Question Undervolting fixed my problem, faulty GPU or PSU?

[tiagom]

Hello there, I have posted in this forum before looking for answers to my problem.

I went through multiple tries to fix my black screen while gaming flickers, with sound in background and then PC restart.

Tried several drivers, DDU multiple times, bought a GPU holder because of GPU sag, updated mobo bios, disable resize bar... Endless tries to fix this.

Only recently I found a way to replicate the problem.

On MSI afterburner I reset all settings to default and let 3dmark running on timespy, and on first or second GPU test, I start seeing screen flickers and then it happens... This is about when GPU is draining about 350w or so.

If I undervolt GPU to a fixed 875mv at 1890mhz then I don't have any issues on that timespy test benchmark.

So why I am asking this question you may ask?

I feel like I am not using the full potential of my system, and I want to replace the faulty part. I just don't know if it's GPU or PSU problem. Any thoughts on this?

I want my system to handle extreme scenarios of power draining and not be limited to a specific undervolting just to make it work. Feels like I am putting a bandage on a bigger problem and not fixing it at all!

My GPU: Inno3d RTX 3090Ti
My PSU: Gigabyte P1000gm 1000W Plus Gold

Thank you all for your time and any help you can provide!

 

[Lutfij]

In order to rule out your PSU being the root of the issue, source another unit that's higher in wattage and higher in quality, to see if the issue persists without an undervolt. If the issue is gone, then it's your PSU. You forgot to mention the make and model of your case, the orientation and number of fans and the ambient room air temps not to mention the temps of your GPU.

 

[tiagom]

In order to rule out your PSU being the root of the issue, source another unit that's higher in wattage and higher in quality, to see if the issue persists without an undervolt. If the issue is gone, then it's your PSU. You forgot to mention the make and model of your case, the orientation and number of fans and the ambient room air temps not to mention the temps of your GPU.

Hi @Lutfij
Thank you for your reply, I didn't mention temps because they looked fine during the benchmarks, around 60~70ºC

I will try to swap the PSU and test it again,
I believe this PSU is not that good right? Which one would you recommend?
Thank you very much!

CPU: i7 12700k
RAM: 2x16GB DDR4-3600MHz CL18
Motherboard: Asus Prime H670 Plus D4
Case: Cooler Master MasterBox MB520
SSD: 2xM.2 2280 Kingston NV1 2TB 3D TLC NVMe

 

[Lutfij]

Swap as in borrow, not buy...and I can't recommend that you throw money at a problem, rather take hardware borrowed from a friend or neighbor and then see if the issue persists.

 

[tiagom]

Swap as in borrow, not buy...and I can't recommend that you throw money at a problem, rather take hardware borrowed from a friend or neighbor and then see if the issue persists.

Unfortunately I don't know anyone that could borrow me one to test it out.
So I guess I will need to order a new one, if problem is fixed I keep it if not I try to return it? I think its allowed in my place.

 

[tiagom]

@Lutfij i forgot to mention in the main post, that the GPU already went through a month of RMA process, it went from the shop to the factory, I have receipt proof of the courier. However I can not be entirely sure which test they did there, but they did mention that GPU was fine and that they didn't found any issue with it. So does this give any lead that can be PSU the real culprit here? Thank you!

 

[scout_03]

use occt to stress test psu and hwinfo for the details on voltages with temp and fan speed in system on sensor page .

 

[tiagom]

Just wanna stop by and say that the problem is fixed.

I can't reproduce anymore the problems I was facing on 3dmark without undervolt.
Reset master gpu settings on MSI Afterburner, ran the 3dmark timespy and now it runs till the end, no screen flickers, no black screens, no restarts.

Board draining around 450w on some times without any issues.
Problem was indeed a bad PSU that couldn't handle the power spikes (it seems?).
Solved with new PSU.

Thank you all!

 

[tiagom]

Sorry it started happening again, if I run the games with my GPU on stock settings the whole screen freezes and then usually i have to restart the computer.

