Log in

View Full Version : AMD, Intel and Nvidia driver issues and last recommended version


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 [27] 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69

SamuriHL
6th May 2020, 13:33
Interesting. Dolby vision support eh?

Sent from my SM-G975U using Tapatalk

nsnhd
7th May 2020, 11:18
Should it be Core-i or a Pentium with HD610 is sufficient ? I think of a G4600 with HD630.

littleD
8th May 2020, 15:18
Not sure what support they claim because such file
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : Main 10@L5.1@Main
HDR format : Dolby Vision, Version 1.0,

Plays as green and pink. (Dolby Vision Profile 5 )

Edit.
Ctually
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : Main 10@L5.1@Main
HDR format : Dolby Vision, Version 1.0, dvhe.08.09, BL+RPU, HDR10 compatible / SMPTE ST 2086, HDR10 compatible

Plays ok. Its is described as Dolby Vision Profile 8.1

el Filou
8th May 2020, 17:59
Are you sure the Dolby Vision support isn't just for games?
I suppose support for media playback would need some proprietary decoding filter?

amayra
8th May 2020, 23:38
is there a different between intel windows driver and i965 Linux driver for 7gen igpu?

Blue_MiSfit
9th May 2020, 03:16
Dolby Vision profile 8.1 is an HDR10 base layer, so you're likely just seeing that. There's also an enhancement RPU (reference picture unit) that lets a compliant Dolby Vision decoder get more quality and get dynamic tone mapping metadata.

Profile 5 is non-backwards compatible, so you're seeing the IPT representation of the image interpreted as YCbCr by your player that doesn't know any better :) You need a real Dolby Vision decoder to play this.

Basically it looks like maybe this driver can output Dolby Vision for games over HDMI. Maaaaybe a custom player could output Dolby Vision video as well, but not sure about that. Lots of special Dolby stuff has to be implemented :)

aufkrawall
9th May 2020, 10:22
is there a different between intel windows driver and i965 Linux driver for 7gen igpu?
i965 is only an OpenGL driver in Mesa and is replaced by Iris. The kernel driver module is still named i915.
The kernel drivers of Intel and AMD support a lot of what is supported on Windows these days, though graphical user space in Linux world can't catch up on HDCP and HDR support due to lack of consensus, interest or developer resources.

VBB
27th May 2020, 18:36
I'll test in a bit to see if the bit-depth/GUI issue is fixed.

SamuriHL
28th May 2020, 03:28
Would be nice. I didn't get a chance to even look at it today.

VBB
28th May 2020, 05:32
I did, and it's not fixed...

But in the process, I found out something I did not know: Nvidia drivers dither by default. See my post over at AVS https://www.avsforum.com/forum/40-oled-technology-flat-panels-general/3138274-2020-lg-cxa-gx-dedicated-gaming-thread-consoles-pc-16.html#post59718362

bleakassassin
28th May 2020, 06:07
I feel like this is gonna be one of those things that doesn't get fixed for months upon months until it just suddenly gets fixed one odd update awhile from now. Unless it's somehow something that can happen in games, too. Nvidia isn't catering to madVR and HDR video users, after all.

Out of curiosity, is there a reason this isn't mentioned in the first post? I feel like it's something that people using one of the last few drivers should know.

VBB
28th May 2020, 07:28
Klaus, could you add the bit-depth/GUI issue to the last few Nvidia drivers in the first post, please? "Playback at greater than 8-bit drops back to 8-bit (and out of HDR) in Windowed Full-Screen mode when player GUI is triggered, regardless of bit-depth of the file. Can be circumvented by using Full-Screen Exclusive mode or by setting madVR to 8-bit to begin with."

Danke aus Long Beach ;)

NoTechi
28th May 2020, 07:49
New AMD driver Adrenalin 20.5.1 (https://www.amd.com/en/support/kb/release-notes/rn-rad-win-20-5-1) released.

NoTechi

Klaus1189
28th May 2020, 17:21
@VBB: What was the first version affected?

Greetings from Bavaria ;)

brazen1
28th May 2020, 18:10
Klaus, you might also want to add that every nVidia driver after 388.59 has a faulty updated HD Audio driver in the newer Windows driver installer packages. Turn on AVR. Right click Audio Icon in tray. Hover over 'speaker setup'. It should show 5.1, 7.1, (whatever you selected running the audio setup). Turn off AVR. It should show 'stereo'. Toggling the AVR on/off should automatically toggle Windows audio properties. This doesn't occur with newer drivers though. It sticks on whatever you last selected and isn't automated. You have to manually change it. This affects flash video in Windows and programs like Kodi player for certain audio like AAC.

After testing new drivers to see if the bug remains, I always unpack new driver packages, delete the included HD Audio driver and replace it with 388.59 HD Audio driver which brings back normalcy. Don't be fooled thinking Windows audio properties have nothing to do with passing audio duties to AVR via HDMI. This only affects folks who use systems with AVR in both on and off states and not just one or the other. Just sayin'...

el Filou
28th May 2020, 18:34
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/AA5emb3Well guess what? Looks like MS *did* implement it: "specialized displays" in Windows 20H1 https://www.tenforums.com/tutorials/154973-how-remove-display-desktop-windows-10-a.html
From the description it sounds very much like what I asked i.e. displays that can be connected and used by specific apps but don't interact with the desktop. However as this is MS we're talking about they decided to restrict the feature to Windows 10 for Workstations... :( so I can't even test if it would work for HDMI audio to an AVR then. Shame.

