Sign in with
Sign up | Sign in
Your question

Serious GPU overheating

Last response: in Graphics & Displays
Share
April 17, 2011 10:46:41 PM

Over the last week or so my computer has experienced strange crashes during which my monitor went to sleep whilst my computer remained on but unresponsive though the fans became incredibly loud. At first I thought it might be the psu as it was running hot, however upon opening the case I found my gpu (ati 5770) was quite hot. I used gpu-z and catalyst to check the temperature and they both gave me a reading of 74c - 78c - despite the fact I was doing nothing more intensive than browsing the web. Twenty minutes after waking the computer up the temperature has climbed from the low forties to near seventy. I have tried giving the case and the gpu in particular a good clean but this did nothing to solve the problem.
My specs are;

processor: i5 750
gpu: ati hd 5770
Motherboard: gigabyte ga-p55m-ud2
Ram: 8gb
heatsink/cpu fan: Quiet nocturna nh-u12
psu: xilence power 600w

Does anyone have any idea what the problem could be? I'm at my wit's end here.

More about : gpu overheating

a c 241 U Graphics card
April 17, 2011 11:52:13 PM

That's hot for idle .... try turning off sleep ..... seems it drives lotsa puters crazy these days.

start up Task Manager and see what's using the CPU when the fans speed up....clock on the CPU columns twice to put the biggest users at top.
April 18, 2011 12:08:27 AM

Apparently catalyst control center and firefox are the biggest users, but my cpu usage is still only 2% at most. I'm not running much. Also, according to core temp, none of my cores are running hotter than 35c.
Related resources
April 18, 2011 12:50:43 AM

How is your motherboard temp? Are your cables neatly tucked away from your components especially fans? What is the ambient room temp? How are your voltages for your components running?
April 18, 2011 12:57:52 AM

Also, another test to see if you have a cooling issue is to pull off the side case cover, and put a house fan to blow air in your comp. If temps drop down say more than 4C then you have a case cooling issue.
April 18, 2011 1:06:01 AM

My cables are tucked away to the side. My room is fairly cool, my windows are open and earlier I even had my huge fan right by me - I was freezing but my graphics card was still on fire. Not sure about the motherboard temp, but when I took the case off it seemed pretty cool. I'm too sure about voltages either (forgive my ignorance) but speed fan reads out like this;

vcore1: 0.86v
vcore2: 1.58
+3.3v 3.39v
+5v 5.00v
+12v 1.28v
-12v -0.92v
-5v -8.78v
+5v 3.63v
vbat: 3.14v

Hope this helps.
April 18, 2011 1:12:12 AM

pocketdexter said:
Also, another test to see if you have a cooling issue is to pull off the side case cover, and put a house fan to blow air in your comp. If temps drop down say more than 4C then you have a case cooling issue.


I tried that earlier and I'm trying it again at the moment. Temperature drops from 76 to 72 and remains stable at that, despite the fan in my room being quite powerful and the fact that the gpu is reporting little to no load/activity. I doubt it's a cooling issue to do with my general set up as I've haven't changed a thing inside it for about a year, I've regularly dusted and it's never been anything like this hot in the past, except on particularly long and intense gaming sessions.
April 18, 2011 1:24:31 AM

gormenghast said:
My cables are tucked away to the side. My room is fairly cool, my windows are open and earlier I even had my huge fan right by me - I was freezing but my graphics card was still on fire. Not sure about the motherboard temp, but when I took the case off it seemed pretty cool. I'm too sure about voltages either (forgive my ignorance) but speed fan reads out like this;

vcore1: 0.86v
vcore2: 1.58
+3.3v 3.39v
+5v 5.00v
+12v 1.28v
-12v -0.92v
-5v -8.78v
+5v 3.63v
vbat: 3.14v

Hope this helps.


The 12 v rail is way too low. However, some programs do not read accurate voltages. Try checking voltages in BIOS. The rule of thumb for voltages is +/- 10% of each voltage is the acceptable tolerance.

