuriHL]

Yea, I've narrowed it down to video related issues. I've pulled the 2070 right now and am reinstalling windows with just the built in Intel whatever on the 9900k. madvr isn't going to love that but this should only be temporary until I can determine if it's a defective card and get it replaced.

[hotripper]

@el Fijou
Thanks. works, So far so good! Its annoying how my 5.1 through dvi/hdmi sleeps but stereo hdmi through tvARC never sleeps so power concern is none for me, I just want both outputs to act the same. Awesome tip!

[SamuriHL]

Alright, my issue is on its way to being solved. After so much troubleshooting it has been narrowed down to a hardware issue with the 2070. nVidia is RMA'ing it and my system is running with my 1060 right now. Good thing is I guess the windows clean install? LOL It is a brand new machine.

[oldpainlesskodi]

The banding issue on Win 10 1903 looks like it's been picked up:

https://hub.displaycal.net/forums/topic/severe-banding-in-windows-1903/

https://www.reddit.com/r/Windows10/comments/bzed4h/windows_10_1903_banding_vertical_strips_issue/

https://www.reddit.com/r/Windows10/comments/bs7vr9/so_i_managed_to_find_out_why_some_people_who_use/

[ryrynz]

Yeah users picked it up ages ago, has it been picked up by Nvidia/MS though? It's entirely the reason why I've held off 1903 on my HTPC. Really nice update though apart from this issue.

[el Filou]

The really depressing part is this: "I’ve reported the issue both to nvidia, and Microsoft multiple times since the issue has been there for months in the insider builds which is now RTM but seems that nothing is being done about it."

They mostly depend on insiders for QA but even when issues are reported officially and widely they don't get fixed before release. :(
It's like Windows is some video game for them now instead of, you know, a critical part of using your computer.

[ryrynz]

That honestly does not surprise me, communication is often poor unless you're dealing with an actual person. Pity Madshi ain't around to help sort it.. providing him all the info and having him forward it on would've been the ideal solution.
Have you spoken with ManuelGuzman or tweeted @windowsdev? Both of them appear to be reasonably useful in making things happen.

[oldpainlesskodi]

Yeah users picked it up ages ago, has it been picked up by Nvidia/MS though? It's entirely the reason why I've held off 1903 on my HTPC. Really nice update though apart from this issue.

Yeah, true...I guess the more noise created, hopefully the better the chances of it getting fixed.

[ryrynz]

It will be fixed, it'll just be better chances of getting it fixed promptly. Wonder if it's even on their radar?

[oldpainlesskodi]

It will be fixed, it'll just be better chances of getting it fixed promptly. Wonder if it's even on their radar?

I'm not a member on the Nvidia or MS forums, so hopefully those that are (and here) will help flag the issue.

[el Filou]

i personally would just ignore this "issue" an audio pop with an resync is sadly relative normal with all kinds of devices.I know but I'm hypersensitive to harsh noises and it's really quite a loud pop if the AVR volume is set to movie listening level. :( I'm also afraid it may damage my speakers in the long run but maybe I'm being paranoid.

You're right that it's due to a change in Windows because I've just cleaned my system of all traces of NVIDIA audio driver and registry settings and went back to the generic MS HD Audio driver and it's still happening. It's probably a change on the graphics side, it must force the GPU to reset the whole HDMI connection when changing refresh rate or something like that and it probably is doing it dirty without cleanly cutting the audio channel first.
I wish if HDMI QMS eventually comes to PCs one day that it will fix this again.
I'm now manually muting the global sound output of Windows before changing my display refresh rate to avoid it, it's bothersome but it's a workaround. :sly:

[janos666]

Did anybody find an AMD driver that works with 1903 (no banding)?
Edit: This is madness. I decided to run DDU and try the latest 19.6.1 (WDDM 2.6) for one last time (before trying to revert to 1809) and now it's all fine. This leaves me wondering what really triggers the banding (or if running DDU changed anything).

[oldpainlesskodi]

Finally managed to use a wddm 2.6 driver....and such a simple fix (I had done this on a previous windows build, but didnt even think to check on 1903 build).

Credit to acesko92 from the nvidia forum:

"I managed to find a fix for the broken gamma, and nvidia dithering in windows 1903: you need to go to Task Scheduler > Microsoft > Windows > WindowsColorSystem, and delete the task called WindowsColorSystem Calibration Loader."


I just disabled the service and rebooted, but each to their own.

[janos666]

"I managed to find a fix for the broken gamma, and nvidia dithering in windows 1903: you need to go to Task Scheduler > Microsoft > Windows > WindowsColorSystem, and delete the task called WindowsColorSystem Calibration Loader."

Hah, funny. This workaround works for me too (with v435.27 on a machine which has a custom ICM profile for the secondary display).
But I am pretty sure I didn't have any custom ICM profiles installed on my AMD notebook (unless Windows Update tried to grab one from the manufacturer but I saw no indication of this anywhere and those rarely have a modified 1DLUT anyway) and it just went away between random driver installs (+ a DDU run but no task scheduler tinkering).

[el Filou]

I have filed a feature suggestion to MS Feedback Hub for adding a 'audio-only display output' feature to Windows (could be useful for HTPCs). Here's the link if anyone is interested: https://aka.ms/AA5emb3

[huhn]

no one cares? that's very disappointing...

[ryrynz]

Yeah, everyone jump into the Feedback Hub and upvote it.

[SamuriHL]

I upvoted it. I just think MS won't care. I've been through this with them already for a different kind of issue. Even when they did eventually add the feature I was looking for, it was for something completely different but they closed my issue and merged it with the other one anyway. So...good luck I guess?

[Nicog]

Hi!

Maybe I'm a noob but I still have stuttering on 430.xx drivers my my RTX. For me last working driver is 418.91...

I played with that (https://forum.doom9.org/showpost.php?p=1871970&postcount=151) for hours without any success (not sure also to have understood the purpose of the attached files with Nvidia inspector…).

I'm the only guy who still have this issue?