VBB
28th May 2020, 22:35
@VBB: What was the first version affected?

Greetings from Bavaria ;)

Let's go with the first 445 release, 445.75. Fairly certain that's when it started.

huhn
28th May 2020, 23:44
But in the process, I found out something I did not know: Nvidia drivers dither by default. See my post over at AVS https://www.avsforum.com/forum/40-oled-technology-flat-panels-general/3138274-2020-lg-cxa-gx-dedicated-gaming-thread-consoles-pc-16.html#post59718362

it's well known. AMD does it too.
it's so well known that they both dither that madshi added a workaround for a bug in the nvidia driver where it would not dither in the past. i'm pretty sure this workaround is still in madVR. in this case madVR doesn't allow sending 10 bit.

VBB
29th May 2020, 02:45
I had a feeling huhn would show up eventually ;) Anyway, it was news to me that Nvidia dithered by default. I knew about AMD.

So, this is probably a dumb question, but here goes: Do we want the driver to dither in addition to madVR? If so, why? What happens if only madVR dithers?

huhn
29th May 2020, 02:54
that's a confusing question.
you can't dithering 8 bit to 8 bit or 10 bit to 10 bit.
so dithering is applied when bit deep is "lowered" if the GPU has to do that for what ever reason.
yes we do want it.

VBB
29th May 2020, 02:59
I see you're up late. OK, so what happens if we disable dithering at the driver level? I noticed it removed some of the smoothing from my gradient tests, but it didn't introduce any new banding. Playback of normal content was noticeably sharper.

huhn
29th May 2020, 03:11
i tried the tool at 6 bit without dithering on the desktop it's "clearly" not applied on everything. so i don't know where it put's it self. does it work similar to reshade and only effact d3d programs?

nvidia was now tested to be bit perfect by madshi for envy development using capture cards.
or in short nothing should happen if no dithering is needs and banding should be the result if it is needed.

VBB
29th May 2020, 03:15
Right, so you would think, but in an all 10-bit chain, there is a very noticeable difference between Nvidia dithering on or off. This applies to playing back both 8- and 10-bit content. That's why I'm asking, because I was always under the impression that the video card doesn't dither if no dithering is needed. I noticed the difference in both RGB and YCbCr output, btw.

huhn
29th May 2020, 03:19
if you output YCbCr you need dithering.
if it is full range RGB you may not need it else you pretty much always need it.

and source bit deep is irrelevant.

chros
29th May 2020, 10:36
VBB uses ycbcr 4:2:2 10 bit output.
madvr 10 bit RGB -> ycbcr 4:2:2 10 bit : dithering is needed after every conversion step (in this case at nvidia driver)

@VBB: how did you test nvidia dithering? sample file?

Klaus1189
29th May 2020, 16:19
@VBB: added info to 445.xx ;)

@brazen1: added in Notes to Nvidia section, because too many versions affected ;)

I had cut shorten a bit, the post is to long, in the future I will have to split AMD and Nvidia into the first two posts, we'll see :)

mclingo
29th May 2020, 19:26
20.5.1 is looking good for me, i can now get FULL RGB with 1080p with HDR, for some reason everythign after 20.3.1 gave me raised blacks, this is the first fully working driver i've had since I bought my RX5700 !!! yippeee....

not tested anything else yet though ;)

VBB
29th May 2020, 21:08
Thanks, Klaus!

@chros - Yeah, after my little discussion here with huhn and also reading the first few introductory posts of the madVR thread again, I now understand better how it all works in theory. Having said that, without any driver dithering, what exactly should I be seeing? Increased banding due to reduced noise?

I tested this once again with my usual banding tests from both the AVS as well as the S&M discs.

huhn
29th May 2020, 21:49
if the tool is working in that eco system you should get 12 bit banding with 12 bit(should be close to invisible) or 8 bit with 8 bit output as an example.
it should be pretty much the same as madVR with disabled dithering.

with full range RGB nothing should change.

VBB
29th May 2020, 22:07
Thanks, huhn. I'll do some tests with full range RGB later. That tool makes it so easy to switch. The only thing missing is refresh rate.

kostik
29th May 2020, 22:49
Thanks, huhn. I'll do some tests with full range RGB later. That tool makes it so easy to switch. The only thing missing is refresh rate.

You can combine it with Display Changer (DC not DC2) which will change refresh rate and create one batch file for each mode

chros
29th May 2020, 22:58
DC2 is also fine, that's what I use to change resolution and refresh rate.

