NVIDIA GeForce 516.59 WHQL driver download & Discussion

Wait so they’ve fixed the issue on 3000 series then. As my strix 3090 was doing it.

 

Ah, got a repro now, only occurs on subsequent runs of the game which means probably that the shader recompiler is a factor (infinite loop is my best guess without being able to pix/nsight)

 

This is a browser core issue in chromium, the new drivers have enabled MPO's on 10bit and now have 4 planes available instead of just 1,

Google and Microsoft are yet to introduce patches that resolve the issue, disabling MPO's for the time being is the known workaround.

Reference list: Issues - chromium

 

I can confirm that, as with 516.40, this driver also works fine with Jurassic World Evolution 2 as long as the game is patched to v1.4.3, which is the update that fixed the invisible geometry bug that affect v1.4.2 and earlier builds of the game. I know the Steam and Epic versions are on v1.4.3 but the Windows Store / GamePass version may not be as it wasn't when I checked it last week after the Steam / Epic versions had been updated.

 

wrong, I experienced it myself on my desktop 3080 ti and laptop 3070

 

Good stable driver. All games working great

Note:
This one resets cache.
No need to restart.

 

Just tested 4K, 4K60FPS, 8K and 8K60FPS and all of them run fine on YouTube.

Do what Astyanax recommended below in his comment. Disable MPO.

You only need to do this once; you don't need to do it after every driver install or uninstall.

Open notepad, copy and paste the code below into it, save it to desktop, then change it from .txt to .reg and run it.

Code:

Windows Registry Editor Version 5.00


[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\Dwm]
"OverlayTestMode"=dword:00000005

If you don't know how to change the extension at the end, use the search bar in the start menu, type "File Explorer Options", open it, go to the "view" menu at the top, and untick the below:

I done this over a year ago and haven't had YouTube issues since. I always use the latest drivers too.

 

mpv/madVR stutter with MPO is still not fixed with this driver, I reported this in November 2021.
I'll try recording a video, showing how the render statistics stutter count increases vs. no stutter with MPO off...

 

Driver seems fine in games on my end

I did notice lower performance in timespy, doesnt seem to hold the same clocks as previous drivers there and core clock bounces 100mhz or more not the usual 15mhz steps

Games seem unaffected by this

 

So destiny2 issue is confirmed?

Will see again today when i get home from work, got new 2x16gb ram now maybe it will act different.

 

its not up to nvidia to fix this, applications need to correct their behavior for compliance with multiplane overlays, madvr included.

 

Interesting, never had this issue myself with MPV, have been using it for years, almost daily. What is your configuration like? What GPU do you use?

Here are my MPV configurations (mpv.conf), you can try those same shaders (they have to be downloaded) for increased quality. But you need a decent GPU, at the very least I know this configuration works on an RTX 2070 Super and up (nvidia only for some configurations here).

Anyway, what will likely fix your dropped frames issue is if you switch to full screen exclusive mode, all you need for that is the first line in this code, the second line opens videos in fullscreen straight away if you want that.

If that doesn't work, use everything in the code except the last 3 lines which are for shaders. Actually, you can remove a lot of it which is just custom stuff for me, I already removed a bunch. But if it works and there's no stutter for you anymore, let me know and I will edit the configuration below to clean it up for you or others if they are interested.

Code:

ontop
fullscreen=yes
priority=high
osd-level=1

osd-bar-align-x=0
osd-bar-align-y=-1
osd-bar-w=25

volume=100
volume-max=100

vo=gpu
profile=gpu-hq
fbo-format=rgba16hf
gpu-api=vulkan
gpu-context=winvk

hwdec=nvdec-copy

linear-downscaling=no

spirv-compiler=auto
video-sync=display-resample
vulkan-swap-mode=fifo

tone-mapping=hable
tone-mapping-max-boost=1.0
hdr-compute-peak=yes

scale=ewa_lanczossharp
cscale=ewa_lanczossharp
dscale=ewa_lanczossharp
dither-depth=auto
sigmoid-upscaling=yes
sigmoid-upscaling

glsl-shader="C:\Users\YourUsernameHere\AppData\Roaming\mpv\Shaders\KrigBilateral.glsl"
glsl-shader="C:\Users\YourUsernameHere\AppData\Roaming\mpv\Shaders\FSRCNNX_x2_8-0-4-1.glsl"
glsl-shader="C:\Users\YourUsernameHere\AppData\Roaming\mpv\Shaders\SSimDownscaler.glsl"

 

Where do Microsoft document this and clearly advise devs what exactly to do about it? Because if they don't, I doubt this is true.