12V = should be 12v or 12.2 not 13v, but most cheap PSU are under 12v
5v = 5v or more but should not reach 5.5v nor 4.8v (some psu do reach these voltages)
3.3v = 3.3 or more but should not reach 4v

Also make sure that your 12v rail has enough amps for your system. If say a gpu calls for a 28a 12 v rail, then make sure you have more amps than say 33 because other components need amps too like fans for example.
April 18, 2011 1:48:07 AM

You might be right about the 12v rail. The Xilence Power isn't terrible but it isn't that great either according to what I've heard. It should deliver about fifteen amps on either rail and I think the 5770 only requires about ten amps. Could the psu be overheating the gpu?

Also, how would I check the voltage in bios?
April 18, 2011 2:12:11 AM

gormenghast said:
You might be right about the 12v rail. The Xilence Power isn't terrible but it isn't that great either according to what I've heard. It should deliver about fifteen amps on either rail and I think the 5770 only requires about ten amps. Could the psu be overheating the gpu?

Also, how would I check the voltage in bios?


When you boot your machine, you get a screen with a logo. Somewhere on this screen should be some letters saying something like F2 Bios, Delete Setup, F12 boot or w/e. Locate it 1st then reboot (you probably wont make it in the menu the 1st time so use ctr+ alt +del) and press the appropriate key. A menu should appear. Look for something like hardware monitor that shows system temps/voltages. Report back on these findings.

If they 12v rail is low in BIOS then your psu is likely the culprit. The proper way of checking voltages is to use a multimeter, but lets keept it simple.

Next thing to do would be to use 2 12v rails to power your card and see if the same behavior occurs
a b U Graphics card
April 18, 2011 2:25:52 AM

Open up GPU-Z, goto the Sensors tab. Aside from temperature what you should look at are :

- GPU Core Clock
- GPU memory Clock
- GPU load
- video engine load


If you really are "Idle" the clock should be running much slower than the rated GPU card speeds. And the loads should be next to zero (0)......... Also check the Fan Speed, should be maybe 2000 RPM if card is running in the 70C range...

April 18, 2011 10:36:43 AM

pocketdexter said:
When you boot your machine, you get a screen with a logo. Somewhere on this screen should be some letters saying something like F2 Bios, Delete Setup, F12 boot or w/e. Locate it 1st then reboot (you probably wont make it in the menu the 1st time so use ctr+ alt +del) and press the appropriate key. A menu should appear. Look for something like hardware monitor that shows system temps/voltages. Report back on these findings.

If they 12v rail is low in BIOS then your psu is likely the culprit. The proper way of checking voltages is to use a multimeter, but lets keept it simple.

Next thing to do would be to use 2 12v rails to power your card and see if the same behavior occurs


According to bios

vcore 1.140v
ddr15v 1.584v
+5v 5.26v
+12v 12.239v
current system temp 35c
current cpu temp 32c

Screen went black again when I was in bios. What the hell is going on? :heink: 
April 18, 2011 10:41:34 AM

jb6684 said:
Open up GPU-Z, goto the Sensors tab. Aside from temperature what you should look at are :

- GPU Core Clock
- GPU memory Clock
- GPU load
- video engine load


If you really are "Idle" the clock should be running much slower than the rated GPU card speeds. And the loads should be next to zero (0)......... Also check the Fan Speed, should be maybe 2000 RPM if card is running in the 70C range...


- GPU Core Clock - 400 MHz
- GPU memory Clock - 900 Mhz
- GPU load - 1%
- video engine load - didn't see this in there.

As for rpm, gpu reports 2525 rpm. Gpu temp is currently around 76c - 80c and the pc has only been on for five minutes.
a c 88 U Graphics card
April 18, 2011 10:45:33 AM

ok those are not correct idle clocks for a 5770, they should be 157/300. 400/900 is for flash and stuff, you wouldn't have youtube open on the browser?

make sure the heatsink on the gpu isn't loose, sounds like it is not making proper contact with the gpu die.
If you take it off, re-apply the thermal paste because the old stuff wont work anymore once you break the bond.
April 18, 2011 11:12:48 AM

Kari said:
ok those are not correct idle clocks for a 5770, they should be 157/300. 400/900 is for flash and stuff, you wouldn't have youtube open on the browser?

make sure the heatsink on the gpu isn't loose, sounds like it is not making proper contact with the gpu die.
If you take it off, re-apply the thermal paste because the old stuff wont work anymore once you break the bond.


