Question RTX 3080 Ti causing restarts

[mihailt]

Hi fellas,

Recently I upgraded my GPU from an RTX 2070 (EVGA XC3 Ultra model) to an RTX 3080 Ti Founders Edition. After several days of using it, I've had my PC restart itself 3 times so far. While I'm pretty sure it's my PSU not handling the transient power spikes and I will update that ASAP, I can't figure out why it happens in the situations it does...

Here are my full system specs (I'm listing even peripherals on purpose):
Seasonic SSR-750PX PSU
Aorus H370 Gaming 3 WiFI motherboard
i7 8700 (non-K) CPU
HypeX 32GB DDR4 at 2666MHz CL13
RTX 3080 Ti Founders GPU
512GB Samsung 970 Evo NVMe
Seagate Barracuda 2TB HDD
4 Deepcool CF120 ARGB fans
240mm AIO with ARGB on the block
2 RGB RAM extensions thingys
Soundblasters Z sound card
2m of RGB light strip powered by the PC
Corsair K70 Rapidfire Keyboard
Logitech G502 Mouse

All the RGB is lit up during use.

So where's the problem? Well, the restarts only happened in older games such as PUBG (and even then only when watching deathcams and replays) and MW2 Remastered. On the other hand, newer titles such as MW 2019 run without a hitch. Another thing which I don't know if relevant, is that the restars happen when my iCue software runs with PlugIn's enabled controlling the GPU lighting and custom fan profiles.

As I said, I'm upgrading to a 1000W next week (either a Seasonic GX-1000 or a Corsair RM1000x 2021 model), but I just got curious why would the older titles trip the PSU (which I guess happens when both CPU and GPU are working hard and the GPU experiences a transient power spike). If it helps, I have not reinstalled my Windows after the update, nor have I cleaned up my drivers with DDU, I just updated them with the latest version from Nvidia after installing the 3080Ti.

Any comments welcome.

P.S. Note that the new PSU is supposed to power up an entirely new system with a 12700K, Z690 MoBo, DDR5 RAM (I'm thinking 5600MHz at CL36) and 2 more NVMe's + 1 or 2 SATA SSDs, so if you have a suggestion other than the models listed above, please share.

 

[Deleted member 2731765]

What's the current Make/brand and model number of your current PSU. You didn't mention this in your OP, or you did ? If newer graphic demanding titles work fine, then I won't assume the PSU to be the culprit here, at least for now.

Any overheating issues on your CPU, or GPU in general ?

 

[mihailt]

Sorry, I forgot this - current PSU is a Seasonic SSR-750PX, i.e. 750W Platinum PSU, but the model from around 2018. I thought being a Platinum it might hold up, but I guess not.

Overheating is not an issue on any component I think. CPU doesn't go over 75 degrees Celsius and the GPU die measures at 75c max.

However, the reason I started using the iCue plugin for custom fan curve was that Memory Junction temperature in HW Info was measure at 100c peak value, which was a bit too close for comfort. With my fan curve, it doesn't go above 90c.

Editing OP with PSU model for more clarity.

 

[Deleted member 2731765]

That's a pretty good/decent PSU if you ask anyone. The Model might be from 2018, but for how long have you been using this PSU ? How many years ? It's kind of odd that older games such as PUBG cause a system restart, and not the newer or graphic demanding titles.

Had the PSU been the culprit, then the PC should restart even while playing newer games as well. Could be some other hardware issue as well, though I can't say for sure. But to narrow this down, try checking the current PSU on some other system/PC as well, IF possible.

 

[mihailt]

That's a pretty good/decent PSU if you ask anyone. The Model might be from 2018, but for how long have you been using this PSU ? How many years ? It's kind of odd that older games such as PUBG cause a system restart, and not the newer or graphic demanding titles.

Had the PSU been the culprit, then the PC should restart even while playing newer games as well. Could be some other hardware issue as well, though I can't say for sure. But to narrow this down, try checking the current PSU on some other system/PC as well, IF possible.

It is a good PSU, but I read somewhere that some PSUs have their OCP triggered easier than others in general. I've been using it for 4 years straight. Powered a 1070 and a 2070 during this time. No overclocking involved whatsoever.

Don't have another system to test it in, unfortunately.

 

[Phaaze88]

While I'm pretty sure it's my PSU not handling the transient power spikes...

Yeah, that should be shut offs, with no restarts. I'm thinking driver or software, and this part here:
"the restarts happen when my iCue software runs with PlugIn's enabled controlling the GPU lighting and custom fan profiles."
... definitely smells of that.

