MSI Afterburner question - Understand my GPU

[LightBringer]

I have had constant long running problems with 'display driver stopped responding and was successfully recovered' forcing mostly newer games to crash. Its been a long running issue for me and I have exhausted every online solution in the book. Frankly by now I've almost given up on it possibly being repairable without new parts (even after already replacing my RAM and PSU). Im trying to understand my GPU better to diagnose if it is truly the culprit. The card itself failing either by wear and use or whatever else.

I ran MSI Afterburner while I benchmarked Cyberpunk, just because its a good test on my PC. By good I mean it will typically always crash within a minute. However, the afterburner readings looked a little wrong to me.... Is this suppose to look this way? https://i.imgur.com/v5nFW8Z.png

The meter on the left significantly surpasses the base and boosted rates put on the GPU. The reason I have them downclocked to the max has to do with it making my games *slightly* more stable and less prone to crash. I started at just -100 and have since just dropped it all the way. But should the meter be passing my Base and Boost rate like this? Is this a significant sign of something or is this usual?

[Set-115689]

Did you flash the card with some sort of weird bios like a mining one maybe?

Is the card new or old to you? Meaning fake card?

A fresh windows install make any difference? Reset the nvdia control panel?

Temperatures ok at normal clock speeds?

Any errors in windows event viewer or view reliability report?

[LightBringer]

โพสต์ดั้งเดิมโดย Set-115689:

Did you flash the card with some sort of weird bios like a mining one maybe?

Is the card new or old to you? Meaning fake card?

A fresh windows install make any difference? Reset the nvdia control panel?

Temperatures ok at normal clock speeds?

Any errors in windows event viewer or view reliability report?

No weird flashing or any of that, purely just for gaming and socializing. Bought it off Newegg back in 2017, it was a Zotac 1080Ti. It didnt have problems until maybe the beginning of 2020 when I moved to a new house. Ive done a windows reinstall atleast 3 times and DDU+driver reinstall atleast 10 times (lol). Temps are all good on both the CPU and GPU, never saw it get above maybe the low 70s.

Windows Event Viewer is where all the meat truly lies though. A standard crash looks like as follows:
~80 errors all stating:
"The description for Event ID 13 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer."

Followed by a single warning after:
"Display driver nvlddmkm stopped responding and has successfully recovered."


I really appreciate you commenting on the matter, regardless if theres a follow up response or not. Seriously, anyone coming to this forum topic to help are actual saints.

[LightBringer]

For some people it seems so, which is good. Unfortunately for myself, even after underclocking it as far as I could it aided me little to any. I appreciate you pitching in of course though!

[Carlsberg]

Downclocking the card may not be helping. Those cards use gpu boost 3, the gpu can dynamically boost itself if conditions allow. They do so by thermally adjusting clock speed and voltage depending on temperature and demand. Possibly the card is overclocking itself too much, try a more aggressive fan curve to control it or maybe there is some way to limit boost clock.

Maybe capping frame rate at a lower rate will reduce the load on the gpu and therefore decrease the amount of overclock it will apply ? (just a thought/suggestion)

[GOD RAYS ON ULTRA™]

Try enabling debug mode in nvidia cp and run the card. Top left> Help> debug mode. It will return your nvidia gpu to a reference clock. Zotac probably comes with a questionable factory overclock beyond nvidia's values. Using debug mode will determine if the zotac overclock imposed from them is stable or not.
The only issue is debug mode setting doesn't stick, so you need to enable it again every time you boot.

[LightBringer]

โพสต์ดั้งเดิมโดย Hardware Hero™:

Try enabling debug mode in nvidia cp and run the card. Top left> Help> debug mode. It will return your nvidia gpu to a reference clock. Zotac probably comes with a questionable factory overclock beyond nvidia's values. Using debug mode will determine if the zotac overclock imposed from them is stable or not.
The only issue is debug mode setting doesn't stick, so you need to enable it again every time you boot.

I saw a recommended solution saying to use debug mode but I couldnt find it. So many thanks for telling me where that was at! Though unfortunately, nothing seems improved (I made sure to kill msi afterburner when running debug). CP2077 has been my game of choice to test just because it often crashes within 30 seconds. I experienced the same visual bugs and eventual crash as before. I do thank you for the help though!

โพสต์ดั้งเดิมโดย Carlsberg:

Downclocking the card may not be helping. Those cards use gpu boost 3, the gpu can dynamically boost itself if conditions allow. They do so by thermally adjusting clock speed and voltage depending on temperature and demand. Possibly the card is overclocking itself too much, try a more aggressive fan curve to control it or maybe there is some way to limit boost clock.

