I keep all my GPUs at 72C or below. Most of them are at 72C because I feel this is the perfect balance between speed, noise and longevity of the GPU itself. If you think this is too hot let me know, I would like these GPUs to make it at least a couple more years. Most are GK104 and GM2xx.

Definitely NOT too hot!

Definitely NOT too hot!

What do you keep your GPUs at?

Definitely NOT too hot!

What do you keep your GPUs at?

I will let mine run up to 86c if no errors. If the hardware is good you'll find they can cope and they'll likely be redundant long before they fail.

Definitely NOT too hot!

What do you keep your GPUs at?

I will let mine run up to 86c if no errors. If the hardware is good you'll find they can cope and they'll likely be redundant long before they fail.

Ah I had a card that I ran at 80C all the time and it recently failed so I'm a bit more conscientious about that now. It started artifacting on screen and erroring out every WU I gave it.

Definitely NOT too hot!

What do you keep your GPUs at?

My triple Slot Palit Gamerock 1080 and Jetstream 1070 normally regulate to 65°C but I have underclocked them a little in order to maintain 55-60°C. Just because I run them 24/7 but want the cards to last very long even so.

80°C is the ceiling limit in my humble opinion as the temperature is not evenly distributed on the board and the sensor having a deviation from "reality". I therefore recommend <70°C for permanent load.

My Pascals cool down by respectable 5°C just from a 20MHz reduction already. So you don't lose a lot of speed because of that.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Definitely NOT too hot!

What do you keep your GPUs at?

I will let mine run up to 86c if no errors. If the hardware is good you'll find they can cope and they'll likely be redundant long before they fail.

Ah I had a card that I ran at 80C all the time and it recently failed so I'm a bit more conscientious about that now. It started artifacting on screen and erroring out every WU I gave it.

That maybe true but doesn't prove the card failed because it ran at 80c.

Temperature is only part of the equation. You can reach max overclock without reaching max temperature.

Ah I had a card that I ran at 80C all the time and it recently failed so I'm a bit more conscientious about that now. It started artifacting on screen and erroring out every WU I gave it.

Artifacts = sounds like this GPU needs some new reballing. Temperature certainly is a crucial factor for aging, however the capacitors are normally more sensitive to temperature than the chip.

Which one is affected if I may ask you this question?

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Ah I had a card that I ran at 80C all the time and it recently failed so I'm a bit more conscientious about that now. It started artifacting on screen and erroring out every WU I gave it.

Artifacts = sounds like this GPU needs some new reballing. Temperature certainly is a crucial factor for aging, however the capacitors are normally more sensitive to temperature than the chip.

Which one is affected if I may ask you this question?

Two actually. My friend's 980ti of which two of the 3 fans fell off while he was computing and the temps sat in the mid to high 80s for a few months. It recently whitescreened every time it got a load. He is RMAing that. And my 280x with an MSI cooler which has thermal pads on the RAM and VRMs and MOSFETs. Also what do you mean by new reballing?

An example for (home) reballing...
https://www.youtube.com/watch?v=yZzFdT5PQxw

Actually that is re-soldering of the graphics chip in case of artifacts.

If you dont have the equipment at home and some experience(!), you may bring the card to a professional repair shop. However it pays only for high end cards and admittedly the re-work does not last forever. Many cards fail again after yet another year. Probably an after effect of heating the other components during the reballing procedure as well.

So I would give the 980ti a try.. but I dont think the 280X is worth the repair cost. Having said this, the whitescreen is not neccessarily a contact error so the reballing may be fruitless anyway. Just taking my first steps with GPU repair, so maybe someone else can venture a better guess.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

An example for (home) reballing...
https://www.youtube.com/watch?v=yZzFdT5PQxw

If you dont have the equipment at home and some experience(!), you may bring the card to a professional repair shop. However it pays only for high end cards and admittedly the re-work does not last forever. Many cards fail again after yet another year.

So I would give the 980ti a try.. but I dont think the 280X is worth the repair cost. Having said this, the whitescreen is not neccessarily a contact error so the reballing may be fruitless anyway. Just taking my first steps with GPU repair, so maybe someone else can venture a better guess.

This is why I wish GPUs were socketed

Well ... and that is why Nvidia and AMD don't do that.
They want your money. Money. Money. ;-)

Having said this, purchasing new GPUs make sense anyway from an ecological way of view (power consumption). Aside from the games needing more speed every year. To qualify my above sarcastic Statement even more, the BGA provides a lot of pins, it would be difficult to arrange them otherwise.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.