Do you have 'automatic restart on system failure' disabled?
How to: https://help.maingear.com/article/62-how-to-disable-automatic-restart-on-system-failure

I have not reinstalled my Windows after the update

Shouldn't need to if just changing a gpu. A different motherboard, or carrying over a C drive from another system is a must.

nor have I cleaned up my drivers with DDU

I know it's not necessary to run before every driver update the user does, but here, you're having problems, and a clean driver install is part of the troubleshooting process.

 

[mihailt]

Yeah, that should be shut offs, with no restarts. I'm thinking driver or software, and this part here:
"the restarts happen when my iCue software runs with PlugIn's enabled controlling the GPU lighting and custom fan profiles."
... definitely smells of that.

Do you have 'automatic restart on system failure' disabled?
How to: https://help.maingear.com/article/62-how-to-disable-automatic-restart-on-system-failure


Shouldn't need to if just changing a gpu. A different motherboard, or carrying over a C drive from another system is a must.


I know it's not necessary to run before every driver update the user does, but here, you're having problems, and a clean driver install is part of the troubleshooting process.

Thank you for responding! "Automatic restart on system failure" is currently enabled.

I presumed it was caused by the PSU because both times the last item recorded in Event Viewer before the restart was Critical, Event ID 41, Kernel-Power, Task 63.

Would you recommend the following steps in this case?

1. Disable Automatic restarts
2. Clean up drivers with DDU and doing a fresh install of Nvidia drivers
3. Disabling plugins on iCue (The reason I had those enabled is because with the default fan profiles the Memory Junction reported by HWInfo hits 100 degrees vs 90c with the custom profile)

 

[Phaaze88]

I presumed it was caused by the PSU because both times the last item recorded in Event Viewer before the restart was Critical, Event ID 41, Kernel-Power, Task 63.

Windows throws that one up every time it doesn't properly shut down, regardless of the cause. That error doesn't really help us or you narrow things down.

Would you recommend the following steps in this case?

1. Disable Automatic restarts
2. Clean up drivers with DDU and doing a fresh install of Nvidia drivers
3. Disabling plugins on iCue (The reason I had those enabled is because with the default fan profiles the Memory Junction reported by HWInfo hits 100 degrees vs 90c with the custom profile)

1)Yes.
2)Yes.
3)Are you talking about ram, or the ram on the gpu?

 

[Phaaze88]

3)Are you talking about ram, or the ram on the gpu?

Never mind - I'm half asleep here.
90C is definitely better long term. The memory is specced for up to 105C: https://www.micron.com/products/ultra-bandwidth-solutions/gddr6x
... but it's safer to keep it below 95C.

 

[mihailt]

Windows throws that one up every time it doesn't properly shut down, regardless of the cause. That error doesn't really help us or you narrow things down.


1)Yes.
2)Yes.
3)Are you talking about ram, or the ram on the gpu?

Well, I did all three in succession - same result, unfortunately.

I should note that when the PC restarts, all lights go out for a moment and then it powers itself back up, which is why I think it's the PSU. I know that HWInfo detects over 140W max power draw for the CPU and ~350W for the GPU, but if a transient spike happens approaching 500W or 600W, it could be enough to trip an OCP, no?

Could the GPU itself be malfunctioning?

 

[Phaaze88]

I should note that when the PC restarts, all lights go out for a moment and then it powers itself back up

You mean the room lights too?

 

[mihailt]

You mean the room lights too?

No, no, just the keyboard, mouse and all RGB inside the PC.

 

[Phaaze88]

Download and run Msi Afterburner. Unlink the power and temperature limits - click chain link icon - drag only the power limit slider to 90%. This should still allow the card to yield high boost clocks.
Click Apply, and play the problem games. Does the system still restart?
If it does, lower it to 80% and test again. Still restarts? By this point, the highest boost clocks likely aren't possible, and it shouldn't be too hard on the psu.

if a transient spike happens approaching 500W or 600W, it could be enough to trip an OCP, no?

Unfortunately, that one's a bit over my head.

 

[mihailt]

Just wanted to do an update here in case someone from the future comes across this thread.

Ultimately, I purchased a new PSU - the Seasonic Fcous GX-1000. After a month of use - not a single problem in any game I tried.

Meanwhile, I sold my old 750W PSU to a local buyer who uses it for a 3070 and has not reported any problems in that month.

Coincidentally, GamersNexus just released a new video detailing these transient power spikes on new and upcoming GPUs:

 

[Phaaze88]

I saw this video last night.
The Corsair AX850 Titanium that's seen less than a year of use might not be good enough for whichever RTX 40 or RX 7000 gpu I eventually settle on... this sucks.