VBB
5th June 2020, 01:22
Some observations on Nvidia dithering after using the tool for a while now:

- The default Nvidia setting for dithering seems to be "auto".
- By default, there is no driver dithering with RGB, as expected, but it can be enabled.
- By default, there is less than 10-bit driver dithering with YUV, even when the output is set to greater than 8-bit.
- Dithering applies to everything, including the desktop. This is most obvious when forcing 6-bit dithering, as the noise then becomes highly visible.

If you enable dithering for RGB, you can get rid of any banding in the LG OLEDs' PC mode, even in HDR. The downside is an increase in noise level, which might not be tolerable for everyone. Certainly not for me, but it's fun to test, and it might be perfectly acceptable for gamers. After all this testing, I am back to my old YUV422, 10-bit, normal HDMI. It's the "Triple Dither™" that makes it look great after all (driver+madVR+display processing).

huhn
5th June 2020, 04:06
what did you do to get 6 bit working it's clearly not working here.

VBB
5th June 2020, 04:15
That tool I linked to earlier allows you to do that. I used nvcolorcontrol.exe 8 (this is the driver bit-depth) RGB (color space) 1 (this enables dithering) 0 (this is 6-bit) 4 (for temporal, but you can use whatever you like) 0 (for SDR), so nvcolorcontrol 8 RGB 1 0 4 0

huhn
5th June 2020, 04:49
ok it doesn't work with dithering disabled or mode 0.

temporal is horrendously bad.

VBB
5th June 2020, 06:01
Well, yeah, disabled turns dithering off altogether. Mode 0 is auto, which you can't be sure what it does exactly. 6-bit just proves that you can make even the worst banding smooth, and it'll make all the old plasma guys feel nostalgic ;)

What do you consider the best type of dithering, since you dislike temporal? With 10-bit and even 8-bit, I can't see the noise, and I sit less than 5 feet from my 65".

huhn
5th June 2020, 07:08
didn't check the rest in detail if they are not random dithering they should be pretty transparent i need to use dithering test images for this but it's hard for me to care if i can just avoid it.

i don't even know why they would add temporal if you can't use it for what it is made for.
i mean nvidia get's 60 hz in and outputs 60 hz so temporal can not be used as it should be...

maybe it's made for half frame rate vsync games.

SamuriHL
11th June 2020, 23:15
Ya know....I really wish Windows would stop F***ING with my drivers! I was content and happy on 442.74. Apparently Windows decided to upgrade my drivers for me. Holy God how can any of you stand the latest nVidia driver?!?! Seriously it broke everything for me. HDR was a complete mess while triggering HDR it was not outputting in the correct color space (even though it claimed to be bt.2020). Even SDR was a complete mess. No one else has seen this? I went back to 442.74 and I'm back to video bliss.

huhn
12th June 2020, 00:20
sorry didn't update in a long time and could be that the windows update driver is broken so everyone that upgrades normally doesn't notice.

SamuriHL
12th June 2020, 01:47
Yea I hadn't updated in a while either for the obvious reasons. I suppose it's entirely possible that WU broke the driver install. It was bad though. Freaked me right out. LOL Downgraded back to my normal driver and all is well so no big deal I guess. Just apparently no one else has run into it.

VBB
12th June 2020, 03:30
I haven't watched anything HDR in a while, but as you know, I've been using the latest drivers. No issues here other than the now well known one mentioned in the first post. What did you notice wrong with SDR? How did you figure HDR wasn't being output in the correct color space?

RXP
12th June 2020, 06:10
Does 442.74 passthrough meta data oK? I don't notice that mentioned on the first page anymore.

VBB
12th June 2020, 08:33
Yes, meta data has been passed through properly for a long time now.

SamuriHL
12th June 2020, 14:55
Yes 442.74 is the driver I'm sticking with until they fix the stupid issues introduced in the later drivers. It works perfectly for me.

webbo
13th June 2020, 08:59
Yes 442.74 is the driver I'm sticking with until they fix the stupid issues introduced in the later drivers. It works perfectly for me.

Hello SamuriHL

But which would be those issues - for example, tested 446.14 myself for a few movies / TV Shows, 4K HDR ones, no issues whatsoever

Bugs in colorspace, you mean?

Thanks for your input

SamuriHL
13th June 2020, 20:04
The issues that have been reported about if you're outputting 10 bit or higher, it'll pull you out of HDR if you bring up an overlay (like your player). It's just distracting. Not that I tend to bring up the overlay very often when I'm watching a movie, but, I have no reason to upgrade that machine unless I have a driver with no compromises to go to. For others, especially those using 8 bit output, it's a non-issue.

Calvi
14th June 2020, 04:55
I am running 442.92 Studio drivers and it seems as good as 442.74

SamuriHL
14th June 2020, 14:10
Maybe I'll give those a try.

Sent from my SM-G975U using Tapatalk

nsnhd
15th June 2020, 05:36
The issues that have been reported about if you're outputting 10 bit or higher, it'll pull you out of HDR if you bring up an overlay (like your player).
So with the 442.74, outputting 10-bit FSW stays when bringing up overlay ? I'm using the very old 385.28 for sticking to 10-bit in all modes but when overlay (seek bar, madVR controls...)comes up it changes to 8-bit shortly.