GPU temp concern

Anonymous
Topic 197303

I am running a GTX 770 standalone (no other GPUs) on Ubuntu 12.04 crunching for E@H only. This GPU is processing 3 WUs at a time while the CPUs are handling various E@H WUs. The temperature of this card is a consistent 79~80 C. The card is rated at 98 C. Does this seem high or is this normal for a GTX 770? Anyone else running a GTX 770? and if so what temps are you experiencing? Do I need to be concerned?

The case has 3 200mm fans (two bring in air and one exhausts air out the top of the case) and two 120mm exhaust fans (one on the power supply) so I believe that there is adequate cross ventilation.

Mumak
Joined: 26 Feb 13
Posts: 325
Credit: 3528130891
RAC: 1456661

GPU temp concern

My personal opinion is that this is a bit higher for a sustained long-term operation.
For example, my GF 660 Ti runs at ~65 C doing 2 tasks. My Dual HD7950's OCed to 1100 MHz run at similar temperatures (DP load), but I had to install 3x 140mm side fans blowing air straight at them.
You might try to change the default GPU fan profile to a custom one (use MSI AB or similar).

-----

WB8ILI
WB8ILI
Joined: 20 Feb 05
Posts: 45
Credit: 1110108013
RAC: 2780966

Robi - I have read 80C is

Robi -

I have read 80C is a safe temp although I like to keep mine as cool as possible.

I think this is your LINUX computer. If you haven't already done so, install NVIDIA-Settings to set your fan speed. I have NVIDIA-Settings-304 installed because I am using version 304 of the driver.

I believe my "default" fan speed is about 30% and won't increase until the temp gets pretty high. I set my fan speed to 70% to get the temp down to about 60C although I don't have the same card as you do.

The fan speed has to be reset each time you reboot.

ROBtheLIONHEART
ROBtheLIONHEART
Joined: 16 Aug 12
Posts: 47
Credit: 58199880
RAC: 0

robl,I run a 770 and even

robl,I run a 770 and even when I was testing 4x to see if I would get a bit more thruput my temp stayed below 56.My card is a Gigabyte oc with their 3 fan cooling system and my case has 2 120 intake fans in front at gpu level along with a 120 side intake aimed at the gpus (I also just put in a 780 in the same box the other day and it runs about 10-15 higher)along with a top 140 and a rear top 120 out. I use Evga precision X for gpu fan control set with a curve that puts the fan % at about the temp number ie 55 temp , fan speed 55% etc. I know X only works with Windows. Is there a similar program for Ubuntu ? I don't know maybe some one else does? IMO I agree with others that you need to get that temp down for the longevity of the card.

Anonymous

RE: Robi - I have read

Quote:

Robi -

I have read 80C is a safe temp although I like to keep mine as cool as possible.

I think this is your LINUX computer. If you haven't already done so, install NVIDIA-Settings to set your fan speed. I have NVIDIA-Settings-304 installed because I am using version 304 of the driver.

I am running driver version 331.20 on Ubuntu 12.04. Here is how to "add" the fan control functionality to your "nvidia-settings".

1. sudo nvidia-xconfig --cool-bits=4
2. reboot
3. launch nvidia-settings from command line or from the "Dash Home".
4. click on "Thermal Settings" for your GPU.
5. Note the "enable gpu fan settings checkbox" and the fan slider.
It will look like the following image:

You can see the effect of increasing the fan speed. Them temp is now around 60 C.

Quote:


I believe my "default" fan speed is about 30% and won't increase until the temp gets pretty high. I set my fan speed to 70% to get the temp down to about 60C although I don't have the same card as you do.

The fan speed has to be reset each time you reboot.

I do not believe there will be a need to do anything after a reboot with this procedure, but I have not done but the one reboot since installing.

Note: there is a "HUGE" disclaimer for using the function within nvidia-settings. Basically you void your warranty if you use this option/function. So user be ware.

I do run a perl script to acquire temps at a predfined periodicity and receive email should a preset temp value be exceeded.

Edit: I am thinking of modifying the perl script to do a shutdown if this predefined temp be violated should the fan on the card fail.

Anonymous

ROBtheLIONHEART, I think

ROBtheLIONHEART,

I think the big difference is your card with 3 fans. Mine has a single fan. As for your other case fan configuration it is very similar to mine. I might disable the side mount fan to see if it is in anyway interrupting the airflow over the 770 and effecting its fan's efficiency.

0
0
Joined: 11 Feb 05
Posts: 11
Credit: 47160013
RAC: 0

See:

See:

http://www.geforce.com/hardware/technology/gpu-boost-2/technology

80C is normal. This is the typical temperature target that the card will aim for.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 579153532
RAC: 203197

RE: 80C is normal. This is

Quote:
80C is normal. This is the typical temperature target that the card will aim for.


Exactly. Which means it's "safe" for occasional gaming during the warrenty period. But nVidia really doesn't expect you to number crunch 24/7 with it. Never mind all this GP-GPU talk they tell you to sell the cards.. they know full well that most of their cards are idle most of the time.

Anyway, as a rule of thumb every 10°C hotter halves the chip's lifetime. This doesn't mean the chip will fail too early at 80°C, but 60 is a significant difference. Personally I shoot for below 70°C, otherwise things would get too noise anyway.

BTW: your power consumption should go down a few W going from 80 to 60°C, unless the fan is getting loud (which would indicate it's starting to draw too much power itself).

MrS

Scanning for our furry friends since Jan 2002

Mumak
Joined: 26 Feb 13
Posts: 325
Credit: 3528130891
RAC: 1456661

RE: 80C is normal. This

Quote:

80C is normal. This is the typical temperature target that the card will aim for.

.. and for typical operation that card is designed to - gaming.
Not for 24/7 operation under high load.

-----

Anonymous

the temp is now oscillating

the temp is now oscillating at 59~60 C. To get this temp really puts the fan into 73% with an RPM of 3030. This cooler temp might "save" the chip but I am not sure how the fan's bearings will manage. If the fan fails then ....

Mac.teh.Knife
Mac.teh.Knife
Joined: 4 Feb 13
Posts: 17
Credit: 865586
RAC: 0

RE: the temp is now

Quote:
the temp is now oscillating at 59~60 C. To get this temp really puts the fan into 73% with an RPM of 3030. This cooler temp might "save" the chip but I am not sure how the fan's bearings will manage. If the fan fails then ....

It will run for a loooong time at that speed. If it fails then the card's thermal protection mechanism should shut the card down. It should but maybe it won't, those mechanisms have been known to fail. For additional protection you could run a script that would monitor temperature and if it gets above 80C and stays there for more than 3 secs and the fan speed is ridiculously low (as in bearing has ceased up) then the script could tell BOINC to suspend all GPU crunching. Additionally the script could send you an email to tell you the fan seems to have failed. I think someone else mentioned a Perl script similar to that in this thread? Check that out if fan failure concerns you, it would be well worth it for peace of mind alone.

You might be able to reduce the fan speed a fair bit by simply removing the side of the case. Lots of crunchers do it as a cheap but very effective way of reducing the air temp inside the case, the idea being it's very difficult to cool a hot thing by blowing hot air at it. Try it, you'll likely be pleasantly surprised. If you have curious cats then put a piece of window screen (cheap) over the open side but I know my cat stays away from my open case, seems he doesn't like the sounds he hears inside it. He's no sissy but he won't even poke his nose in there to quickly snatch his favourite treat if I put one in there.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.