Topic: How hot is too hot? (Read 1066 times)

Thats right. Cgminer and BFgminer can do this, too. But you have to set it

Nice one - boiling in waterblock... expensive cup of tea...
I thought miners can be set to not go beyond certain temperature. At least ufasoft miner can do this.

OK, I'll admit I ruined 2 Ati cards, 1 died at almost 2x stock clocks at 2.5V under dry ice, the other died at stock clocks when the water cooling failed and the fullcover waterblok reachted >boiling temperature for prolonged time (memory fried, artefacts..)

Anyway, mining bitcoin isn't very memory intensive (I hear LiteCoin is, but not sure..), so for Bitcoin, you could dial the memory right down and make sure that doesn't fry..

Yeah, I'd definitely agree with what folks say here and argue that you really want to keep it around 70-75. Overclock as much as you stably can while still staying in that range! I suppose it might be fun to actually Vmod a few cards and get back into the overclocking game...ah, goodness. Back in the day, it had to be done for its own reward -- now, there's a completely legitimate reason to do it! Exciting.

Well if you're mining bitcoins, with cgminer you can also very easily drop the RAM clock on it, that saves a LOT of thermal output. Bitcoin mining don't rely on memory as much as raw computational power. Personally I found a sweet spot at memory:300 and dynamic gpu_clock:600-800 so it can throttle if it does overheat, then I set the thermal ranges in cgminer to be 75 base target so it will clock up or return to normal if it's up to that, and 80 as the overheating where it will start to raise fan and throttle back, and 85 where it will cutoff and just take a break as it's overheat range. It's been working great.

Before cgminer had the really good thermal and clock code put in, last year I let it run and tried to clock with the regular program adjusters and I couldn't seem to keep the temperature down, so for a long while I had extra case fans and they still wanted to crank 85-95 range for a few months. I would say 90+ is bad, 80-90 is strenuous, and 75-80 is a fine range for a long few years lifespan.

adding posts. checking things out. Chillin. Peace.

Hmmm. I'm running 100C but fan is not 100% yet.

[Correction update] Sapphire Trixx tells me. 90C and 99% Fan.
GUIMiner-scrypt on 'low usage' setting at these temperatures. Man, what's up?

Thats basiclly how every nvidia card since the 8xxx series dies 2 weeks after warranty expires..

I mine at 68-71 and have had no problems

Some cards might handle 90c or higher, but that is dangerous to some, and depending on the setup can also affect other components (an open-air rig would be a different story however, and can run a bit hotter so long as the air vents away properly)


the other thing is, as i mentioned, rapid heating or cooling will do the most damage. running at 70c 24/7 is okay so long as the card isnt forcing 95%+ fanspeed (indicator that temps are VERY high), and a constant temp is fine for hardware. however, if after 24hrs you turned off the PC and walked away, the card could cool from 70c to 30c fairly fast, causing metal heatsinks and traces to flex every so slightly as they shift/contract. Same thing heating up from room temp to full operation too quickly. It can effectively open a solder point and disrupt the data/power flow of the entire chip.

everthing under 90C° is fine, look at the fan in automatic mode. If it spins 100% all the time, its to hot

Running my HD5870 at 80~85 degrees here.. Never had problems with ATi products running around those temperatures.. Will go back to watercooling soon, mostly for silence.

I've been running one of my cards 24/7 for a month at 75...no ill effects yet. I do need to ventilate better though.

Awesome, thank you for the reply man

70C is about as high as I would go.

Well, you are always "shortening its lifespan" if you use it, at least vs. not using it. But the point is: Will it last "long enough" for you? My guess is probably yes - unless you care about it still in 3-5 years or so from now, when it's hopefully outdated. 70°C chip temperature feels rather low for me, should be fine I guess.

So if I'm running my 7870 at 70C 24/7, am I shorting its lifespan? If so, by how much? It's my only gpu and I'd like to keep it for a bit while longer lol

Of course lower is better: Arrhenius' rule (google it) estimates roughly double lifetime for every 10K lower temperatures. 70°C sounds like a very sane, even conservative value for not-too-recent GPUs.

Background: For the silicon itself even over 100°C is no problem. For the interconnects (bumps between die and substrate), underfill, PCB etc. of course it is - but since you don't have an Nvidia card, a company which is infamous for f*ck-ups with heat related underfill/interconnect issues, your longevity might be OK. 

Finally, depending on who made your card, the quality of the components (mostly the electrolytic capacitors) and the related design (for how many working hours at what ambient temperature the voltage regulator design was calculated) will vary. Still, broken capacitors can be fixed much easier than a broken chip - so this should not be the biggest concern either. But still: Be aware that you are going to use for 24/7 a card that was most probably designed for at most 8 hours per workday, since it is a consumer product!

From my point of view, the biggest thing to look out for is thermal throttling: At some temperature your cards will start slowing down to avoid overheating.

To check this, keep an eye on mining speed as the temperatures go up. If you hit a thermal throttling barrier, you will see a sharp decline in speed.

To sum it up: Don't worry, most halfways modern GFX chips will typically take care of themselves not to get so hot that they are actually damaged. Just make sure you get the ebst performance by having them not throttling. Also, keeping the card cool will help with the lifetime of the capacitors - but these can be exchanged more or less easily in case of trouble.

most cards should not go above 70 degrees, or the risk of overheating or failure increases dramatically. 79xx series cards might handle 75 okay.

anything higher is straying into the danger zone, and may work for a while but pre-mature failure can be several times more likely at 80c vs 75c


also, rapid heating or cooling of the card can create issues. (ie: starting the miner immediately after turning on the system, or vice versa, without allowing a low amount of load to run for a while and bring the card to a moderate operating temperature (40-50c) and warm up before taking it to extremes, mostly due to thermal expansion being able to create gaps between the heatsinks and chips or having the chips flex away from the PCB and break connection.


I found I could run a 6850 at 63c and 95% fan for a consistent 200Mhash, and am not planning to push it much further in overclock unless the heat level stays the same or can be dropped (mem/voltage underclock)

I'm mining with a GTX570 (Since I already had that on hand - not planning to dump a bunch of money into this mining thing), and I run at about 77-80C overclocked to 850mhz. Shouldn't be a problem at this level.