Why do my GPU's get so hot when apparently not in use ?

jives11

New Member
Joined
Oct 25, 2023
Messages
7
Reaction score
3
Credits
70
Hi,

First off thank you for having me .
I have a general hardware question , and this is more an "out of interest" question, but I've just resurrected an old 775dual-vsta based PC and installed Peppermint 11 on it. I plan to use it for various command line activities, maybe some FAH as well. Just a bit of fun.
The PC has a PCI-E slot and several PCI slots too. It has a couple of Graphics cards, one AMD radeon in the PCI-E and a Geforce 8500 Nvidia PCI (not express). Fine, I'm running it headless and have disabled the graphical interface, so just a command line login. No X windows or window managers are running. I access the machine vis ssh, with the occasional vnc session.

if I look at the processes, nothing running to do with X windows etc . All good.
So I installed lm-sensors and when I run the sensors command , after 15 minutes uptime it reports that one GPU is at 45 degrees (Radeon PCI-E) and the nVidia PCI is at 65 degrees. Now both are passively cooled , so have no onboard GPU fan, but I'm kind of surprised that they are getting so hot while seemingly running no load. I'm running FAH on the core and that is cooler than the second GPU at 100% load. OK it has a fan but I'm wondering what these cards are doing ? is this just the heat dissipation of having them powered ? they are old and maybe the grease between the GPU and the heatsink needs redoing , but that's pretty hot

This is the output from the sensors command after 30 minutes uptime. FAH is running 100% CPU, but is not using either GPU (I actually would love it to, but it's not seing these old GPU's - too old)
truncated output from $ sensors command

w83697hf-isa-0290
Adapter: ISA adapter
fan1: 1493 RPM (min = 84375 RPM, div = 8) ALARM
fan2: 1985 RPM (min = 1318 RPM, div = 8)
temp1: +33.0°C (high = +0.0°C, hyst = +0.0°C) ALARM sensor = thermistor
temp2: +38.0°C (high = +80.0°C, hyst = +75.0°C) sensor = thermistor

nouveau-pci-0300
Adapter: PCI adapter

GPU core: 1.32 V (min = +1.20 V, max = +1.32 V)
temp1: +64.0°C (high = +95.0°C, hyst = +3.0°C)
(crit = +125.0°C, hyst = +2.0°C)
(emerg = +130.0°C, hyst = +10.0°C)

radeon-pci-0200
Adapter: PCI adapter
temp1: +46.0°C (crit = +120.0°C, hyst = +90.0°C)
coretemp-isa-0000
Adapter: ISA adapter
Core 0: +56.0°C (high = +78.0°C, crit = +100.0°C)
Core 1: +60.0°C (high = +78.0°C, crit = +100.0°C)

Thanks, I'm happy with how it all runs , just curious as to what is happening here ?
 


The temperatures from the outputs shown in post #1 indicate that they are well within normal limits and there is plenty of leeway still there, which means that things are not abnormal at this point.

The 64 degrees C is well within it's designated range that runs to 95 degrees C, and doesn't become critical until 125 degrees C.

The 46 degrees C likewise, doesn't become critical until 120 degrees C.

The way to assist cooling for passively cooled graphics cards, if necessary, is to install another fan or two in the box. Fans vary in size, speed and airflow, but on the figures shown, there is nothing to be concerned about at this point in my view.
 
The temperatures from the outputs shown in post #1 indicate that they are well within normal limits and there is plenty of leeway still there, which means that things are not abnormal at this point.

The 64 degrees C is well within it's designated range that runs to 95 degrees C, and doesn't become critical until 125 degrees C.

The 46 degrees C likewise, doesn't become critical until 120 degrees C.

The way to assist cooling for passively cooled graphics cards, if necessary, is to install another fan or two in the box. Fans vary in size, speed and airflow, but on the figures shown, there is nothing to be concerned about at this point in my view.
Thank you, It's good to know that the GPU's are within their temperature envelope, but my question was more , why are they getting this hot when they are idle ? No monitor is plugged into either, nor is any form of gui running on them, just a text only login on the primary one, I have disabled the gui entirely ( $ sudo systemctl set-default multi-user).

Is this heating just caused by come current being drawn through them by virtue of being connected to the PCI, some aspect of the kernel interrogating the PCI system, or perhaps they are doing some processing, though I don't believe so
 
Maybe bad airflow in your system, my Nvidia gpu 27C when idle.
 
G'day @jives11 and welcome to linux.org :)

...but I've just resurrected an old 775dual-vsta based PC

Maybe a silly question, but what about dust? Did you go inside and give it a clean?

Cheers

Chris Turner
wizardfromoz
 
Maybe bad airflow in your system, my Nvidia gpu 27C when idle.
Yes - certainly, both these cards are passively cooled, so have a large heatsink, and its quite possible that the case fan is insufficient. I'm just surprised they get hot at all ?
 
Thank you, It's good to know that the GPU's are within their temperature envelope, but my question was more , why are they getting this hot when they are idle ? No monitor is plugged into either, nor is any form of gui running on them, just a text only login on the primary one, I have disabled the gui entirely ( $ sudo systemctl set-default multi-user).

Is this heating just caused by come current being drawn through them by virtue of being connected to the PCI, some aspect of the kernel interrogating the PCI system, or perhaps they are doing some processing, though I don't believe so
The graphics cards heat up simply because they draw the amount of power which keeps them ready, when idle, for any demand that is put upon them. It's a matter of "initialisation" so to speak. It's a feature, not a bug, as the saying goes.
 
