Can't control fan speed on my GPU

[IAmAndre]

Hi,

 

I've having an issue for several months now with RTX 3070 from Gigabyte and at this point I'm even wondering if the card isn't dying. The fans keep running at full speed when I'm gaming. It's louder in some games than in others, but I mostly play Age of Empires 4 and it typically runs at 3000 RPM, making a huge noise. Normally it would get at this level late in the game when there's a lot going on on the screen, but right now it stays consistently at this speed. Well actually the speed keeps fluctuating between 1000 and 3000 RPM with a temperature constantly around 75°. The worst part is the PC crashed after a while if it keeps running like that, so it's hard to finish a single online game at this point. Very frustrating.

 

I've tried many things to control the speed, including Afterburner, Fan Control, and now the official Gigabyte software. I've come to the conclusion that the fan speed curves are ignored. I just did an experiment by setting up a flat curve at 1100 RPM, but as soon as the game begins the fans start running close to 3000 RPM. I've also had some crashes by doing other experiments, probably because I forced the speed to be too low.

 

All in all, it looks quite unstable to me and I don't know what's going on. Also, I've taken the time to physically clean the GPU. I didn't take it apart but I blew some air in it. It's a white GPU and it looked clean, but I still wanted to try this. As expected, there wasn't much dust inside. Could it be just the GPU dying?

 

The fully detailed PC specs are on my profile.

 

Thanks.

[Sjaakie]

Full speed fans usually do indicate that something is wrong with the gpu. Can you check if the gpu fan header is fully inserted? It's a small cable from the fans to some header on the gpu itself. If it's missing the PWM pin signal it will only be off or run at max speed.

 

Does windows leave any crash logs / messages in system log during crash?

 

Also use ddu to purge nvidia drivers and reinstall them fresh to rule that out.

[PDifolco]

39 minutes ago, IAmAndre said:

Hi,

 

I've having an issue for several months now with RTX 3070 from Gigabyte and at this point I'm even wondering if the card isn't dying. The fans keep running at full speed when I'm gaming. It's louder in some games than in others, but I mostly play Age of Empires 4 and it typically runs at 3000 RPM, making a huge noise. Normally it would get at this level late in the game when there's a lot going on on the screen, but right now it stays consistently at this speed. Well actually the speed keeps fluctuating between 1000 and 3000 RPM with a temperature constantly around 75°. The worst part is the PC crashed after a while if it keeps running like that, so it's hard to finish a single online game at this point. Very frustrating.

 

I've tried many things to control the speed, including Afterburner, Fan Control, and now the official Gigabyte software. I've come to the conclusion that the fan speed curves are ignored. I just did an experiment by setting up a flat curve at 1100 RPM, but as soon as the game begins the fans start running close to 3000 RPM. I've also had some crashes by doing other experiments, probably because I forced the speed to be too low.

 

All in all, it looks quite unstable to me and I don't know what's going on. Also, I've taken the time to physically clean the GPU. I didn't take it apart but I blew some air in it. It's a white GPU and it looked clean, but I still wanted to try this. As expected, there wasn't much dust inside. Could it be just the GPU dying?

 

The fully detailed PC specs are on my profile.

 

Thanks.

Some deep clean + repaste may  help

Maybe the fan header is badly plugged

[IAmAndre]

17 minutes ago, Sjaakie said:

It's a small cable from the fans to some header on the gpu itself. If it's missing the PWM pin signal it will only be off or run at max speed.

Interesting. I've tried to open the card to check that but couldn't as one of the screws has rust on it. However, I managed to take a picture and if that's what you are referring to, it looks fully seated.

 

18 minutes ago, Sjaakie said:

Does windows leave any crash logs / messages in system log during crash?

It does, but I couldn't figure out what was causing the error and can't find the error code in my search history. However, it was definitely something related to the graphics card.

 

57 minutes ago, Sjaakie said:

Also use ddu to purge nvidia drivers and reinstall them fresh to rule that out.

I'll give this a try.

 

[leclod]

My Afterburner has a few toggles. I don't understand all of them all the time. But ...

For example now I'm forcing my gpu fans at 70% manually. Can you do that ?

Edited December 1, 2023 by leclod

[IAmAndre]

20 minutes ago, leclod said:

My Afterburner has a few toggles. I don't understand all of them all the time. But ...

For example now I'm forcing my gpu fans at 70% manually. Can you do that ?

As I mentioned, I set a flat curve but it didn't help. The fan curves seem to be ignored entirely.

[leclod]

9 minutes ago, IAmAndre said:

As I mentioned, I set a flat curve but it didn't help. The fan curves seem to be ignored entirely.

