Games crash if I do not Underclock. Why?

[Drogin2]

Ladies and Gents,

I've been having problems with my PC for over some time now. Didn't care as much before but with Dragon's Dogma coming out recently - I wanted to optimize my experience with the game.

I need some of your wisdom people.

What I've noticed: New games or the ones in Alpha/Beta states - are more prone to crashing for me. Like Palworld or Cyberpunk when it came out.
GAMES CRASH, NOT THE PC ITSELF.

The game does not have to be demanding on my PC to crash. Elden Ring was working fine on high/ultra.
Decreasing Graphics settings does not change much. Games on Low will still crash if I do not underclock. Ex. When I play Cyberpunk 2077 on Ultra, the game runs perfectly 50-70 FPS but then suddenly just turns off. If I change to low, I will gain some FPS but the game will still crash.

My PC:

OS Name Microsoft Windows 10 Pro System Type x64-based PC
GPU NVIDIA GeForce GTX 1080 8GB
Processor AMD Ryzen 7 1700 Eight-Core Processor, 3000 Mhz, 8 Core(s), 16 Logical Processor(s)
BaseBoard Product ROG STRIX B350-F GAMING
BaseBoard Manufacturer ASUSTeK COMPUTER INC.
Installed Physical Memory (RAM) 16.0 GB
Total Physical Memory 15.9 GB
Available Physical Memory 8.05 GB
Total Virtual Memory 27.9 GB
Available Virtual Memory 15.8 GB

Never had issues with overheating or anything like that. Benchmarks were always doing great. Please let me know what can I do about it.

P.S. I've been playing around with Afterburner to see the best settings.
What I've learned:
- Games will crash randomly with GPU at 65C or 77C.
- Decreasing settings to Low does not change the outcome.
- No Crashes with Core Clock -58 and Memory Clock -52
- Way less Crashes with Core Clock -17 and Memory clock -20

OC Scanner Report:
21:21:30 Start scanning, please wait a few minutes
21:44:35 Scan succeeded, average core overclock is 62MHz, memory overclock is 200MHz
21:44:35 Dominant limiter
21:44:35 No load
21:44:35 Results are considered unstable
21:44:35 Overclocked curve exported to MSI Afterburner

 

[The King]

You seemed to have left out an important part which is the PSU.

If you have to under clock the GPU core clocks it can be the GPU is unstable if it is an OC edition or your PSU maybe having high ripple or something similar.

GPU temp is one thing and GPU VRAM temp is something that never gets reported in most cases.

Have you check for WHEA errors ?

On another note I have a GTX 1070 and have a profile where I run the GPU -50 on the clocks with 80% power limit it uses far less power
and you hardly notice any fsp loss in games. Well in my setup that is.

Also you can use HWINFO64 to monitor GPU information not available In Msi afterburner.

Some cases doing a clean install of Nvidia driver can solve crashing issues has well. Don't forget to select custom clean install option.
I would try this first if you have not done so already.

 

[Drogin2]

Thank you for your answer. I do not remember my PSU. Will need to open the the PC tomorrow and check.

There do not seem to be any WHEA errors + I just did a clean reinstal of drivers yesterday.

Those are my best setups for Afterburner.

More details on the left from the gaming sesh.

CPUS temp around 40C. GPU -65-72C

 

[GerKNG]

from my experience pascal can degrade a bit over the years. i have a 1080 Ti Strix OC which degraded to a point where i almost need to go 30Mhz down.
Check in HWInfo your 12, 5 and 3.3V Rails while gaming and if a DDU clean install and fresh install of the latest driver does not help it might actually be just as mine.

 

[basco]

your settings in msi afterburner are all messed up.
1st pic shows power limit 88% and in the graph below you see power limit to 130%
your card boosts to 2076mhz and you have an overclock applied not a underclock.
memory overclocked 200mhz
and putting voltage to max and power limit down is no help either because you force a power\volt limit earlier.
and why say underclock and then show oc scanner result ?

maybe you have set the automated oc profile trough oc scanner at startup?

i think you messed up your over\underclock with profiles
try to use your card standard

 

[Vayra86]

Revert to stock GPU settings. Set power target to 95% or 90%. Bump memory clock to +500. Leave the core voltage alone and run core frequency +0.

If that is stable... bump core frequency by no more than 50mhz. Thats the best an aging 1080 will provide. If not stable, reduce memory OC.