Define "Safe" Temps...

[redsquirrel0249]

Specific overclocking details aside, what temps are "safe" for graphics cards?

 

This idea stemmed from asking whether or not I should run fans at idle and low temps. Then I asked myself "well, when do I need to start running fans?" So now my questions are:

1. How low should max temps be to have almost negligible effect on the life of the card? Obviously voltage unavoidably wears down a card, but is there a point where warmth begins to impact lifespan, and therefore a point under it I can shoot for?
2. If the previous question provides an answer, then I should be able to not run fans until nearing that threshold, right?

 

These are really the main questions. My other questions hinge on these.

[Princess Luna]

Each individual graphical processing unit through its TDP rating has a specific target temperature considered "safe/desirable".

[Syntaxvgm]

18 minutes ago, redsquirrel0249 said:

Specific overclocking details aside, what temps are "safe" for graphics cards?

 

This idea stemmed from asking whether or not I should run fans at idle and low temps. Then I asked myself "well, when do I need to start running fans?" So now my questions are:

1. How low should max temps be to have almost negligible effect on the life of the card? Obviously voltage unavoidably wears down a card, but is there a point where warmth begins to impact lifespan, and therefore a point under it I can shoot for?
2. If the previous question provides an answer, then I should be able to not run fans until nearing that threshold, right?

 

These are really the main questions. My other questions hinge on these.

It really depends on what you want and what card.
The number one problem I see with mining cards these days that aren't blower style is actually the fans going out. It used to seem to me that this was an exclusive feature of old gigabyte windforce cards, but now I see that with all of them, gigabyte having a lower rate than others. This is especially true with cards that have a fanless mode, they cheaped out because it wouldn't be spinning all the time I guess? We have a ton of mining rigs at the office, and out of the ones that aren't blower cards, seems about half have some sort of fan problem by now just a few months in. A good mix of different cards too. All ran quiet enough to be in office with, so not like really fast.

For temps on the card itself, it really depends. Some just fold under constant heat. I've seen some card have very low failure rates at 90+ all the time (R9 2xx reference cards, notably) but I've seen some other cards with surprisingly high failure rates at like 80 mining, and this may vary per the actual GPU if they're made differently, ie a bunch of I dont know 580s could be different depending how they're made I guess. There's more than just core temp I suppose, memory cooling and whatnot. But most failures there I see are solder joint, and can be fixed by reballing. This indicates to me that the constant stress is more of the problem, not cumulative damage from gaming at like moderately high temps or something.

If you run at or just above 70, I would not expect a problem with any card, aside from dead fans if you're not lucky.

This is all of course mining loads, which are much more harsh. The number one think I see that kills cards is simply very high temps constantly (24/7) under 100% load (synthetic) causing solder joints to fail, not chip failure of any kind.

TDLR you're probably ok if you're not mining, though you didn't give any temps.

[Cookybiscuit]

Unless you intend to use the card 24/7 or near-24/7 I wouldn't give any consideration to temperatures so long as they are under 90C.

 

You'll not be able to set up a fan curve like you describe as it'll just keep cycling the fans on and off even if you have a large hysteresis. Best is to decide what temperature you want to run at, have a conservative fan curve below that and a aggressive one above that. It'll need tweaking to get it right, considering ambient temperatures fluctuate, dust accumulates and different applications draw different amounts of power regardless of them all causing the card to be at "100%" usage. 

 

This is of course all assuming you care about noise, if not, just set a aggressive profile and be done with it.

[Windows7ge]

Personally I like to keep temps under 80C. 85C at the most. Just my preference for any long session. Overclocking wise look to keep temps below this figure. If you can keep them below this completely passively that's great but when overclocking you'll need some fan spin to keep air moving.

[redsquirrel0249]

1 hour ago, Princess Cadence said:

TDP

 I know this, but everything is relative. TDP temps perform fine, but my question is whether or not I can extend the life of the graphics card by running it at lower temps.

 

1 hour ago, Syntaxvgm said:

depends on what you want

I understand this as a general precept of overclocking, but I'm asking about extending the health of the graphics card as long as possible. That's interesting what you say about fan durability on cards, especially with the reputation of my card, EVGA's FTW cards, and other good cooling cards. Good to know about constant stress versus moderate gaming temperatures, but wouldn't a certain temperature of gaming induce "constant stress"? You mentioned certain GPUs will vary in performance based on how they're made, and I understand this, but I figured, even among similar heatsink designs (such as reference models) they would perform similarly against temperature loads, regardless of what silicon lottery they're given as far as clock speeds. Though, I suppose clock capability is intertwined with temperature resistance, right?

 

1 hour ago, Syntaxvgm said:

at or just above 70

Well my stock furmark temps are 64C, so I've got some headroom there.

 

1 hour ago, Cookybiscuit said:

wouldn't give any consideration to temperatures so long as they are under 90C

Yeah, I understand and have heard this general guideline, I'm just trying to maximize my margins. I built my case around good airflow, I've got 5 noctuas in a Meshify C for cryin out loud.

 

1 hour ago, Cookybiscuit said:

hysteresis

I set it to 2C, and it does fluctuate. I never notice the whole of my case's noise at all, though. Noise is not an issue given my setup, max fans are quiet. But I'm not going to run max fans all the time if it will shorten the life of my card faster than overclocking to high temps would. Though, I could rely on a fan mod later on (i.e. just ziptie noctuas to the heatsink), and in that case, focusing on temperature only, with an aggressive fan curve, I still want to understand what max temps will maximize the life of the card.

[Syntaxvgm]

14 minutes ago, redsquirrel0249 said:

 I know this, but everything is relative. TDP temps perform fine, but my question is whether or not I can extend the life of the graphics card by running it at lower temps.

 

I understand this as a general precept of overclocking, but I'm asking about extending the health of the graphics card as long as possible. That's interesting what you say about fan durability on cards, especially with the reputation of my card, EVGA's FTW cards, and other good cooling cards. Good to know about constant stress versus moderate gaming temperatures, but wouldn't a certain temperature of gaming induce "constant stress"? You mentioned certain GPUs will vary in performance based on how they're made, and I understand this, but I figured, even among similar heatsink designs (such as reference models) they would perform similarly against temperature loads, regardless of what silicon lottery they're given as far as clock speeds. Though, I suppose clock capability is intertwined with temperature resistance, right?

 

More about the PCB design  too

14 minutes ago, redsquirrel0249 said:

 

Well my stock furmark temps are 64C, so I've got some headroom there.

 

and stop right there, that's excellent.

[redsquirrel0249]

On 6/11/2018 at 1:34 PM, Syntaxvgm said:

stop right there

Yeah, was having a conversation in another thread that basically amounts to maxing out case fans and minimizing GPU fans. Which is simple enough, given relative cost to replace. I mean, the 64C in furmark WAS stock speeds, but it's still good.

[Christophe Corazza]

1 hour ago, redsquirrel0249 said:

Yeah, was having a conversation in another thread that basically amounts to maxing out case fans and minimizing GPU fans. Which is simple enough, given relative cost to replace. I mean, the 64C in furmark WAS stock speeds, but it's still good.

 

Indeed, if you max out your case fans and you reach "safe" temperatures at full load, I don't see any reason why you would run the GPU fans.

Obviously, you must bear the noise of a jetplane taking off, but if you're fine with that... just go for it