1. Already RMA'ed the GPU card before and they told no issues on it, I guess they run benchmarks, which i also run (timespy) and no issues with the new PSU (check comment before this one).
2. As I said, I already changed PSU to rm1000x shift series.

What else can it be? I can only play with my GPU undervolted at 0.875v , this is very odd.


3. Already tried also undervolting my CPU to fixed 1.250v and manual offset [-] 0.150v

What else can I try to find where the problem is?
Thank you!

 

[JCarax]

Sorry it started happening again, if I run the games with my GPU on stock settings the whole screen freezes and then usually i have to restart the computer.

1. Already RMA'ed the GPU card before and they told no issues on it, I guess they run benchmarks, which i also run (timespy) and no issues with the new PSU (check comment before this one).
2. As I said, I already changed PSU to rm1000x shift series.

What else can it be? I can only play with my GPU undervolted at 0.875v , this is very odd.


3. Already tried also undervolting my CPU to fixed 1.250v and manual offset [-] 0.150v

What else can I try to find where the problem is?
Thank you!

It might a component/driver incompatibility?
If that is the case it would be near impossible to replicate your issue on a test bench since the GPU itself isn't at fault here.
Also just wondering, when you sent the gpu for RMA did you send it with the PCIE cable or without it?

 

[tiagom]

It might a component/driver incompatibility?
If that is the case it would be near impossible to replicate your issue on a test bench since the GPU itself isn't at fault here.
Also just wondering, when you sent the gpu for RMA did you send it with the PCIE cable or without it?

I have send it without.
Between that time and before this thread, I had bought a gpu vertical holder with a specific riser cable, problems continued with the riser cable and even with gpu directly connected to mobo.
This was like 1 or 2 months ago.
Then last week bought new PSU and connected everything back with the new PSU cables, this PSU came with a 12vhpwr cable which I used to connect to the GPU, instead of using adapters like it was before.
Problem persisted.

What works:
- Play games with GPU undervolted to 0.875v at 1890MHz + 600 memory
- Run timespy benchmark not undervolted, gpu at stock settings.

What doesn't work:
- Play games more than 10~15minutes with GPU at stock settings.

I notice on MSI when I am playing and testing GPU at stock, the core clock go as high as 2100 or so...
And this GPU is listed to have boosted core clock at 1890MHz, this is crazy isn't it?
I guessed it came too much factory overclocked? And my only option is running it at undervolt 0.875@1890?

 

[tiagom]

I wonder if it can be motherboard related? Maybe this motherboard is not good enough? Bad voltage regulators or so? And what's why I can't run the GPU on stock? What you guys think? Thank you

 

[ChardBurger]

Hi bud, I think I might be experiencing the same problems you are, you have any updates on yours?

 

[tiagom]

Hi bud, I think I might be experiencing the same problems you are, you have any updates on yours?

Sadly no, it's still undervolted

 

[ChardBurger]

Sadly no, it's still undervolted

Kinda doing the same things with my system tho mines' mid-range and with an AMD GPU (6600XT). Also noticed that when I share screen at Discord, my stream flickers. Don't know if it happens to you also and if it's connected somehow.

 

[DarkForceX]

Sadly no, it's still undervolted

Any news about your issue?. Writing from my PC with 'undervolted' Gigabyte 2080 gaming super OC. Which is ment to work on 1.050v @ 2010mhz max @ stock settings. For some reason that's how i used the card for like 6 months before '' black screen + fans at full speed '' problem started on heavy gpu load at stock voltage and clock speeds. undervolting to 0.843v and downclocking to 1830mhz runs smooth. But i also think this is more like putting a temporary bandage on a broken leg. I got a spare GTX 1060 6GB which also uses 1.050V max and it doesn't crash as the 2080 super does. i know that the 1060 is much less powerful gpu but since its able to use that amount of wattage without problems tells me that the issue is related to the other GPU itself. i also renewed it's thermal paste. temps are between 60 and 70 during heavy load.

