What temperatures do you keep your GPUs at ?

[bulanula]

I am referring to longevity.

Is it OK if they are at about 70 degrees Celsius 24/7 or should I increase the fan speed to keep them at 60 degrees or lower ( fan is less expensive if it dies compared to GPU core electromigrating ).

At what temperatures would you consider running your cards 24/7 as a dedicated miner ?

What temperature is needed to avoid electromigration ?

Thank you !

[malevolent]

I am referring to longevity.
Is it OK if they are at about 70 degrees Celsius 24/7 or should I increase the fan speed to keep them at 60 degrees or lower ( fan is less expensive if it dies compared to GPU core electromigrating ).
At what temperatures would you consider running your cards 24/7 as a dedicated miner ?
What temperature is needed to avoid electromigration ?
Thank you !

I don't think it is possible avoid electromigration but it doesn't mean the GPU will die, it may just operate only on stock clocks and voltages if you are unlucky.
For me it is <100C for VRM (all are <90C), <90C for rest (everything is <85C), except GPU, none of mine exceed 80C. But VRMs can sustain 105-140C and GPUs around 100-110C. But these temperatures will affect longevity

[DeathAndTaxes]

Electromigration can't be stopped.  Every single silicon product every built WILL (as in absolute certainty) fail from electromigration.

I don't think there is any specific temp -> life correlation but we do know some general ideas:
1) All Silicon devices will die due to electomigration
2) Lower temps and wattage will slow the effect.
3) The relationship between temp/wattage & electromigration is non-linear.

What does this mean practically for me?
I keep GPU <80C.  I don't overvolt.  I have reduced my overclocks somewhat to keep wattage down. I keep fan at a constant 85% (hoping constant speed and not running at 100% will make them last longer).

Is that enough?  I don't know.  I would be interested to see what others do.  Maybe I should make a poll. Poll made see below for link.

[Chefnet]

do the same normally <75°C and look for low fan speeds and low wattage (better MH/W)

[DeathAndTaxes]

I made a poll because I am interested in seeing what the consensus is:

Vote here: https://bitcointalk.org/index.php?topic=62814

[Chefnet]

just voted

[BkkCoins]

I use auto-fan and set a target temp of 72C and fan range of 40%-85%. This means my fans vary from 45% (coolest night) to 85% (worst afternoon), but seem to usually run around 55-60%.

My worst card (aka "The Dud") heats up to 75C even when it's fan hits 85%.

[_Vince_]

I have the GPUs at 70o and adjust the fan automatically with a control script

Max acceptable fan speed for me is about 70%; more than that level it is very noisy and I can't stand it. Besides, I also afraid that high speed could kill the fans quickly

Ambient temp is around 25oC at night to 33oC noon.

[bulanula]

I have the GPUs at 70o and adjust the fan automatically with a control script

Max acceptable fan speed for me is about 70%; more than that level it is very noisy and I can't stand it. Besides, I also afraid that high speed could kill the fans quickly

Ambient temp is around 25oC at night to 33oC noon.

Yeah exactly what I am doing as well but I was wondering if it would be better to raise the fan speed so as to bring down the temps to 60 degrees constant instead of 70 ATM. I wonder what the VRMs are doing at that temp !

In Linux using radeonvolt I don't think I am getting the VRM temps because they are too similar compared to the core temps so I think it is getting something else.

[aistto]

my rig in cold basement.
Now i don't now exactly temp (i think 50-60 C). But in summer I try to set it on 60-70 C.

[m3sSh3aD]

I have 2 rigs with 4x 5850's in each and i have them running 100% Fan, Overclocked and overvolted (965/300 @ 1.187V) to get 400+ m/hash a card. These run with VRM's just below 100 (below 110 is fine with me and 120 is my personal limit but lower is better) and no Core goes over 80 with 90 being my limit for GPU. I got a gaming PC which creeps over 90 when mining cause its 2x 5870's in a case where my rigs are open framed (2x tables ). My personal limits i like to stay in though:-

GPU <90
VRM <110

After them i just feel like i'm putting the card under too much with heat & voltage. Although i know my cards are good for 1000/300 @ 1.250V with sufficent cooling solutions. But i play it 'safe' and am happy with 400MH per card, 1000 only bumps it to 420MH a card (5%) and needs a considerable Voltage boost to achieve this.

[minerops]

BUMP! Very happy to be up and mining profitably again. I'm at 90C on two 6870s constantly

[trinsic]

80 or lower on 7950s.