Do you happen to know if new GPU upscaling resolutions still disable MPO?

Most simple config:
hwdec=d3d11va
video-sync=display-resample
interpolation

When I play 720p 59.94fps video in 1440p fullscreen, it reliably accumulates delayed frame in statistics (shift + i) with visible stutter. It's 100% gone with MPO disabled, I've been able to reliably reproduce this for over a year now (ruled out anything else, including power saving features etc.).


Btw: At least with some older drivers, apparently MPO wasn't always activated reliably also with 8 bit and upscaling resolutions disabled with at least 4k 144Hz HDMI 2.1. At least Amazon Prime app sometimes couldn't make use of VRR for no obvious reason, which indicates that MPO then wasn't active. And this would also explain why some users aren't able to reproduce.

What a giant mess MPO is on Windows...

 

Was the stuttering ever fixed? It used to stutter for me all the time, thing it was a shader compilation thing.

Does that just go away with time?

 

Microsoft doesn't publicly document all aspects of it unfortunately, so without a microsoft liason to the DWM/Compositor team, it makes it difficult to identify what changes are necessary, this is one of the reasons Chromium has had such a long standing issue with the feature,

the two options are making your application completely immune to MPO promotion, or making it work perfectly with it, making it work perfectly means nailing down every possible case that could cause the overlay dropping out of hardware overlay back to composited.

These are things like reporting the letterbox dimensions when presenting to within 1px of precision - as if you tell the hardware you're going to draw at 1920x800 but then something in the undrawn regions is updated, you drop from the hardware overlay, like wise you also need to implement seperation of canvas and content, an issue Google is having with chrome and video players especially those that render from an iframe container.

You also need to be sure you use a format compatible with the hardware overlay, not all hardware supports NV12 (Intel especially), which is YUV_420_BIPLANAR

The basics aare here https://docs.microsoft.com/en-us/wi...play/multiplane-overlay-hardware-requirements

AFAIK, the problem madvr has is its used in conjunction with software that isn't mpo compatible, some of the apps draw gdi glyphs on top of madvr rather than through it.

It should, as it manipulates the hardware plane.

 

As stated in my pervious post, I can confirm this. I am experiencing the exact same issue.
It appears that not everybody is affected by this though.

 

I have same problem with Division 2 on GTX 1080 Ti, and it happen offen

 

MPO stutter recorded on video:


No stutter with workaround MPO off:


I've also tested it on my Gemini Lake notebook with Intel IGP, and there I can play that video without any tweaks or workarounds without any stutter. No idea if it enables MPO, but it underlines how wrong situation is that well-coded video players don't work as expected without tweaks/workarounds/limited functionality on expensive Ampere GPUs.

Interestingly, the OpenGL renderer of mpv with WGL window context backend is not affected. But everything else (D3D11, Vulkan) is.

It would be really nice to have a statement from Nvidia about the situation. I'd gladly report it to the mpv devs, but it would be only fair to provide them with any information possible, in order to hopefully improve the situation.

 

Which monitor or TV model do you have? Perhaps it's specific to that model (could be a driver incompatibility with it). Does it support 23/24Hz at 3840x2160 natively or did you add it via Nvidia Control Panel, CRU, etc.?
Do you have VRR/G-Sync enabled (assuming your monitor supports it)? If so, try to disable it and see if that helps.
Did you try a different cable?
Have you filed a full, detailed bug report with Nvidia yet?

 

Div2 has had random crashes since launch. I highly doubt it will ever be fixed at this point