Is my GPU dead?

[IAmAndre]

Hi

 

I've been having issues with my RTX 3070 for years and today I decided to do something about it. Basically the fans would run at full speed when gaming, and the PC would eventually crash. A quick fix for the past few years has been to lower the power limit in the Nvidia app. It has worked for a couple years but then even that wouldn't prevent the crashes, unless I play older games.

 

Today I decided to repaste it, and while I'm seeing some small progress, it still didn't do much. Before the repasting, the Heaven benchmark could only run for a few seconds before the PC crashes. After the repasting, I was able to run for a minute or so, then I closed it manually. The fans were super loud but I assumed it was normal.

 

Then I started playing Age of Empires 4, and I ran into the issue again: the game starts, the fans get super loud, and then it crashes. Yesterday the game crashed after a few seconds, today it did after a couple minutes. According to GPU-Z, the GPU usage is around 80%, the hot spot at 105° and the GPU temperature is around 84°.

 

As for the repasting methodology, I used Linus' approach, consisting in cleaning the previous thermal paste and a "nut" of the new thermal paste in the middle. I didn't spread or anything.

 

Is there anything that I can try or should I just replace the graphics card?

[Mumintroll]

In the Nvidia control panel, what power setting do you have there? If it is maximum power then switch to normal.

When put to normal try then to run without any app that controls the fan curve and power limit. Let it use its own default values.

[IAmAndre]

9 minutes ago, Mumintroll said:

In the Nvidia control panel, what power setting do you have there? If it is maximum power then switch to normal.

When put to normal try then to run without any app that controls the fan curve and power limit. Let it use its own default values.

It was in normal. Actually I did a fresh Windows install just yesterday and didn't mess with the Nvidia settings so far.

[leclod]

I have a 3070.

I did undervolt mine heavily because my airflow is limited.

I recommend you undervolt, you could try my setting (with MSI Afterburner).

Did you make pictures when you repasted ? how did the pads look ?

I recently repasted and repaded. I reduced 8°. I prefer to spread the paste.

But it seems it's tricky. On my Ventus it seems I've got to be carefull that the 4 standoffs  for the 4 screws are well positionned in the pcb.

I did retry a few times with varying success. I also suggest you retry more carefully. 

 

Edited July 5, 2025 by leclod

[xAcid9]

1 hour ago, IAmAndre said:

As for the repasting methodology, I used Linus' approach, consisting in cleaning the previous thermal paste and a "nut" of the new thermal paste in the middle. I didn't spread or anything.

Did Linus really do that? For GPU? 

GPU die need overall coverage of thermal paste since it doesn't have heat spreader, unlike CPU. 

 

Try repaste again but this time spread the thermal paste evenly and thinly rrom edge to edge of the GPU die. 

[leclod]

8 minutes ago, xAcid9 said:

Did Linus really do that? 

Are those the specs of your system ? ddr3 ?

[IAmAndre]

9 minutes ago, xAcid9 said:

Did Linus really do that? For GPU? 

Yes, he did on 5:43

I also watched JaysTwoCents video, where he spreads but ended up taking the Linus approach 

 

 

[IAmAndre]

58 minutes ago, leclod said:

I have a 3070.

I did undervolt mine heavily because my airflow is limited.

I recommend you undervolt, you could try my setting (with MSI Afterburner).

Did you make pictures when you repasted ? how did the pads look ?

I recently repasted and repaded. I reduced 8°. I prefer to spread the paste.

But it seems it's tricky. On my Ventus it seems I've got to be carefull that the 4 standoffs  for the 4 screws are well positionned in the pcb.

I did retry a few times with varying success. I also suggest you retry more carefully. 

 

That's interesting. I didn't take pictures, but I didn't move the pads since they were still on the PCB 

 

So did you speak your paste or let the cooler spread it for you?

[leclod]

31 minutes ago, IAmAndre said:

Yes, he did on 5:43

Linus seems young and unexperienced here.

What he says about too much paste insulating is wrong.

I believe the amount of paste he uses to be not enough or very close.

29 minutes ago, IAmAndre said:

but I didn't move the pads since they were still on the PCB 

You probably are alright, but it's wise having the right thickness pads available when repasting.

29 minutes ago, IAmAndre said:

So did you speak your paste or let the cooler spread it for you?

I spread my paste over the whole die and still let the cooler spread it afterwards.

(It's not possible to spread that thin)

 

Which paste are you using ? MX-4 or 6 are good choices for example

 

(It's wise making pictures along the way when doing something like repasting)

Edited July 5, 2025 by leclod

[xAcid9]

27 minutes ago, IAmAndre said:

Yes, he did on 5:43

I also watched JaysTwoCents video, where he spreads but ended up taking the Linus approach 

It's not really a problem if you know how much thermal paste you need for it to spread evenly and cover the whole die. 
If it too little, you'll get area/s of the die that doesn't have thermal paste coverage and ended up getting high hotspot temp. 

[IAmAndre]

3 hours ago, leclod said:

Linus seems young and unexperienced here.

What he says about too much paste insulating is wrong.

I believe the amount of paste he uses to be not enough or very close.