Ok, but that's not what I proposed.

[IAmAndre]

2 minutes ago, leclod said:

Ok, but that's not what I proposed.

OK I'll give this a try if installing clean drivers don't change the situation.

[IAmAndre]

1 hour ago, leclod said:

Ok, but that's not what I proposed.

As I suspected the changes in the fan settings got ignored.

 

However, the clean driver install seems to have made things worse as the GPU fans are now spinning at nearly 4000 RPM now in Age of Empires, with the GPU utilization at around 60%.

 

I've tried playing other games to see if it was an issue with AoE but the problem seems to be GPU itself. I first launched Assassin's Creed 3 and as I suspected, it was silent because the GPU utilization was below 40% for the most part. Then I launched Ethan Carter Redux and the fans would start yelling every time the GPU utilization was over 50%.

[leclod]

10 hours ago, IAmAndre said:

As I suspected the changes in the fan settings got ignored.

Did you move that ruler to 70% ? show me a screenshot

(Fan Control seems to be at the top of the hierarchy, so you've got to disable the control of the gpu fan in Fan Control before trying my "70%")

Do (did) you use other similar software controlling fans ?

Edited December 2, 2023 by leclod

[IAmAndre]

4 hours ago, leclod said:

Did you move that ruler to 70% ? show me a screenshot

(Fan Control seems to be at the top of the hierarchy, so you've got to disable the control of the gpu fan in Fan Control before trying my "70%")

Do (did) you use other similar software controlling fans ?

I moved it even lower, something like 50%. Fan Control wasn't running so AfterBurner was on its own. I then tried to limit the fan speed to 60% on GeForce Experience but it was ignored there too (although Afterburner was running too).

[leclod]

38 minutes ago, IAmAndre said:

I moved it even lower, something like 50%. Fan Control wasn't running so AfterBurner was on its own. I then tried to limit the fan speed to 60% on GeForce Experience but it was ignored there too (although Afterburner was running too).

Could you make me the same screenshot I did ?

[Bagzie]

16 hours ago, IAmAndre said:

Interesting. I've tried to open the card to check that but couldn't as one of the screws has rust on it. However, I managed to take a picture and if that's what you are referring to, it looks fully seated.

 

It does, but I couldn't figure out what was causing the error and can't find the error code in my search history. However, it was definitely something related to the graphics card.

 

I'll give this a try.

 

How on earth does the screw have rust on it? 

[IAmAndre]

2 hours ago, leclod said:

Could you make me the same screenshot I did ?

Sure, I'll make a screenshot this afternoon.

2 hours ago, Bagzie said:

How on earth does the screw have rust on it? 

It's a chemical reaction between iron and the oxygen in their air: 4 Fe + 3 O 2 → 2 Fe 2O 3

[Bagzie]

57 minutes ago, IAmAndre said:

Sure, I'll make a screenshot this afternoon.

It's a chemical reaction between iron and the oxygen in their air: 4 Fe + 3 O 2 → 2 Fe 2O 3

It's a great calculation but you forgot the fact gpu screws are stainless steel with a fair amount of chromium in them which effectively stops the reaction.

[leclod]

1 hour ago, IAmAndre said:

It's a chemical reaction between iron and the oxygen in their air: 4 Fe + 3 O 2 → 2 Fe 2O 3

I'm no expert, but I doubt iron rusts with air only. Always thought it needed H2O.

Edited December 2, 2023 by leclod

[IAmAndre]

1 hour ago, Bagzie said:

It's a great calculation but you forgot the fact gpu screws are stainless steel with a fair amount of chromium in them which effectively stops the reaction.

That's probably why not all the screws have rusted. There are different types of screws on it, I guess with different levels of protection.

1 hour ago, leclod said:

I'm no expert, but I doubt iron rusts with air only. Always thought it needed H2O.

No, you don't need H2O. Some metals can rust in water (and there's oxygen in water) but just the oxygen in the air is enough to make iron rust. Different types of air can cause different types of reactions i.e. iron may rust more easily in some areas than in others. I'm no expert either.

 

Back to the topic, I don't think there's much investigation needed at this point: I just started a new game, Gotham Knights, and the intro custscene was playing with a lot of stutters. Then before the end of the cut scene, the game crashed: it was a black screen with the game audio still playing and the fans running at 100%. It clearly looks like a faulty GPU.

[leclod]

I don't believe in the gpu dying thing.

Did you try that card in a clean system ? In a friends desktop. Or install a Windows on a spare drive.

If the gpu runs hot it should throttle, not crash.

It seems you've got some fan curve going.