Last edited:
Maybe bad airflow in your system, my Nvidia gpu 27C when idle.
You didn't mention whether it's passively cooled, or fan cooled. It makes a difference. And there's also the box itself with the numbers and relative locations of the various elements, the fans, the cards, the ventilation apertures.
 
You didn't mention whether it's passively cooled, or fan cooled. It makes a difference.
Fan cooled. I can't remembered what the temps were of the last passively cooled gpu that I had.
 
Fan cooled. I can't remembered what the temps were of the last passively cooled gpu that I had.
The OP has passive cooling. On this machine, also passive cooling, there is a similar temperature:
Code:
nouveau-pci-0100
Adapter: PCI adapter
GPU core:    900.00 mV (min =  +0.85 V, max =  +1.00 V)
temp1:        +49.0°C  (high = +95.0°C, hyst =  +3.0°C)
                       (crit = +105.0°C, hyst =  +5.0°C)
                       (emerg = +135.0°C, hyst =  +5.0°C)

Just checked on second machine with passively cooled graphics card and it's similar again.
 
Hmm...

I thought Folding@Home could do computations on a modern GPU, not just unused CPU cycles.

And I looked...

It can, my problem is the GPU's I have are far from modern , and mine are too old , or my motherboard too old, for FAH to detect . No matter. The machine is near silent and running FAH nicely, while consuming around 95 watts , most of which ends up as heat in the room. My next project will be something bigger.
 
I'm in much the same boat here.

I run an Asus GeForce GT 710 (the one the gamers all love to take the p**s out of!) This is a relatively new card, despite the GPU itself being an old, Kepler-based design; Asus have kitted this one out with GDDR5, as opposed to the original GDDR3, and have outfitted it with 4 HDMI ports.

VRAM is a "mighty" 2GB, and yes.....it's a passive-cooler. I didn't buy this for gaming, anyway; I wanted it primarily for video-rendering, since both Openshot AND Flowblade will offload to the GPU if, as & when they detect one. This is something this card is actually pretty good at, and it also lets me run the FlightGear flight simulator. It's far more capable than most folks would give it credit for...

I went for this because of its low power draw.....all of 19W, through the PCI-e slot itself. This HP Pavilion desktop rig has been fitted by HP with a weird, custom, low-power slimline PSU of just 180W output - pretty well impossible to replace! - so I have to watch total consumption quite closely. Of course, with newer hardware we're slowly moving away from the need for large, high-power PSUs that generate a ton of heat, due mainly to the fact that modern gear is all now deliberately designed specifically to be low-power, and to draw as little as possible.

~~~~~~~~~~~~~~~~~~~~~~~

This thing usually starts out at around 26-28C when it first fires-up, and over the next 15-20 mins will rise to around 42C-48C on average, though it doesn't ever seem to rise very much higher, even when under full load. I think the highest I've ever seen reported was just under 60C on one occasion, and it's often "at max" around 54-56C.

I think "passive-coolers" are all like this, and tend to run that much warmer even at idle because there's no forced-air cooling; all they do is to radiate the heat away slowly & gently. Frankly, I prefer it like this, because the temps are fairly predictable.......and you don't have to worry about temps going "leery" without warning because a fan suddenly decides it's had enough, or a connector works loose, or whatever....

I'm happy with mine.


Mike. ;)
 
Last edited:
I'm in much the same boat here.

I run an Asus GeForce GT 710 (the one the gamers all love to take the p**s out of!) This is a relatively new card, despite the GPU itself being an old, Kepler-based design; Asus have kitted this one out with GDDR5, as opposed to the original GDDR3, and have outfitted it with 4 HDMI ports.

VRAM is a "mighty" 2GB, and yes.....it's a passive-cooler. I didn't buy this for gaming, anyway; I wanted it primarily for video-rendering, since both Openshot AND Flowblade will offload to the GPU if, as & when they detect one. This is something this card is actually pretty good at, and it also lets me run the FlightGear flight simulator. It's far more capable than most folks would give it credit for...

I went for this because of its low power draw.....all of 19W, through the PCI-e slot itself. This HP Pavilion desktop rig has been fitted by HP with a weird, custom, low-power slimline PSU of just 180W output - pretty well impossible to replace! - so I have to watch total consumption quite closely. Of course, with newer hardware we're slowly moving away from the need for large, high-power PSUs that generate a ton of heat, due mainly to the fact that modern gear is all now deliberately designed specifically to be low-power, and to draw as little as possible.

~~~~~~~~~~~~~~~~~~~~~~~

This thing usually starts out at around 26-28C when it first fires-up, and over the next 15-20 mins will rise to around 42C-48C on average, though it doesn't ever seem to rise very much higher, even when under full load. I think the highest I've ever seen reported was just under 60C on one occasion, and it's often "at max" around 54-56C.

I think "passive-coolers" are all like this, and tend to run that much warmer even at idle because there's no forced-air cooling; all they do is to radiate the heat away slowly & gently. Frankly, I prefer it like this, because the temps are fairly predictable.......and you don't have to worry about temps going "leery" without warning because a fan suddenly decides it's had enough, or a connector works loose, or whatever....

I'm happy with mine.


Mike. ;)
Thank you Mike , yes I think it's a combination of the board being powered and the cooling characteristics of passive coolers
 

Staff online

Members online


Latest posts

Top