24

Everest Ultimate is suddenly telling me that the CPU temperature (and core temps) for my E6850 Core 2 Duo is 72 degrees Celsius. When I stress-test the machine, the temp goes up to 91 degrees and the CPU actually throttles. System remains stable though.

For over a year now, my CPU has run very cool (40's) with a large commercial copper heatsink/fan that I bought separately.

To top it off, I removed the cover of the box and felt the cpu heatsink and it wasn't even warm.

Is there such a thing as a CPU temp sensor showing the wrong readings?

Any tips would help.

UPDATE #1

Temp is also just as high in BIOS. So that leads me to believe it's a CPU seating issue (even though I used thermal paste to seat it two years ago when I built the machine)

UPDATE #2 Well. I removed the heatsink and cleaned off the original thermal paste (which was somewhat crusty). I polished the surface, re-applied some new paste, and reseated the heat sink. After powering it up, there was no noticeable change in the temp - ideling at 74. Ran the stress test and it went up to 94 degrees before being 100% throttled. I let it sit at 94 degrees for 20 minutes straight and the computer didn't even flinch. I then immediately shut it off and opened the case and felt around. The heatsink was completely cold to the touch. Even the copper rods were cold. The area near contact with the CPU was slightly warm but not hot to touch.

Then I ran REALTEMP, which is supposedly more accurate and it told me the CPU was at 104 degrees. (LOL)

At this point, I'm thinking no doubt the cpu's sensor is wrong. Sidenote: the BIOS has the latest version so no option to flash there. Reverting hasn't been known to help from what I've read.

What angers me is the false temps force the CPU to artificially throttle from 3GHz down to 2GHz and my CPU fan is cranking at full force all the time.

Should I call intel and tell them to send me another E6850?

SOLUTION UPDATE

I switched the processor out with another one and got the same obscene temperatures with the new processor followed by a heatsink that was cool to touch. My suspicion in the heatsink was suddenly renewed. I swapped it out with the stock heatsink/fan and lo and behold the temperatures returned to the normal 35C-50C. Even though the thermal paste was visibly flattened out every time I removed it, it looks like the heatsink was still not pressing hard enough on the CPU to effectively conduct the heat. The heatsink is a Masscool 8Wa741, which screws into a standard position on a mount on the back of the MOBO. Only thing I can surmise after 2 years of use was that, over time, the heatsink pressure on the CPU gave way until the heat began to be ineffectively conducted.

Lessons learned:

  • Intel CPU's can run SUPER HOT (upwards of 95C) and still be stable.
  • Heatsink's need to be VERY firmly pressed against the CPU to conduct heat.
Improve this question
6
  • I wouldn't risk the sensor being wrong. The more time a CPU runs at hotter temperature the shorter life span it will have. 90c+ is pushing as it is close to the die temp of most modern cpus.
    – Tony
    Commented Oct 15, 2009 at 15:58
  • Isn't it possible to disable the throttling from within the BIOS?
    – sYnfo
    Commented Oct 16, 2009 at 12:53
  • And by the way, 104°C is indeed very close to melting point of whatever compound there is, inside the CPU, as alex suggested.
    – sYnfo
    Commented Oct 16, 2009 at 13:14
  • As far as I know, there is no way to disable CPU throttling. The only option is to force it into the slowest state. It looks like I'm going to have to do a CPU swap test, and if that proves the CPU sensor is broken I will ask intel to send me another one (intel CPU's have a 3 year warranty for boxed orders)
    – Matias Nino
    Commented Oct 16, 2009 at 17:31
  • solution found - see update
    – Matias Nino
    Commented Oct 28, 2009 at 5:07

7 Answers 7

Reset to default
17

If the fan is very cool and the processor is reporting very hot, I would first suspect the connection between the fan/heatsink and the processor package. Make sure the heatsink is properly seated with thermal paste.

Let's get some terminology straight from Lavalys' KB:

  • CPU Diode or Core temperature is the temperature measured across a temperature sensitive diode on the processor die. The most common implementation uses a temperature sensor part external to the processor with the CPU Thermal Diode Anode (+) and Cathode (-) connections to it.
  • CPU Temperature is the temperature measured by a remote temperature sensors part on the motherboard near the CPU.

It is possible for these temperature sensors to fail although I haven't seen it personally.

Try installing a different processor and see if the readings are still incorrect. This can help point to where the problem lies. If the readings are OK, I would suspect your processor, if they are off as well, then suspect the parts on the board.

Unfortunately, if it isn't the heatsink, there is no easy user fix other than replacing the offending part.

