well my rx 580 is 52C and 1080 ti is 61C on air the 2 1080's on water are 46 or so, I'd aim for 75C at most 24/7 long long term

What you have to look for is safe operating temps for cheap components installed on the board. These are typically capacitors with 90C rated and will have drastically shorter life span the longer they operate over that. But, keep in mind that is for the temp of the component in question and not the gpu core itself. You could have core under 80C and some parts of that board could be over 90C.

In my honest opinion the best thing to do is just have excessive amounts of airflow over the cards, not only to keep the gpu core temps down but also all the other parts of the board as well. ( same reason why servers are so load, the have lots of air flow )

Edit: oh and gamers nexus on YouTube has some videos on this topic, i would give them a watch.

I would try and keep AMD gpus under 60c and Nvidias under 70c.  Control the mining so that your cards last.  Flat out, they will last 8-10 months, fans usually go first and if your fail safe's aren't set the whole card will go poof..  The temp sensor on most cards is on GPU chip, not memory or other components.

Agreed. Seems under powering them makes sense with custom profile.

Have any good suggestions for a 1070 or 970? I've heard this too, but I'd rather go with some more trustworthy sources.

Oh, it didn't quote the post about "undervolting"

Here's a message I received from EVGA when I emailed tech support about this:
-----------
Hello,
We do guarantee the functionality of the cards for the 3 year warranty. Obviously running the cards 24/7 will put the cards under more stress, and might lower the lifespan of the cards. In terms of temperature, the cards can reach up to 90c and still be within the thermal limit. so there should be no issues running the cards in the 70s.
Hope this information helps.
Regards,
EVGA Support 
----------
I'm mining on Nicehash with a variety of cards... I've got all my power limits maxed out in MSI Afterburner, so anywhere from 104-120% TDP...
Currently, my warmest cards run around 73C, but its winter... in summer, they'll get warmer... for now, I've decided I will limit them to 79C when the time comes... we'll see if i change my mind then, depending on profitability, GPU availability, fan noise, etc...
Edit: Right now, I keep my fans in the 60% range, and will limit them to no more than 80% 24/7... I've heard running 100% kills them faster too, though whether that true...

Tech's name was Gabriel Talamantes
 

MacII
Here's a message I received from EVGA when I emailed tech support about this:
-----------
Hello,
We do guarantee the functionality of the cards for the 3 year warranty. Obviously running the cards 24/7 will put the cards under more stress, and might lower the lifespan of the cards. In terms of temperature, the cards can reach up to 90c and still be within the thermal limit. so there should be no issues running the cards in the 70s.
Hope this information helps.
Regards,
EVGA Support 
----------
I'm mining on Nicehash with a variety of cards... I've got all my power limits maxed out in MSI Afterburner, so anywhere from 104-120% TDP...
Currently, my warmest cards run around 73C, but its winter... in summer, they'll get warmer... for now, I've decided I will limit them to 79C when the time comes... we'll see if i change my mind then, depending on profitability, GPU availability, fan noise, etc...
Edit: Right now, I keep my fans in the 60% range, and will limit them to no more than 80% 24/7... I've heard running 100% kills them faster too, though whether that true...

have you ever tried lowering your power limit? most lower cause the hash speed vs watt usage is a great factor. I know people as low as 65%. I personally run 70- 90% max and get good profits

I normally aim for about 70C on cards in an air-conditioned room (where my "setup and test bench" is at) as that will tend to keep them under 80C even in the summer in my "production mining room", for better longevity.
I have NEVER seen a card designed to run a 95C, though some of the AMD R9 2xx/3xx series cards had a thermal limit a 94C (and did NOT like running at close to that limit).
 
On the Nvidia side, almost ALL of my cards run at about 60% of factory TDP (the exceptions are the 1050/1050 ti cards I have that CAN'T be set lower than 70%, and some 1070 models with the "standard" 151 watt TDP that end up close to 70%) as they are more efficient at that range, which lets me pack more cards AND MORE HASHRATE into my existing electric supply limits AND helps keep them running cool and reliable for a long time at lower fan speeds (which help the FANS last a long time).
 
Fan wear is not linear with speed - it's a curve where the wear gets noticeably worse faster than the speed does.
On the other side, fan AIRFLOW is also a curve, where the last 10% gives very very little more airflow.
There is no real point to running most fans at 100% or even higher than 80% - wear goes up a LOT for small gains in cooling airflow.
 
I also refuse to buy any more cards that do NOT have ball bearing fans (not a factor on EVGA cards, they're ALL ball bearing) due to too many of the fans dying way too soon even at fairly low fan settings.
I MIGHT make an exception for a REALLY good deal on a "refub card" with the deliberate intent of replacing the fans VERY early on, figuring on the cost of the ball-bearing replacement fans being covered by the savings on the card itself.