Maybe capping frame rate at a lower rate will reduce the load on the gpu and therefore decrease the amount of overclock it will apply ? (just a thought/suggestion)

I play all my games with Vsync enabled on a 4k 60hz monitor. I have the display set to 1440p however just because it helps with playing games borderless and at better frames, since not all releases are fairly optimized for 4k 60.

[GOD RAYS ON ULTRA™]

โพสต์ดั้งเดิมโดย LightBringer:

CP2077 has been my game of choice to test just because it often crashes within 30 seconds. I experienced the same visual bugs and eventual crash as before. I do thank you for the help though!

And you are sure it's the graphic card doing it? What kind of psu do you have? Do you get the same symptoms if you cap cyberpunk to 30fps?

[LightBringer]

โพสต์ดั้งเดิมโดย Hardware Hero™:

โพสต์ดั้งเดิมโดย LightBringer:

CP2077 has been my game of choice to test just because it often crashes within 30 seconds. I experienced the same visual bugs and eventual crash as before. I do thank you for the help though!

And you are sure it's the graphic card doing it? What kind of psu do you have? Do you get the same symptoms if you cap cyberpunk to 30fps?

I recently went from an 850W to a 1000W. Replaced it maybe a month ago in attempts to fix everything. So im fairly certain it cant be my PSU. I tried capping CP2077 to 30 and got the same crash and visual bugs. Whats weird is this is the ONLY game where I have visual issues like this. All other games have look just fine.
https://imgur.com/a/EE77FhS

[GOD RAYS ON ULTRA™]

โพสต์ดั้งเดิมโดย LightBringer:

Whats weird is this is the ONLY game where I have visual issues like this. All other games have look just fine.

I understand. The thing is cyberpunk plays perfectly on my pc with a gtx 1070, so I find it odd that you are having a graphical issue. In a case like this you will have to share your full pc specs through the cpu-z app. Submit a validation and copy/paste the link here.

[LightBringer]

Through CPU-Z? I know my specs enough to list then here if that helps

CPU: i7 7700k
GPU: ZOTAC GeForce GTX 1080 Ti AMP Extreme Core
Mobo:MSI Z270-A PRO LGA 1151 Intel Z270
Ram: Corsair Vengeance RGB PRO 16GB (2x8GB) DDR4 3200MHz
PSU: EVGA SuperNOVA 1000 G5, 80 Plus Gold 1000W, Fully Modular
Samsung 500GB m.2 and 4TB HDD

[Bad 💀 Motha]

What version & build of Win10 are you using?
Also which version of NVIDIA GPU Driver?

Since you moved and all that, first thing to do after moving a desktop in such a fashion, is to open it up and fully remove the GPU and re-seat it all again. Also check everywhere inside the PC tower, especially CPU cooler, GPU fan and heat fins areas, and your PSU... for any dust and blow all of that out of the system.

For GPUs such as yours, it's best to stick to NVIDIA Driver 461.xx or older.

[Carlsberg]

โพสต์ดั้งเดิมโดย Bad 💀 Motha:

What version & build of Win10 are you using?
Also which version of NVIDIA GPU Driver?

For GPUs such as yours, it's best to stick to NVIDIA Driver 461.xx or older.

Did consider the same, especially as there's no benefit for 1080 in later drivers and many users have reported issues with 465.89 which OP is using. Decided not to suggest because the issues seem to predate the driver updates.

[Bad 💀 Motha]

Yea for GTX 10 and older; you can pretty much just stop updating GPU Drivers beyond 461.xx for now, unless we reach a point where Win10 or Games require something newer.

Drivers newer then 461.xx at this present time are not stable for most non RTX GPUs

So if you use and rely on NVIDIA GFE, then set its Update setting to "Check for Updates Only"

[plat]

Do other monitoring software display the same thing? Like HWiNFO64 and/or GPU-Z? Did OP try an older version of Afterburner to see if it displayed the same things?

You can google this and come up with a number of threads regarding Afterburner's sometimes questionable clocks readings. Toms Hardware, LinusTechTips, HWiNFO all have threads about this.

I would reset everything to default and monitor for crashes, if not done already. I understand about Pascal cards and latest NV drivers. I run the latest 465.89 on my GTX 1080 just fine but that doesn't mean it should work for everyone else. For stability's sake you might have to go back to older drivers--hopefully it's not something inherently wrong with the gpu, just a setting that's off.

[Bad 💀 Motha]

In MSI Afterburner, to the right of the "gear" icon is a "circled arrow" or a "back arrow"
Click that circled arrow to restore the Clocks and all back to defaults. It gets this "default" from the GPU Firmware itself, not from anything in the application. So that default will always be correct per what your GPU Firmware has for default settings within it.