Topic: Whats the lifespan of a GPU? (Read 2166 times)

 They haven't existed long enough to be sure.

 For perspective though, I bought 6 x HD 7750 back in Litecoin GPU mining days - ballpark 5 years back I think?
 5 are still running abet all of them needed fan transplants (the Sapphire JUST had it's fan flake out late last month, the HIS cards only lasted a few months on the original fans).

 One of my pair of HD 7870 dates from the same timeframe (HIS Ice-Q super-blower model), still running fine.
 The other one I bought about a year back used, no clue what it had been used for before I bought it.

 ALL of these cards have spent all of their time in my ownership either doing cryptocoin mining, or running the Distributed.net RC5-72 client (which is a different crypto program that puts a VERY similar load on a card to most cryptocoin mining) on a 24/7 basis with the only downtime being for fan replacement, system getting moved, power outages, moved to a different machine, or something ELSE in the system (usually HD) died.

 If you keep the cards fairly cool, they should last a VERY long time.



 If you run the Milkyway project on BOINC, you can earn GRC on a Tahiti card pretty efficiently.

warranty +1 day

yes, from my experience thus far between the 6 GTX 1080tis i own now.. this is the case.. 40MH/s less per card on air... thus why i have been going back and forth on just building one larger watercooling rig, in the old days i used submersible aquarium pumps for pumps on my watercooled system, i had an old AMD computer i built gosh i forget how long ago.. i used a beer keg for the re-savoir for fluid and just sat the aquarium pump in the bottom of the keg pumping water into the computer and dumped back into the keg.. the keg metal surface acted as my heat exchanger (kinda like a radiator) and actually worked flawless for a lot of years for me tell watercooling became more popular and people manufactured radiators and pumps specific for watercooling...

One really nice thing about watercooling is the ability to move the heat out of the room more consistent.. when i first built my 1080mm external radiator rig i am currently using, i placed the radiator near a room suction vent for the house a/c system.. i ended up having to move it because the pressure from the suction on the vent would cause the fans to spend on the radiator when the a/c cycled on and off.. was a little annoying, because it would spin them at super high rpms on 1 or 2 of them if the suction caught them just right... causing fan noise basically... the positive to this was the a/c would literally suck the heat right off the radiator.. and the room would cool down in no time flat from the a/c..... you cant place the radiator next to the discharge, because if its to cold of air it will condensate on your waterlines... bad bad bad... but if you wanted to run a mining operation you could set the a/c to always fan on mode, then just place the radiator in front of the a/c suction vent.. because when the a/c would come on the water loop would drop down as low as the mid 20c

Some of my 2013 7970 are still running and mining every day.
Undervolted to 1000mV

You CAN see them mining with there Fury Friends
http://www.monitorig.com/#?panel=dashboard&id=1086db9224632d26671ab42df5ee1d92bf0187bbdd

In addition to newer cards I am still running some rigs with R9 290s (hair drier) on ETH  and R9 280X on ZEC that I purchased and ran in 2013 and into 2014 before I shut them down.

I never sold them, because I did not want to give them away, and then turned them on again when the second GPU round came about last year.

I had to RMA some 280Xs back in 2014. I have also had to replace some fans on the 280Xs.

I would say they are very reliable, if you don't push the limits, and keep temps reasonable. Its the heat that kills.
     

I have 750 ti's that have been running non stop overclocked under volted for four years and have not lost a card yet, but I have replaced fans

If I understood correctly, you are running on a watercooled system and getting 40MH/s per card more than with the common build? Interesting.

My gpu is working about three years. but the second fan is broken, i put a new fan.

Huh?

EU has a 2 year minimum limit of warranty on electronics.

There are also laws in some places of the world with a lower limit (one year), like Europe.

This is actually why i have been contemplating going to watercooling for my future mining rigs, granted its an extra cost upfront.. but i think the extra cost is worth the better cooling.. I have been mining going on 2 weeks with my gaming pc on my 2 x 1080ti's and after finally getting around to learning how to mod them, im running at roughly 68% of TDP power on both and getting roughly 40Mh/s less per card compared to completely stock hashrates... and lowered my water loop temp from 45c to 38c now..

Just to throw more into the idea.. my friend bought his GPU with me when i ordered the gt1080ti's he is on fan cooling and running the same settings as me, he redurced fan noise, fan speed, and heat by a lot.. but his gpu is producing about 90Mh/s less than stock compared to my 40MH/s drop... so somehow there is a 50Mh/s drop between his setup and mine running the same miner, same settings.... that 50Mh/s will add up over time and multiplied per card...