You probably are alright, but it's wise having the right thickness pads available when repasting.

I spread my paste over the whole die and still let the cooler spread it afterwards.

(It's not possible to spread that thin)

 

Which paste are you using ? MX-4 or 6 are good choices for example

 

(It's wise making pictures along the way when doing something like repasting)

 

3 hours ago, xAcid9 said:

It's not really a problem if you know how much thermal paste you need for it to spread evenly and cover the whole die. 
If it too little, you'll get area/s of the die that doesn't have thermal paste coverage and ended up getting high hotspot temp. 

Alright I'll try again. Should I wipe all the existing thermal paste or can I just add some more (assuming that I don't spread)?

 

Btw this was 2015, I wouldn't consider Linus to be inexperienced at that time

 

Update: I disassembled the card again and I really don't think there weren't enough thermal compound because that's well spread and there's much more than originally. I still added some more but it didn't change a thing. It might be time to retire this card.

[leclod]

7 hours ago, IAmAndre said:

Update: I disassembled the card again and I really don't think there weren't enough thermal compound because that's well spread and there's much more than originally.

GPUs don't get hotter over time as an aging process. You shouldn't change your gpu because of that.

How's the airflow in your case ? Did you check the heatsink for dust ?

Did you consider undervolting as I wrote before ? it's easy and nothing can go wrong (apart from the system being unstable)

Which's your thermal paste ? (a picture would have been nice)

Edited July 6, 2025 by leclod

[IAmAndre]

8 hours ago, leclod said:

GPUs don't get hotter over time as an aging process. You shouldn't change your gpu because of that.

How's the airflow in your case ? Did you check the heatsink for dust ?

Did you consider undervolting as I wrote before ? it's easy and nothing can go wrong (apart from the system being unstable)

Which's your thermal paste ? (a picture would have been nice)

So I'm currently keeping my case open so airflow isn't an issue. I also didn't notice much dust anywhere on the GPU. I suspect that I could still use the GPU by undervolting it or lowering the power limit like before, but that's still very annoying and I switched from AMD to Nvidia to avoid running into thermal issues again so I'm kind of frustrated. I'm using a no-name thermal paste but I don't think that's the issue because I saw good online reviews and I'm using it on my CPU as well. I remember posting on this forum before using it and I was told I should be fine.

 

At this point, the only thing I can think of is the thermal pads so I'll post a picture when I can to make sure that they are placed properly.

[leclod]

7 minutes ago, IAmAndre said:

At this point, the only thing I can think of is the thermal pads so I'll post a picture when I can to make sure that they are placed properly.

Thermal pads have nothing to do with die hotspot hitting 105° with fans at full blast.

7 minutes ago, IAmAndre said:

I suspect that I could still use the GPU by undervolting it but that's still very annoying

Undervolting isn't annoying, my gpu is undervolted from day one.

Same performance from less power so less heat and less noise.

Edited July 6, 2025 by leclod

[IAmAndre]

20 minutes ago, leclod said:

Thermal pads have nothing to do with die hotspot hitting 105° with fans at full blast.

Undervolting isn't annoying, my gpu is undervolted from day one.

Same performance from less power so less heat and less noise.

So can we rule out the thermal pads entirely? If so, what could be the reason why a GPU suddenly requires to be undervolted in order to just run at stock speed?

 

I mean GPUs are supposed to be functional out of the box, and for me it doesn't make sense to do any tinkering like that, especially out of the box.

[leclod]

You're right, your gpu shouldn't crash your system.

When too hot it should throttle but not crash as far as I know.

84° with fans at full blast is weird though (how much rpm or % is your full blast ?).

21° between die average temp and hotspot temp sounds like a lot. Mine is 15° under load.

Crash could be due to VRam getting too hot due to the thermal pads. Or something else is the cause...

Undervolting would help but do as you want.

Difficult to help you from here.

 

What does your "out of the box" mean when you've had issues for years ?

Which gpu have you got ?

Edited July 6, 2025 by leclod

[IAmAndre]

1 hour ago, leclod said:

What does your "out of the box" mean when you've had issues for years ?

Which gpu have you got ?

Edited 5 minutes ago by leclod

This comment was referring to you undervolting your graphics card from day one. But before that one, I've been using a Vega 56, with the blower style cooler, and the noise levels were atrocious. The fact that I had a 4K monitor didn't help either. I also had to undervolt it to reduce the noise but also reduce the crashes and the power consumption because my PSU were struggling, but I ended up replacing it with a 750W one, which I'm still using. I would still get some crashes, even with all that tinkering, so I ended up replacing it with the 3070. I bought this one because the consensus even here on this forum was that the Vega 56 *required* some undervolting so I thought going Nvidia would avoid me all this trouble.

 

Anyway now I think I'll just lower the power limits for now and just wait for Panther Lake laptops in 6 months.

[leclod]

4 minutes ago, IAmAndre said:

I also had to undervolt it

You know undervolting and underclocking are two different things, right ?

[IAmAndre]

4 minutes ago, leclod said:

You know undervolting and underclocking are two different things, right ?

I know, I had to undervolt the Vega 56 (and I found the process very annoying) but I haven't undervolted the 3070 so far.