Improve this answer
4
  • Very good points, hanleyp. See Update #2 in the Q. I'm hoping to avoid having to swap out the processor to verify the deficiency. If I can be fairly certain it's the CPU and there's nothing I can do to prevent the throttling, I'll probably just get a new CPU.
    – Matias Nino
    Commented Oct 16, 2009 at 5:02
  • You could try to return or exchange your processor, but since it has been more than a year, your warranty may have expired (I don't know much about processor warranties). Swapping out a friend's processor temporarily would really help, though. The last thing you want to do is to buy a new processor only to find out it was the sensor on the motherboard that gave the false reading. What motherboard is it anyway? I will look to see if I can find their sensor info. You should be able to read them through the SMBus or Super I/O.
    – hanleyp
    Commented Oct 16, 2009 at 13:10
  • Intel's warranty is 3 years and its only been 2 so far. The motherboard is a Gigabyte p35-ds3p 1.1. As far as I know, the core/cpu temp sensing comes from the CPU itself, though I'm sure the value is processed by the BIOS at some point.
    – Matias Nino
    Commented Oct 16, 2009 at 17:24
  • solution found - see update
    – Matias Nino
    Commented Oct 28, 2009 at 5:07
10

It could be a bad sensor, but I wouldn't wager on it. Check layer of conduction paste, seems like that could be the problem.

Improve this answer
7
  • 2
    Or the fan/radiator set is not properly secured on the CPU anymore.
    – Snark
    Commented Oct 15, 2009 at 5:41
  • First thought I had was missing thermal paste. If the heatsink should feel warm to the touch pretty much any time the computer is on as 40c is 104f. If it doesn't there is a good chance the heatsink is not properly installed. +1
    – Tony
    Commented Oct 15, 2009 at 15:54
  • Is it possible for thermal paste to degrade over time? I used plenty of it when I seated the CPU 2 years go. I guess I could re-seat it and see what happens.
    – Matias Nino
    Commented Oct 15, 2009 at 17:26
  • It is possible. Also, I'd suggest taking a look at some of many How To Apply Thermal Paste tutorials on the web, if you haven't done so already.
    – sYnfo
    Commented Oct 15, 2009 at 17:59
  • Thanks for the tips sYnfo. Further test still indicate something wrong with the sensor. See update #2 in question.
    – Matias Nino
    Commented Oct 16, 2009 at 4:51
2

I had same overheating problem with intel i5-661 and ASUS P7H55M mobo. This has a push-fit heatsink and fan and one of the push-fit posts had come loose. PC was switching off at 93 celsius. Fixed push-fit retaining post for fan and normal operation at 46 celsius resumed.

Improve this answer
1

Once you have lm-sensors installed, make sure you run sensors-detect from command line, which will detect your sensors and give you a more thorough read out.

Another handy piece of code, is a coretemp script 

function coretemp {

   /usr/bin/clear;

   while : ; do 

      /usr/bin/sensors | /bin/grep ^Core | while read x; do 

         /usr/bin/printf '% .23s\n' "$x"; 

done; /bin/sleep 1 && /usr/bin/clear; done; }

Then you just type 'coretemp' into your terminal and it gives you realtime stats on each of your cores.

Improve this answer
1

I think your temperature sensor is wrong. As all third party software uses the same sensor of the board they all show the same result.

You already checked the heat sink and processor immediately after shutdown and found it's cool. Check once again to assure it.

If the wrong sensor boosts up the speed of the processor fan, the fan noise becomes uncomfortable. You can overcome this by installing SpeedFan. It can reduce the speed. For this to work correctly you should disable the automatic fan speed control of your CPU from the BIOS setup. After disabling it, reboot again and run SpeedFan. Reduce the speed to 50%. This will put your PC in a silent state.

You can also check if the CPU temperature rises while you reduce speed to 50%. If not, definitely the readings from the thermal sensors are wrong.

Improve this answer
0

while defective sensing components are not unheard of, they are certainly not common. I'd be looking at the mechanical connection between the heatsink and the cpu.

is it possible that you have too much heat transfer agent installed?

Improve this answer
1
  • I layed down a very thin film with a credit card. The heatsink is screw-in (not snap-in) so it's clamped down as hard as it goes. Besides, if the CPU was really running at 104C for over 20 minutes, the heat-sink should have melted and yet was cold to touch.
    – Matias Nino
    Commented Oct 17, 2009 at 6:05
-2

If your temperature reading is immediately over 50C or 60C in the BIOS from a cold boot, a sensor is definately defective somewhere. Even running without a heatsink from cold boot wouldn't get that hot that fast.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .