AMD, Intel and Nvidia driver issues and last recommended version

[mclingo]

dont feed the troll

[Plutotype]

Hi folks,
Thanks for the PDF, but reading it and the last 10 pages of this thread I could not figure out what is the last recommended Nvidia, AMD and Intel driver version ( madvr / JRVR user here ) and for which GPU/OS.
Thanks
Pluto

[Klaus1189]

Latent Nvidia is fine. AMD has Problems described in Post 1, other than that nothing to worry now.

Back when I startet the thread both Nvidia and AMD hat lots of problems.

[chros]

Take a look.at my signature what I use on Win10 1809.

[StainlessS]

Is there perhaps an error in OP PDF version number for 472.47 (there are two of them).


And from here [Official Advanced Driver Search | NVidia]:- https://www.nvidia.com/Download/Find.aspx?lang=en-uk#


472.84 from 2nd image seems to be latest standard WHQL Studio Driver, and given same date as the latest 472.47 in pdf, ie 13 Dec 2021.

EDIT: To Below, thanks

[Klaus1189]

You are right, I correct it soon, some problems now.

[GTPVHD]

https://blogs.windows.com/windowsexp...1-2022-update/
https://www.microsoft.com/software-download/windows11

Windows 11 2022 Update Version 22H2 released, WDDM 3.1 and Shader Model 6.7.

[Balling]

22H2 also has support for HDR calibration with ICC profiles, finally! Color.org people are so slow.

WHY IS THE APP CALIBRATION ONLY min 0.005 NITS!??? Useless, my TV is 0.000006! Yes, PQ 65 is also below 0.00001 and 64 is black body 0 nits perfectly! Crazy!

[SamuriHL]

Only until Arc dies off.

[cesar73]

Recommendation needed for Nvidia firmware version without madvr HDR not switching off issue for geforce RTX 3060ti.

My HTPC configuration: Windows 10 LTSC 21H2, Kodi 17.7 dsplayer, madvr 0.92.17, lavfilter 0.76.1.25.
and LG oled G1 TV
Madvr settings: hdr passthrough, windowed mode ( had display issue with full screen exclusive)

I tried randomly different versions of nvidia firmwares compatible with rtx3060 ti cards ( with clean ddu driver deletion in between): 462.65 , 512.96 , and latest 516.94

Always same hdr issue: hdr activated correctly when video starts but remained enabled when video stops making kodi gui color saturated.

Any help will be appreciated

[brazen1]

Originally Posted by cesar73

Recommendation needed for Nvidia firmware version without madvr HDR not switching off issue for geforce RTX 3060ti.

My HTPC configuration: Windows 10 LTSC 21H2, Kodi 17.7 dsplayer, madvr 0.92.17, lavfilter 0.76.1.25.
and LG oled G1 TV
Madvr settings: hdr passthrough, windowed mode ( had display issue with full screen exclusive)

I tried randomly different versions of nvidia firmwares compatible with rtx3060 ti cards ( with clean ddu driver deletion in between): 462.65 , 512.96 , and latest 516.94

Always same hdr issue: hdr activated correctly when video starts but remained enabled when video stops making kodi gui color saturated.

Any help will be appreciated

Using Nvidia 517.48 with official Kodi v20 passing through madVR HDR mode engages when MPC-BE video is started and disengages when video has stopped.

Fwiw, I use the official Kodi internal VideoPlayer for HDR because there is no need for madVR when passing through. HDR mode engages and disengages using Windows HDR and/or Nvidia HDR and video is identical because it's passed through.

Besides, official Kodi handles iso's complete with menus. 17.7 DSP is getting pretty long-in-the-tooth. Expect problems imo.

[jayer]

I'm trying to figure out if any of the Nvidia drivers will work with my combination of Windows 10 21H1 and an RTX 3080 (12GB version) without breaking HDR? So far, every driver I've tried presents the 442.92 bug where SDR videos opened after playing an HDR video get played as HDR too. Makes me miss my old RTX 2080 (well, not really).

If non of the approved drivers for my 3080(12GB) can correctly handle the HDR/SDR switching, could I modify the installation file for one of the older drivers, like 442.74? Or would I likely cause even worse problems with my graphics card by trying to force it to use an older driver that isn't officially supported for my hardware?

Thanks in advance for any help/suggestions!

[nevcairiel]

This sounds like the good old madVR bug, not really a driver failure.

[Klaus1189]

Are the RX 5000 issues still present? Also do RX 6000 cards also still have these? I am referring to the ones described in first post.

[mclingo]

I have an RX5700 and have no issues with it at the moment at all, everything works as it should using LAV filters, EVR renderer, MPC renderer, MADVR , HDR SDR etc, however I dont run my TV is PC mode as I use TRUMOTION, this option is greyed out in PC, when i've tried PC mode in the past the colour space is all wrong but I havent tried it for ages, I doubt it will be fixed.

on latest driver, win10

[QBhd]

Same here... 5700XT, no PC mode on the TV (Banding is the main issue for me with PC mode)... everything works amazing

QB

[huhn]

you listed issues are still present.

there are even new ones very very odd ones.
if you take mpc-vr and set it to more than 8 bit internal bit deep the d3d11 and DXVA2 processing falls apart except if you downscale a lot.

this is the driver thread btw. so the pc mode part is well just that.
but nice to see that user here on this forum are now lying into other people faces.

i can only speak for 5000 series issue the 6000 may actually be perfectly fine just like the polaris series.

[Klaus1189]

So is it possible that the issues are also present in new RX 7000 series? All RDNA had these issues.
I just tested my old heater RX480 a few days ago and deinterlacing works, also frame adaptive, but if I watch a interlaced DVD I get some portions of the image where I see interlacing artefacts. These are strage roundly shaped areas. So not good deinterlacer and also bugged. There was or still is a screenshot in this thread, but I must look for it. I don‘t want to put the shitty Polaris back in just for a sreenshot.

I still do not know if those are hardware issues or „only“ software issues.

Also the Intel issue mentioned in post 1 … still present in new Intel driver?

I have also to mention that the Intel versions are driving me crazy! I don’t have an Intel CPU with iGPU here, just an old HEDT 4930k, so if anybody wants to test it, please let me know.

About the versions of Intel drivers, let me know if I should do anything different, I don‘t know any other way, but I don‘t like it that way.

[mclingo]

Originally Posted by huhn

you listed issues are still present.

there are even new ones very very odd ones.
if you take mpc-vr and set it to more than 8 bit internal bit deep the d3d11 and DXVA2 processing falls apart except if you downscale a lot.

this is the driver thread btw. so the pc mode part is well just that.
but nice to see that user here on this forum are now lying into other people faces.

i can only speak for 5000 series issue the 6000 may actually be perfectly fine just like the polaris series.

whos lying, about what?

[huhn]

Originally Posted by Klaus1189

So is it possible that the issues are also present in new RX 7000 series? All RDNA had these issues.
I just tested my old heater RX480 a few days ago and deinterlacing works, also frame adaptive, but if I watch a interlaced DVD I get some portions of the image where I see interlacing artefacts. These are strage roundly shaped areas. So not good deinterlacer and also bugged. There was or still is a screenshot in this thread, but I must look for it. I don‘t want to put the shitty Polaris back in just for a sreenshot.

I still do not know if those are hardware issues or „only“ software issues.

you can easily assume a software issue the driver was always able to do bob or just weave(do nothing) and they simple don't care to change that. it was an option back in the CCC days.

the current state deint on win 11 22h2 with amd 22.10.1 5700 xt is:
madVR weave sometimes simple bob
mpc-hc EVR CP weave
mpc-hc EVR weave
mpcvr weave but different with massive ringing (i will retest with a current version)
windows media player bob
windows movies app bob
mpc-be EVR "working" atleast sometimes or it does this: https://abload.de/img/mpc-benhfej.png 16 bit format float.
with 10 bit int it works on the first playback doesn't artifact.

so the deintlacer is still available somewhere.

the level issue is now a different beast. it's now full range all the time. even when set to limited range only tested in mpc-be. so still broken but in a far more acceptable state.

it's kinda odd to see the "best" realtime deinterlacer work after all these years...

BTW. nvidia is also generally buggy with all DXVA and D3D11 processing with RGB output.
they move the chroma to the wrong position it's half or a quarter a subpixel to far to the right. like they move the image by half a subpixel before scaling instead of after scaler.
this bug is very old and very hard to explain to nvidia most user will never see an issue in it.

only tested on a 1060.