NVIDIA Hotfix GeForce 496.61 WHQL driver download & Discussion

I did not say that the cause of crash is low memory capacity. What I've meant is that IW8 engine eats extreme amounts of memory, regardless of how much RAM or VRAM you have.


P.S. ... and I'm not talking about a specific DEV ERROR, of which there are many, mind you.

 

I'm the only here that said Warzone@4K/High textures requires more than 12GB of VRAM since i'm getting stutters with a 3080Ti after extended period of gameplay. Normal textures are fine no matter how many hours i play. I'm curious about the upcoming new map though..

 

My machine is more than perfect with all drivers before 496 crap. But hey, it's my problem that i pay double price for ironic attitude, thanks.

 

and majority of others have no issues with 496, believe it or not but the people that do have issues are a minority no matter how loud they are or many alts they create to repost issues to make them seem bigger.

nothing gets fixed with forum posts though, also i think @ManuelG needs to start requesting msinfo32 text files attached to the survey because sometimes it really is due to third party compatibility issues.

 

Hello
I'm having issues with 496 series drivers

out of two main issues, I have only found a work around for one.

Issue 1
496 displays polygon artifacts in 2D desktop where small black squares appear on 2D desktop surrounding app windows.
Issue is 100% resolved when driver is installed without MS store/control panel automatically and control panel is installed manually from extracting the appx (7z) and placed in the correct directory, with ms store off and appx service disabled.

Reverting to 472.12 resolves this issue.

Issue 2
A. Performance drop compared to 472 in mgpu app games or mgpu setup. Examples include SOTTR, ROTTR, Sniper 4, Quake RTX, and a few others.
Single GPU games that are affected to a lesser degree are Cyberpunk, Agony, Succubus, Metro Exodus, Ghostrunner and Necromunda. all games running DRM free, no interference from overlays or 3rd party.
I use max eye candy in settings at 4k, 10-bit, no vignette, no film grain, no motion blur, no depth of field and no DLSS.
I don't have any apps running in background, no chrome derivatives, no ms store apps, and no 3r party such as steam, EA, etc.
(Dual 3090 RTX FE and Asus PA32UCG)

Benchmark drop in Gravity-Mark from 71,000 average, to 64,000 average in Vulkan. Also showing a 67,000 average tp 63,000 average in DX12 performance.

https://gravitymark.tellusim.com/leaderboard/


B. I'm experiencing a new micro delay between opening and closing 3D apps or games (3D medical software) and games (Vulkan DX12, DX11, OpenGL) -where momentarily (1-2 second) the screen goes completely black, occasionally lasting as long as 2 seconds, but normally slightly less than 1 second.

Reverting to 472.39 retains the micro delay and games / benchmarks have a slight drop in performance compared to 472.12. 472.39 solves the 2D artifacts - this is probably due to standard driver implementation as it doesn't use the ms store.

Reverting to standard 472.12 resolves all issues. Benchmark and game performance is returned to normal, screen artifacts are gone.

Peak performance for drivers and most stable was standard 471.41 driver (so far).

 

What the heck?

Can you turn on the Handle and GDI object fields in task manager and sort by them?

 

went through that, captured the list and did performance counters, and sent ticket in to nvidia with all the prerequisite info, as well.

Some additional info to help troubleshoot:

I usually do not run the nvidia container service. I make most changes that I need with profile inspector, However when I do need the control panel, I have a svc start associated with the panel shortcut - otherwise service is set to manual. During the testing I left the nvidia service on as automatic to keep everything generic and took it as a default install.

So now I have two worlds to pick /live from for stability: Run the standard 471/472 driver or run the DCH 496 with the manual install of control panel.
If I let the dch driver do a normal install it all works but I do get the artifacting. I think I posted a screen shot and video of the effect in the prior 496 thread section.


I realize that I am not the 'majority user' because I have to use server for the OS -- I get that and understand. However, the medical software I use has a database license that only works on a server OS. With that said, I've had to use server since 2006 and it has been with some compromises. Also, I have to use Ubuntu for the other part of the software and DL/ML, so I spend a lot of time in dual boot. So my PC generally has the latest server edition and the latest ubuntu LTSC. The medical software vendor did not get their stuff compatible to windows 10 till July of this year (just in time for windows 11, and if one needs the administrative console of the software then yer only able to run it on a server license).

With that said, I did do a windows 11 boot, and the artifacting does go away. The driver install went smooth and easy. However I noticed the performance drop is still there and the micro delay/blank screen still does occur on Dp1.4 and HDMI 2.1. However, that was just an experiment and had to return to server.

