Question RTX 4060 Ti crashes with loud sound when doing simple tasks on Ubuntu and Windows 11 ?

[Mar2012sep]

I recently got a RTX 4060 Ti but I’m unable to use it properly. Every time I try to play even a relatively simple game like Planet Coaster, the GPU starts making a very loud noise, spinning like crazy and the entire computer freezes. In some cases, the screen also goes black.

This issue occurs on both Ubuntu 24.04.2 LTS and Windows 11, so it doesn’t seem to be specific to the operating system. However, since I primarily use Ubuntu, I would prefer solutions that work in Ubuntu.

For additional context, my computer is custom-built, and the motherboard is an ASUS Prime. All cables appear to be properly connected, and everything else in the system seems to be working fine.

 

[Lutfij]

Welcome to the forums, newcomer!

I recently got a brand-new RTX 4060 Ti
So you performed a GPU upgrade. Did you run DDU in Safe Mode(for Windows 11) to remove all GPU drivers(Intel, AMD and Nvidia) to later manually install the latest GPU driver sourced from Nvidia's support site in an elevated command?

You forgot to mention the make, model and age of the PSU.

For additional context, my computer is custom-built, and the motherboard is an ASUS Prime.
When posting a thread of troubleshooting nature, it's customary to include your full system's specs. Please list the specs to your build like so:
CPU:
CPU cooler:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
include the age of the PSU apart from it's make and model. BIOS version for your motherboard at this moment of time.

 

[Mar2012sep]

Thank you for welcoming and reply!

I haven’t performed a GPU upgrade, I bought the components and built a pc from scratch.
I apologize the missing info, here it comes:

CPU - AMD Ryzen 7 7700 (16) @ 5.389GH
CPU cooler – AMD, it came as a package with the CPU
Motherboard – Asus Prime X670-P WI-FI
RAM – Kingston Fury Beast 64x2 GB KIT 5600Mhz DDR5
SSD/HDD - Kingston NV2 M.2 2000GB
GPU motherboard - AMD
GPU – MSI GeForce RTX 4060 Ti VENTUS 3X OC
PSU – Corsair RMe Series RM850e
Chassis - HYTE Y40 S-Tier Aesthetic Case Mid Tower
OS - Ubuntu 24.04.2 LTS x86_64 (had the same problem on Windows 11)
Monitor – Samsung G5 Odyssey C34G55 34”

All components are bought on september 2023 and the motherboard BIOS version is 1654.

 

[qxp]

Video cards aren't really supposed to make loud noises, so it makes sense to investigate that first. Maybe the fan is damaged and its overheating. Or, if it is a high pitch squeal, it could be a voltage regulator of some sort.

Can you run it with case opened so you can locate where the sound is coming from ? There is also a spectrum app for android that would let you plot sound spectrum in real time - if there is a specific frequency peak it could be a clue.

 

[Mar2012sep]

It seems that the loud noise comes from the fans, air friction sound, not squeal like.
When the fans went in berserk mode, the peak was around 2.800 Hz (spectrum analyzer. as suggested).

You can hear the sound and see what's happening when I start a game: Here

 

[qxp]

Ok, the noise is normal. The fans are just going full blast. Nvidia should have control panel that displays GPU temperature. Check that before starting the app and watch it as it starts. If it ramps up really fast, try finding a smaller GPU load, maybe even as small as moving a window around as compositor uses GPU. If you see temperatures peak on small load I would suspect a problem with thermal interface between the chip and the heatsink.

 

[Mar2012sep]

I've started the app, GPU went from zero to 44% usage, 61 degrees C. Really fast I closed the app and even now after a few minutes, while I'm typing here, the fans are om rampage, temperature and usage same. The values are blocked on 44% and 61 degrees.
I've restarted "mission center", something similar to windows task manager, and it only sees one GPU, the motherboards (AMD), GEFORCE is not even listed, and poor thing still full blast, but invisible...

I also tried moving a window around, the temperature was ok, between 44-45 degrees celsius, no peaks.

Thank you so much for your suggestions, what next?

Ok, the noise is normal. The fans are just going full blast. Nvidia should have control panel that displays GPU temperature. Check that before starting the app and watch it as it starts. If it ramps up really fast, try finding a smaller GPU load, maybe even as small as moving a window around as compositor uses GPU. If you see temperatures peak on small load I would suspect a problem with thermal interface between the chip and the heatsink.

 

[qxp]

On Linux there is a command line tool "nvidia-smi" - what does it show before starting and after start and stop ?

 

[qxp]

Also try waiting 10-30s and run nvidia-smi again - it should show reduced temps and fan speeds when the card is not used.

 

[Mar2012sep]

Also try waiting 10-30s and run nvidia-smi again - it should show reduced temps and fan speeds when the card is not used.

Before starting: Here
After start, fans full blast, froze: Here
I stopped the program and it's the same info.
I have to restart my pc in order to stop the fans running.

 

[Mar2012sep]

Also try waiting 10-30s and run nvidia-smi again - it should show reduced temps and fan speeds when the card is not used.

Another funny thing, when the pc boots I have to plug off the cable to GPU, otherwise it will never start. I get a black screen and nothing. I unplug the cable and linux starts. Plug in the cable to GPU after starting and all is good again. Well... as good as it can be....
Could it be a compatibility problem? I only have one screen, HDMI in motherboard card and Display port i GEForce.
Does it make sense?

 

[qxp]

Display cable issue is weird. Try to go through your BIOS settings in case there is something talking about preference for motherboard graphics vs the card. Also anything talking about slot or memory allocation. Maybe something interferes.

Another thing to try is to upgrade to latest NVidia drivers - I think they have CUDA 12.7 out by now, and a newer driver as well. It could be the drivers included with Ubuntu are not quite new enough for your card.

 

[qxp]

Also, do I understand correctly that on the second screenshot nvidia-smi is refusing to connect ? This definitely points towards a card crash, any messages in dmesg if you can still access it ?

One way to get up to the second info is to have a terminal for dmesg running
while [ x==x] ; do sleep 1; dmesg ; done

And another with "nvidia-smi -l"

You might also find useful command "nvtop"

 

[Mar2012sep]

Also, do I understand correctly that on the second screenshot nvidia-smi is refusing to connect ? This definitely points towards a card crash, any messages in dmesg if you can still access it ?

One way to get up to the second info is to have a terminal for dmesg running
while [ x==x] ; do sleep 1; dmesg ; done

And another with "nvidia-smi -l"

You might also find useful command "nvtop"

Yes, that's right, the GPU refuses to connect, " GPU is lost" it says.
nvidia-smi-l command:

 

[Mar2012sep]

nvtop:

It says that there is a problem with system program and asks me if I want to report it.

 

[qxp]

Looks like nvtop has a problem with your AMD driver, nothing related to GeForce shutdown.

The "GPU is lost" message is new to me, but it looks like this can happen with NVidia cards that go overtemp:

https://forums.developer.nvidia.com...aring-after-installing-nvidia-driver/289770/8

In the link above someone was trying to use a server GPU without a fan. You have a fan (three in fact), but maybe it is something similar - try to follow the same steps and compare temperature logs with what you see for the other poster.

Also, double check the card for any obstructions - maybe there is a bit of sticky tape obstructing the heatsink ?