[IAmAndre]

5 minutes ago, leclod said:

I don't believe in the gpu dying thing.

Did you try that card in a clean system ? In a friends desktop. Or install a Windows on a spare drive.

If the gpu runs hot it should throttle, not crash.

It seems you've got some fan curve going.

It's not running hot at all. It just has its fans behaving oddly then crashes. It's just unstable.

[leclod]

10 minutes ago, IAmAndre said:

It's not running hot at all. It just has its fans behaving oddly then crashes. It's just unstable.

75° and fan speeds like 3000rpm sounds hot to me. Sorry if I got something wrong.

But Did you try that card in a clean system ?

Edited December 2, 2023 by leclod

[IAmAndre]

12 minutes ago, leclod said:

75° and fan speeds like 3000rpm sounds hot to me. Sorry if I got something wrong.

But Did you try that card in a clean system ?

The thing is, the temperature stays the same no matter how fast the fans are running. If the fans were somewhere between the 1000-2000, it would be around the same temperature. Now the fans suddenly start peaking at 3000-4000 for no reason, with little change to the temperature. So it's not like it's increasing the speed because it needs to, it just behaves in an unstable fashion, and even limiting the fan speed doesn't change it. Now the moments when it does run fine, I'm still noticing things that weren't there before like stutters and even audio delays.

 

Unfortunately I don't have any other system to try. I could set up another boot drive but at this point I don't think the results would be different. Especially considering that even a clean driver install (and reinstalling GeForce Experience) didn't help.

[Bagzie]

9 minutes ago, IAmAndre said:

The thing is, the temperature stays the same no matter how fast the fans are running. If the fans were somewhere between the 1000-2000, it would be around the same temperature. Now the fans suddenly start peaking at 3000-4000 for no reason, with little change to the temperature. So it's not like it's increasing the speed because it needs to, it just behaves in an unstable fashion, and even limiting the fan speed doesn't change it. Now the moments when it does run fine, I'm still noticing things that weren't there before like stutters and even audio delays.

 

Unfortunately I don't have any other system to try. I could set up another boot drive but at this point I don't think the results would be different. Especially considering that even a clean driver install (and reinstalling GeForce Experience) didn't help.

How is it possible for the temperature to be the same no matter how fast the fans are spinning.....

 

I don't need you to give me a formula for it because that's not possible unless there's a temperature sensor issue.

 

Have you spilled liquid onto it at any point? The screws on a GPU don't rust under normal conditions....they just don't.

[IAmAndre]

17 minutes ago, Bagzie said:

How is it possible for the temperature to be the same no matter how fast the fans are spinning.....

 

I don't need you to give me a formula for it because that's not possible unless there's a temperature sensor issue.

 

Have you spilled liquid onto it at any point? The screws on a GPU don't rust under normal conditions....they just don't.

My guess is the fans don't maintain that speed long enough for big changes to happen. It just fluctuates. So that's 4000RPM for 10 seconds then back to 1000, then it keeps cycling. There is some change but only by a few degrees. The only moment when it maintains that full speed is when it crashes and the screen goes black.

 

As for liquids, the only thing I could think of would be when I blew some canned air in it, but that is supposed to be dry air so I can't think of any occasion when some liquid might have been in contact with it.

 

By the way, I just launched Gotham Knights and for a minute or so the game was running at 4-5FPS with the fans at 1100RPM and power draw at 35W. Then it suddenly went back to 60FPS and the game crashed a couple minutes later. It's super unstable.

[leclod]

It could be an issue with power delivery

[IAmAndre]

5 hours ago, leclod said:

It could be an issue with power delivery

Could be. I may have found a temporary fix. What I did is, I open the Performance settings on Geforce Experience and lowered the Power Maximum slider. It used to be at 100% and the PC would crash with fans running at full speed when running a 3DMark Spy or a stress test, but the benchmark would run just fine with this value set to the minimum. As it turns out, the fan speed settings are ignored when the power drawn is too high for some reason (or it could be that increases the temperature and therefore the fans run at full speed no matter what, to prevent a crash). I'm not sure why this started happening but it is what it is.

 

So I played with those settings and 65% power maximum seems to be the sweet spot. I also set the temperature target to 85° and left the fan speed to Auto. As a result, when I play AoE 4 the power draw now maxes out at around 160W (around 175W in stress tests)  instead of 250+, the fan speed stays at around 2000RPM, the GPU speed is at around 1600-1650Hz, GPU temperature around 75°, and GPU memory speed at 7000MHz (sounds like a lot). Now the benchmark results for the GPU are about 5% less than the average 3070 but I'll take that over the crashes. I'll do some more testing but I think I'm onto something.