Closed firefox and clocks fell to 157/300. Temp has somewhat dropped but still quite high - fluctuating between high sixties and low seventies. Gpu reporting 0 - 1% load.

Would the heat sink for the gpu be inside the card itself?
a c 173 U Graphics card
April 18, 2011 1:28:27 PM

How clean is the card really? Did you just blast it and popped it back in? If not you need to go ahead and just pop the cooler aka take it off and replace the stock compound for arctic silver 5 or any thing better while cleaning the cooler of any trapped dust.
April 18, 2011 1:57:12 PM

nforce4max said:
How clean is the card really? Did you just blast it and popped it back in? If not you need to go ahead and just pop the cooler aka take it off and replace the stock compound for arctic silver 5 or any thing better while cleaning the cooler of any trapped dust.


Forgive my ignorance, but would this entail opening up the card itself? Is this more likely to be the problem than an issue with the psu or some other problem?

I'm sorry if I'm asking too many questions, this is just doing my head in.
a c 88 U Graphics card
April 18, 2011 5:35:27 PM

gormenghast said:
Would the heat sink for the gpu be inside the card itself?

it's the metal fin thingie beneath the fan/plastic shroud (depending on the model you have)
take alook at this
http://www.techpowerup.com/reviews/HIS/HD_5770/3.html and the next page
http://www.techpowerup.com/reviews/HIS/HD_5770/4.html
it shows the reference design 5770 disassembled

and heres one custom design
http://www.techpowerup.com/reviews/Sapphire/HD_5770_Vap... and the next page as well

the coolers are attached with screws thru the card. The important paste is the grey goo on the copper base of the cooler where it makes the contact with the gpu chip.


and psu wont cause the card to overheat, it just isn't possible. Either the fins are full of dust or the whole heatsink is loose (and the paste has dried up becoming more of a insulator rather than heatconductor)
April 18, 2011 6:19:37 PM

Ah, that makes a lot of sense! Thank you :) 

In the interim I have been exchanging emails with the fellows who bilt the rig for me. Rather than attempt to repair the card myself, they have agreed to send me a new card free of charge provided I send my current one back. I believe I'll take that option, as I'm clumsy and only liable to mess up trying to fix the card. Hopefully this is the end of my problem, but thank you all for your advice, it will doubtless come in handy in the future. :D 
a c 173 U Graphics card
April 18, 2011 6:22:58 PM

OK read and fallow this slowly. On the back side of the card there is a x clamp with 4 screws. Unscrew then and the clamp will fall away free from the card. The cooler may still cling to the card but twist back and forth then it to will separate. Then you will see the card as it really is and the back side of the cooler. Clean both and remove all the old nasty stock compound and with a thin piece of plastic apply a new but thin layer of fresh compound on the gpu die. There should be no spots ware the die can be seen except for the edges. Once that it done place the pcb aside and try to clean the cooler of any old compound as well dust. Put it all back together and temps should gradually decline over time by a few c once is is back together and installed.

edit: you posted before I could reply :pt1cable: 
a c 88 U Graphics card
April 18, 2011 6:28:46 PM

gormenghast said:
Ah, that makes a lot of sense! Thank you :) 

In the interim I have been exchanging emails with the fellows who bilt the rig for me. Rather than attempt to repair the card myself, they have agreed to send me a new card free of charge provided I send my current one back. I believe I'll take that option, as I'm clumsy and only liable to mess up trying to fix the card. Hopefully this is the end of my problem, but thank you all for your advice, it will doubtless come in handy in the future. :D 

that works too :D 
!