So some issues are relevant to the dch component and hopefully that will get solved. The micro blank between apps could be driver or possibly a port firmware fix in combination with the driver, - I don't know, just speculating at that point. I had a 3080 ti and did the firmware update and it seems to have solved the issue through the DeviceID fix using the same model monitor. However the firmware DeviceID tool did nothing for the 3090 and reported that the fix "was already applied" (This component doesn't have as much to do with this driver, but I was hoping it would help with the blanking)


Work in progress I guess. Regardless of artifact and screen blanking - I have noticed a drop in performance on apps that invoke AFR or MGPU. While I realize that mgpu is not as popular, however it also seems to affect some single gpu titles. I really don't understand the drop in performance between standard 471/472 and the 496 series. I'm trying to see if its something on my end, however a driver reversion seems to solve it every time - so my deductive reasoning keeps looking for the change to the new driver.

Seems greater stability and performance was around the standard 471.41 iteration of driver. I'm putting a little more attention in this because my hope is that the new series of drivers do not perpetuate the issues I'm experiencing in the 496.

Grazie

 

The 457.51's are such amazing drivers compared to the latest ones, they perform so much better in Warzone with higher minimum's and average's in the frame rates, no crashes AND the Sharpen Filter's work perfectly well with the Nvidia control panel (not black listed) so you don't need Geforce Experience.

 

@Astyanax do you have some fix or workaround for FM7 Graphic glitches in the Hud and Car view that was happening a few month ago in a Driver release?
The Bug is still not fixed and i reported already a few times but sems nvidia have all hands full ^^

Also Pascal and Maxwell is not affected only Turing and Ampere Cards

 

I wanted to verify this is a real thing, when I googled the issue it brought me to your post.

Straight after updating graphics driver with geforce experience. All apps using GFX cards micro lagged when switching between them. Maybe driver crashing since screen kept flickering? And this appeared.




I'm gonna stick with 471.41, thx for the info.

 

is gpu scaling enabled in the sharpening control?

 

The issue venturi is talking about is presumably not the one in your picture.

his was solid black squares, yours have actual artifacts in them

 

The issue you're reporting is different than the polygon issues. The polygons on the issue I reported are ripples of black squares in circumference to some windowed applications on 2D desktop. The polygons I reported are sensitive to the position of open windows, the artifacts you're reporting are populating regardless of content on the screen.. While not pretty, the issue you're reporting is different in nature.

 

And by the way this driver is an awful slow download.

 

Can you capture a dmp file as we have not been able to reproduce any crashes on WRC 10:

https://nvidia.custhelp.com/app/answers/detail/a_id/3335

Send it to driverfeedback@nvidia.com

 

We are still looking into this issue

 

@ManuelG
https://forums.guru3d.com/threads/n...ownload-discussion.440233/page-3#post-5959188
Sorry to disturb you, can you reproduce and fix this bug in the latest driver? thank!
I have been using the latest driver for nearly twelve years. If this is a rare bug, I cannot use the latest driver if you don’t fix it.

 

On older motherboard chipsets, MSI enablement would introduce stability issues but on newer chipsets, it would probably be safe to use.

 

Why not have the installer check the chipset and if newer than XXX enable MSI? Or if it's a hassle to string match a whole bunch of chipset versions, just check for a BIOS date or something and go by year

Edit: And how old is old? I'm running a Sandybridge CPU with Intel X79 chipset, getting long in the tooth but still running like a dream! And what do 'stability' issues look like? I've enabled MSI via the registry tool and I don't -think- I've seen any issues, but are we talking game crashes, or frametime pacing etc?

Edit: Another fun fact, my Sandybridge i7-3970X was the CPU range with the 'not quite properly implemented' PCI-3.0 spec from Intel, and indeed enabling it via the Nvidia PCI-3.0 reg tool (thank you for this all these years) used to cause crashes with my 980Ti cards in games. However since upgrading to 1080Tis (two in SLI -- still working wonderfully for many games) all those years ago, I've been able to toggle PCI-3.0 with rock solid stability since. But I had always wondered, if the issue was Intel not getting the timings right with its PCI-3.0 implementation, why did upgrading to a new Nvidia GPU work? Was the 1080 just more tolerant of signalling on the bus? Anyway, just a bit of what is now ancient history in tech

 

How is the NVRLA.EXE High CPU Usage bug still happening? Even when GFE is removed why is it still running? .. Look at this disaster.. I dont use GFE so i had no qualms in removing it, but i had to rename this file again to make it stop running amok.