LinuxQuestions.org

LinuxQuestions.org (http://www.linuxquestions.org/questions/index.php)
-   Linux - Hardware (http://www.linuxquestions.org/questions/forumdisplay.php?f=18)
-   -   Is my xeon really running at 82C? (http://www.linuxquestions.org/questions/showthread.php?t=4175461070)

xmrkite 05-07-2013 04:02 PM

Is my xeon really running at 82C?
 
I have an x3220 intel xeon cpu in a case. It's a decent case with a nice fan blowing out the back. The air is not hot coming out the back.

I have lm sensors installed and it reports my cpu is at 81 or 82c with about 65% cpu usage. When pretty much idle it lowers down to 45c.

Could it really be this high? Is there another way to check or verify this? It uses a no-frills intel server motherboard that gives no temp information.

I'm running lubuntu 12.10 on it.

It's not hot at all, infact today is pretty cold and rainy. The hard drives report 38c, and that's two of them, one on top of the other.

Beryllos 05-07-2013 07:19 PM

First, let me say I have no experience in CPU thermal management, short of installing a couple of fan heat sinks with white thermal paste on Pentium I processors clocking at 166 MHz, but I found this:
http://www.intel.com/support/process...s-007761.htm#7
which suggests that 82 degrees is pretty darned hot for Xeon processors (spec. for max. case temp. typically in the 70s). If you scroll upward from that tag, there are various recommendations for keeping Xeon processors cool.

However, I would not assume the temperature reading is accurate without knowing a little more about the sensor system. Pardon me, what is an lm sensor? Are they supposed to be pretty accurate? If the sensor is easy to remove and re-install, perhaps you could test/calibrate it next to a trusted thermometer, or in contact with something that has a well-defined temperature (for example, boiling water 100C, ice water 0C).

TobiSGD 05-11-2013 07:13 AM

lm-sensors is a software program that reads out the internal sensors of CPUs. Usually those internal sensors aren't pretty accurate in the lower temperature range, but they are in the higher temperature range.
82 is indeed pretty hot, close to thermal shutdown for those CPUs. If the air that is blown out of the case is not hot there may be a different problem, for example cooling system blocked by dust or a not working fan (heat transfer from cooler to air is not working correctly), dried up thermal paste or a cooler that is not seated correctly (heat transfer from CPU to cooler isn't working correctly).
Looking at the temperatures you give us I would guess from experience that this is either a not working fan or simply dust in the CPU cooler. Check that first.

thorkelljarl 05-11-2013 08:05 AM

New paste...

You might have that high a CPU temperature. Sometimes there is poor air flow through the CPU cooler fins due to dirt or some obstruction, or the cooler paste between the CPU and the CPU cooler has dried out resulting in poor heat transfer.

If you conclude that the temperature is indeed that high and you can't find any other cause, you might try re-installing the cooler with some Arctic Silver paste.

onebuck 05-11-2013 08:56 AM

Member Response
 
Hi,

I would add that you should enter BIOS via setup and confirm the temperatures via that interface. Depending on your lm-sensors setup with the chipset, you can get misconfiguration thus errant readings.

lm-sensors has shown great improvements over past chipset setups so the chance of proper configuration is probable but with newer motherboards with different chip-sets then you could still have problems.

Check out: lm-sensors.org & CHANGES

xmrkite 05-14-2013 11:46 AM

Since it's my server at my office, I will have to wait till I have time off hours to reapply some cpu thermal compound. I'll post back my results.

-Thanks

xmrkite 05-18-2013 01:51 PM

Well, I redid the thermal paste and put it all back together and it went down to the low 40's when idle. it's not under a good load and is at 56C or so. So that may have been the solution...still, it seems about 10 Degrees C too hot.

Any ideas?

-Thanks

TobiSGD 05-18-2013 06:34 PM

If your CPU is too hot although you have applied the thermal paste correctly, the CPU cooler is seated correctly and you have configured power management than your only solution would be to get a better cooler or maybe try to get better airflow in the case.

Beryllos 05-18-2013 07:52 PM

Quote:

Originally Posted by xmrkite (Post 4953875)
Well, I redid the thermal paste and put it all back together and it went down to the low 40's when idle. it's not under a good load and is at 56C or so. So that may have been the solution...still, it seems about 10 Degrees C too hot.

Isn't 56C all right? It's well under the maximum, which I assume is in the 70s. While 56C feels hot to the touch, it's only lukewarm for silicon chips.

I didn't understand what you meant by "it's not under a good load." What is that in percent CPU utilization?

xmrkite 05-30-2013 11:15 PM

It goes to about 40c at idle and 56c-60c when at about 50% utilization on all 4 cores.

TobiSGD 05-31-2013 06:14 AM

You should test you cooling system always under maximum load, 100% on all cores.

xmrkite 05-31-2013 01:35 PM

I don't know how to get it to max out though. Is there a simple program that can do that?

thorkelljarl 06-01-2013 05:58 AM

AS TobiSGD mentioned...

It could be that whoever built this server used a cooler that was just good enough. If you want added security, and it's possible to switch coolers, you could install a heavier duty cooler such as used by gamers when over-clocking.

This of course assumes that there is a good exchange of air in the server cabinet to remove the hot air exhausting the CPU cooler. Good Luck

PS I believe one way to stress a CPU is to use it to compile the linux kernel. Otherwise the Internet tells me that there is an Ubuntu program called "stress" to stress.

xmrkite 06-01-2013 01:07 PM

Ok, i'll try that stress program. But wouldn't the air coming out of the fan be hot if the cpu was really that hot? The air is not even warm. It's actually cool.

The case is a regular tower case with what looks like an intel stock 775 fan on it and then you have a nice intake fan in the front of the case and a great exhaust fan in the back. Cooling seems pretty well done.

Beryllos 06-02-2013 02:09 AM

I'm curious, so I looked up some information available from Intel. Your xeon processor has a "thermal design power" of 105W. That is their estimate of the maximum power at full CPU load for the x3220. Intel's "reference design" for the heatsink system specifies a thermal resistance of 0.42C/W between the CPU and the surrounding air (the air inside your tower case). Your actual heatsink might have a different thermal resistance, but let's assume their figure is accurate. Then the CPU temperature will be (105*0.42)=44C above the air temperature. If the temperature inside your box is a slightly warm 30C, then the CPU temperature will be 74C.

The reason the air does not feel hot is because the airflow through your case is relatively good. I don't know exactly how many CFM is flowing through your tower, but let's assume it is 70 CFM (middle of the range for case fans?). The formula for air temperature rise is roughly deltaT(C)=1.8*power(W)/airflow(CFM), so that would be 1.8*105/70=2.7C= about 5F, so the air exhaust is going to feel barely warm. All the other stuff in the case (hard drives, graphics processor, etc.) will raise the temperature a bit more. If we assume the total system power at full load is 250W, then the temperature rise at 70CFM would be about 6C (about 12F). So if your room temperature is 24C (75F), the air temperature in the case could be 30C.

By the way, one simple way to exercise or stress the CPU is this:
Code:

cat /dev/urandom > /dev/null
Start one of those per core. When you're done, you can stop it with Ctrl-C. I have only tried it with dual-core Pentium processors, and according to "top," each such process runs one processor above 99%. I am not sure if that is equivalent to maximum CPU power dissipation, but it's probably close to it.


All times are GMT -5. The time now is 11:45 AM.