Warning: Overheating Video Cards

[Larirawiel]

Hihihi, that's the reason why i love the FF14 benchmark as an instability test. Because it can spoil issues, which other tests don't.

And if the GPU overheats then replace the thermal paste/pad. If it still overheats then your air circulation in your computer case is very bad.



Cheers

[GrizzlyTank]

You should use GeForce control panel to set fps cap.

I have set it to 60 so my 3070 never goes above 60C and fans are quiet enough not to be audible through headphones.

Yeah always cap it to monitors cap in the GPU's setting directly. Had certain cases (Crash 4 ) where the LOADING SCREEN of all things would tell the gpu to go full force netting me 400 fps 100% gpu usage and massive coil whine...

[Larirawiel]

You should use GeForce control panel to set fps cap.

I have set it to 60 so my 3070 never goes above 60C and fans are quiet enough not to be audible through headphones.

You should cap the fps around 3 fps under the monitor refresh rate. So if you have a 60 hz monitor then set the fps cap to 57 fps. It is a reliable method to prevent tearing without enabling vsync.





Cheers

[Granyala]

I mean this is a lot of assuming;

2 regularly cleaned/maintained PCs run FFXIV fine at mid-low spec, install DT Update, video cards overheat at low-low spec.

Videocards don't overheat out of nothing, though what worries me is that you described a burnt smell.
That usually does not happen when a card dies to overheating (especially newer models that just clock themselves down).

You should assess the health power supply as well. It's possible that it was overtaxed, if it was a cheap one and doesn't have the proper protective circuitry.

Alternatively, limit the gpu's power level.

FPS cap is easier to implement for the average user and pretty much achieves the same result.
Provided you do not choose a resolution/FPS scenario that maxes out/overtaxes your card.

The thing people in these situations don't want to hear, is that there is simply no way for a properly cooled and well ventilated PC to overheat. Components are designed to run at 100% load for prolonged periods of time. The game is more taxing than it has ever been, yes but that is very much a user issue.

Technically: yes.
Practically: no.

I don't think monitoring the temperatures during full screen 3D applications is something the average user does. Also: repasting a card, while not difficult to do for the people so inclined is WELL beyond what is reasonable to expect of the average joe.

If a card turns non-functional because the manufacturer skimped on 2$ worth of heat transfer compound... that's not a user issue.

Yeah always cap it to monitors cap in the GPU's setting directly. Had certain cases (Crash 4 ) where the LOADING SCREEN of all things would tell the gpu to go full force netting me 400 fps 100% gpu usage and massive coil whine...

IIRC New Worlds from Amazon could actually REALLY fry GPUs due to no FPS limit in the menus. :'D

Did you monitor your GPU temperature before that including hot spot?
DT is demanding and sometimes can use 100% of 20 series GPU.

Not just 20 series. My 3080 is taxed 100% in 4K 60FPS and reaches Hot-Spot temperatures of 80°C on a custom loop setup.
Granted, it's summer and my room ambient is at 30°C as well but still.

FF-XIV now needs to be taken more seriously on the hardware front.

Except VRAM.... 4GB usage in Tuliyollal, lul.

[kaynide]

Videocards don't overheat out of nothing, though what worries me is that you described a burnt smell.
That usually does not happen when a card dies to overheating (especially newer models that just clock themselves down).

You should assess the health power supply as well. It's possible that it was overtaxed, if it was a cheap one and doesn't have the proper protective circuitry.

That's a good call;
We've had the two PCs since ARR launch and kept them in good condition, although obviously they're not tip top condition because wear and tear.

The actual power supply interior is a place I can't easily clean so it could very well have cat hair or whatever from 10+ years of use. I'll be looking into that. Thanks!

Not sure about the heat transfer compound as it was put together by a local PC shop...but with10 years good I'd like to think they did right, but I'm also leaning towards it doesn't matter how clean I think I'm being because there will always be random bits of hair/dust that I just couldn't see/notice.

[MXMoondoggie]

What you describe doesn't just happen out of nowhere, a game isn't going to make your components blow up and melt. Modern components are made in a way that they will downclock themselves if they begin to hit temps that are too high, it doesn't just keep getting hotter until it melts. This is more in the realms of an electrical fault. Even if you had one of those teeny tiny zoomer cases that are popular right now that cram everything into a tiny space it'll downclock or shut down long before it bursts into flames even in a hotbox like that. Locking framerate to your display in the settings is a good way to reduce heat as the GPU doesn't have to work as hard to maintain that framerate instead of going full power to give you 300fps for no reason. You can also undervolt with Afterburner or a program that allows you to do so if it's toasty. My 3080 in game running my 144hz display at a constant 144fps gets to about 65 and it's summer here so the room is pretty warm alone.

If your temps are out of control somethings either very old and full of dust or fixing framerate lower will fix it.

[Mr_Gyactus]

Then, my wife's overheated (She also has a similar 20xx series card, slightly better than mine). It hasn't burned out, but it got blazing hot and eventually blue screened.
Note: I updated mine to a 4060 and it's running fine.

RTX20 is a 2018-2019 video card. Not THAT OLD.

Probably you were using DLSS + high level of shadow softening + perhaps NO lod. In short the AI part was working like crazy, and a entry level GPU card has not enough tensor core.

My advice is to let geforce experience decide the best for you, add a FPS limit (always a good idea) and if it is still overheating remove the other shadows.

[angienessyo]

Always cap your frame rate through your GPU’s software. I have a MSI 3080 and have kept cool temps with an fps capped through the NVidia control panel. After seeing New World and Diablo 4 kill cards, it’s just a good practice to have.

[GrizzlyTank]

Always cap your frame rate through your GPU’s software. I have a MSI 3080 and have kept cool temps with an fps capped through the NVidia control panel. After seeing New World and Diablo 4 kill cards, it’s just a good practice to have.

In their defence however, some of those cards weren't properly attached so the pins for the power melted.

[Granyala]

The actual power supply interior is a place I can't easily clean so it could very well have cat hair or whatever from 10+ years of use. I'll be looking into that. Thanks!

DO NOT OPEN YOUR POWER SUPPLY unless you are an engineer and know what you are doing!
Even plugged out it has rather large capacitors and there is still a possibility of electric shock.

Random dust/hair will always be present and should not be the culprit. Even my computer has some dust bunnies in there.
As long as the fins of the coolers / intake filters aren't clogged and the fans are spinning appropriately, you are good.

There is the possibility of HTC degradation, in order to find out, grab Hardware info and Furmark. Stress the system for a while and observe the temperatures. Both CPU and GPU should stay below 80°C at all times.

I know modern stuff can reach 100°C safely but I have never been comfortable with that for extended use.

Hihihi, that's the reason why i love the FF14 benchmark as an instability test. Because it can spoil issues, which other tests don't.

Wouldn't Furmark be a lot easier? It can even test stability of power delivery by maxing out CPU & GPU at the same time.

RTX20 is a 2018-2019 video card. Not THAT OLD.

Probably you were using DLSS + high level of shadow softening + perhaps NO lod. In short the AI part was working like crazy, and a entry level GPU card has not enough tensor core.

My advice is to let geforce experience decide the best for you, add a FPS limit (always a good idea) and if it is still overheating remove the other shadows.

Again: a GPU should not overheat even if it is pegged at 100% use. Just employing arbitrary limits is a band-aid fix. He needs to assess the root cause of the instability.