My old days of watercooling are starting to pop back in my head.. like 10 years ago i watercooled an AMD dual core system i had using a 10 gallon beer keg with aquarium pump submerged inside cooling my system.. i used no fans just the heat transfer from the metal keg and was able to overclock with stable numbers for a lot of years on that system... thinking of bringing back the hack days of watercooling back and build a larger watercooling system with enough pump volume to flow like 30-50 cards constantly... when you think about it.. when you have 30 GPU's typically your gonna have like 90 fans spinning.. the power draw of those fans is probably more or equal the amount of a watercooling system with a commercial style pump..

I see no one has yet mentioned another very important component besides GPU core temps and fans,
I'm talking about VRM - build quality and temperature,

we all know various 3rd party card manufacturers have different designs for their graphic cards, they put together different quality components such as heatsink/vents, memory, VRM, phase number, chokes, FETs and lots of other electronic components I'm not aware of...

I understand some manufacturers use worse VRMs and some use better, to my knowledge better ones are rated at 125°C and cheaper at 105°C
Generally when buying a graphic card for mining you want your card made of best quality components or at least as much higher quality components as possible(at least that's what I'm going for as a top priority when picking cards after I decided which gpu I want)

I really wish 3rd party manufacturers would share at least what electronic components they used when assembling their cards instead of aggressively marketing we have this much core/mem speed, that much better boost in performance and such, no other way than buying a card and stripping everything down attached to it
miners besides overall efficiency in gpu chip care a great deal about card durability...

for a specific gpu chip, let's say, Hawaii 390 model
Not gonna talk about performance but just life durability,

It's a big difference deciding if you wanna get Sapphire's or Vtx3d's 390 card
of course the price is also different but not always necessarily by a lot.

I used sapphire/vtdx3d comparison because Sapphire is one of the most experienced amd card manufacturers and well known as a good quality card manufacturer alongside msi, asus and probably some others(I wouldn't know for sure)
vtx3d's main selling point was that they were cheaper choice/variant of amd cards manufacturer, they went bankrupt and were bought by powercolor few years ago.

In general cheaper components means less durable card but as xleejohnx has already said it, sometimes it depends on your luck,
you can get e.g. 1 sapphire card out of 1 000 cards that is crappy built, badly manufactured, it happens...
of course you can get a good vtx3d card too which will outlast average sapphire's 390 lifespan

back to temps,
I try my best to keep my gpu cores at least below 70°C if not around 60°C or below for 24/7 workload
Since I'm in mediterranean climate it's not uncommon to see during daytime core temps jump to 68°C~69°C even when underclocked(note that I have power hungry sapphire 390/390x cards in my rigs) and it's not even summer yet.

My VRM on these cards usually go around 10°C above the gpu core temp, which is fine considering they have higher temp rating than core.

It's a shame that 390(x) are so far the last generation of amd cards that have VRM temp sensors included, since Fury gpu-z doesn't show anything else but core temps
Anyone had good ol' 5970? that beast had 3 sensors per core and not even mentioning vrm sensors

Avg lifespan if I were to guess would probably be about what the manufacturers warranty gives you - they aren't dummies

until someone makes a tahiti coin that features an algo which  only mines on amd tahiti core gpu's.

p.s. can someone please do this?

i would think as long as you get 2-3 years out of them your golden.. because by then new technology cards will be 2-3x faster probably and the mining difficulty will increase also making the older cards less and less efficient

depends on your definition of lifespan.. undervolted and modestly overclocked they will never shut down or display artifacts on you under reasonable conditions for 5+ years I would say.

of course they dont make things like they used to so who knows if component selection/quality control is slipping over the years.

if they are purely on mining duty I could possibly forsee adjusting voltage and clock rates/intensity well into a decade of mining lifespan.

If you treat it right, you'll get a couple of years out of it before the fans start getting noisy.  Luckily, most of the newer cards have easier fans to replace than in previous years.

with a decent temperature a very long time, the temperature is a very important factor, the lower the better lifespan i owuld keep under 70 or even 60°, but usually every gpu should last at least two years

Awesome, thanks for the replies!

With a decent power supply, no overvoltage and decent temperature limit (~70-75°C) it's probably at least 4-5 years I guess. The fans will give in in that time though in a dusty enironment.