If it helps im using the Evga 750Watt Gold. Ryzen 5600X at 4.650mhz with precision overdrive on. XMP on. mobo: gigabyte ds3h b550m. 32GB corsair vengeance 3200mhz ddr4. scythe ninja 5 cpu cooler which has 2 fans connected on both sides. + i got 1 120mm be quiet case fan at front. and a 80mm case fan at the back. which are connected to a input extender since mobo doesn't have that many fan slots. And that's exactly how i used the system for over 6 months. So i don't get what the problem is. and why now. I also updated the bios to the latest. also reinstalled windows 11 x64 pro downloaded from official microsoft website using windows media creation tool on a 32gb usb stick. and installed it a fresh system so without restoring backup settings of the past windows 11 it was on. i also tried using the driver which windows update automatically installs. as well as using DDU to install the newest nvidia driver for this gpu. both of them giving the same problem. i don't think that it's a dying gpu issue because i had a dead gpu before and the symtoms were looking way too different as it does now. since undervolting helps that has to be a problem with the power connector components of the GPU. but again.. why work for 6 months and then decide to stop working.. thats the part i don't get. I might send my gpu to a hardware repair shop to let them take a look at. I just hope they won't charge me a amount which i could use to buy a rtx 4000 series for.

 

[tiagom]

Any news about your issue?. Writing from my PC with 'undervolted' Gigabyte 2080 gaming super OC. Which is ment to work on 1.050v @ 2010mhz max @ stock settings. For some reason that's how i used the card for like 6 months before '' black screen + fans at full speed '' problem started on heavy gpu load at stock voltage and clock speeds. undervolting to 0.843v and downclocking to 1830mhz runs smooth. But i also think this is more like putting a temporary bandage on a broken leg. I got a spare GTX 1060 6GB which also uses 1.050V max and it doesn't crash as the 2080 super does. i know that the 1060 is much less powerful gpu but since its able to use that amount of wattage without problems tells me that the issue is related to the other GPU itself. i also renewed it's thermal paste. temps are between 60 and 70 during heavy load.

If it helps im using the Evga 750Watt Gold. Ryzen 5600X at 4.650mhz with precision overdrive on. XMP on. mobo: gigabyte ds3h b550m. 32GB corsair vengeance 3200mhz ddr4. scythe ninja 5 cpu cooler which has 2 fans connected on both sides. + i got 1 120mm be quiet case fan at front. and a 80mm case fan at the back. which are connected to a input extender since mobo doesn't have that many fan slots. And that's exactly how i used the system for over 6 months. So i don't get what the problem is. and why now. I also updated the bios to the latest. also reinstalled windows 11 x64 pro downloaded from official microsoft website using windows media creation tool on a 32gb usb stick. and installed it a fresh system so without restoring backup settings of the past windows 11 it was on. i also tried using the driver which windows update automatically installs. as well as using DDU to install the newest nvidia driver for this gpu. both of them giving the same problem. i don't think that it's a dying gpu issue because i had a dead gpu before and the symtoms were looking way too different as it does now. since undervolting helps that has to be a problem with the power connector components of the GPU. but again.. why work for 6 months and then decide to stop working.. thats the part i don't get. I might send my gpu to a hardware repair shop to let them take a look at. I just hope they won't charge me a amount which i could use to buy a rtx 4000 series for.

Hi, for me it's no longer a issue because I simply gave up on that gpu and bought a new one...
Sorry can't be of more help!

 

[DarkForceX]

Hi, for me it's no longer a issue because I simply gave up on that gpu and bought a new one...
Sorry can't be of more help!

No worries. Thank you for letting me know anyways. im happy to hear that your issue doesn't occur with the new gpu. May i as last ask which GPU you bought?. To see if it's consumes the same amount of voltage etc. because i did saw that the gtx 1060 was using less power compared to 2080 super in OCCT software. I need to know if i have to repair this gpu. or either replace my psu with a 1000w or such lol. very annoying issue.