Hot Processor
 

Hello.

I understand that this is not strictly a Linux issue, but I thought that as there is a lot of expertise to be found in these forums, I should be able to get a reliable answer.

I was compiling a kernel for my Debian box and later I received messages that my processor temperature was beyond its threshold, at 97 degrees Celsius. So I immediately instructed the machine to halt and inspected my heatsink and fan attached to the CPU. Nothing seemed out of the ordinary, but I reapplied the thermal paste again after using a chemical agent to remove it and purify the surface.

When I booted in again, the BIOS reported that the CPU temperature was now at 48 degrees Celsius.

This issue seems resolved but my worry is that damage could have been done to either the CPU or the motherboard during the time that the CPU overheated, as I left the machine to make some tea and dinner whilst it compiled, leaving it at this temperature for quite a while.

What are the chances that I have damaged something?

Thanks in advance, peace out.

It's really hard to estimate the probability of damage because some processors are close to a short circuit when they ship from the factory. It isn't possible for the manufacturers to detect this so any processor could be close to failing during its entire life.

All you can do is the same thing that the manufacturers do. Run it and see if it works or fails. The operating system will probably detect machine errors and generate system log entries or just simply crash.

I like to set my motherboard BIOS so that it generates a warning when the cpu is about 85 C and to have the motherboard force the processor into running at half speed if it gets over 90 C. I've currently got one computer that blasts an alarm about once a month. I dismount the encrypted partitions and the motherboard alarm stops. This computer is several years old. Built it myself. Used Arctic Silver thermal grease. I don't know if the thermal grease has a problem or what is happening but it keeps running so there is obviously no damage to the cpu.

So in summary, I doubt that you can estimate the probability of damage from overheating.

I have a spare chip, but it's the motherboard I am worried about the most. Everything seems to be working just fine. I'm in KDE and working as normal, but it worries me all the same.

I've had sessions with games that lasted hours where my graphics card was over 100c and it still works fine...

...but I'd say your BIOS is right as sometimes programs read the temp sensors incorrectly. I usually find sensors works OK, assuming you configured it correctly.

It really only matters if the insulation inside the "grooves" or "lanes" ... I'm having a little trouble articulating this as I'm not the last word but ... if that insulation becomes damaged your CPU will short out.

Also, what CPU do you have? Some CPUs get that hot and 97c would be perfectly normal.

There's no way to know what the consequences are if the temps were beyond the max for your processor. Perhaps the best you can do is gain some confidence by running some lengthy processor diagnostics to verify expected results, Run them periodically over the next week or so; I suspect there is a short threshold where if it were going to malfunction, it would do so immediately, or fairly soon afterward.

Mind telling us what kind of processor you have ?

I mean I know old AMD processors didn't have any kind of overheat protection, it was similar to nuclear meltdown, "china syndrome", the processor would turn into a molten metal slab that would melt right through the motherboard ... lol. That won't happen with new AMDs (which will automatically cut power to the processor on overheat, not the best way, but it seems to work), and most certainly not with any Intel (which have thermal throttling that will throttle down the processor on overheat).

If it's bad, you won't have gotten that far. Plus 97C is not that hot. Not hot enough to damage much, as for the motherboard, the only way a hot CPU can be a problem is when you have a lot of very hot and cold cycles, which strains the PCB and could eventually cause stuff like delamination. If this has only happened the first time, theres really nothing to worry about, however I'm sure if you load it up again with a task such as compiling, you will be seeing that message. I doubt there was anything wrong with the heatsink compound the first time.
48C on idle in BIOS is kind of hot, I would get a different cooling device.

I don't think 48 C is hot. That's what I would consider normal, cool temperature for CPUs and GPUs. That's the normal temperature on all my desktops, it's hard to get it too much cooler than this. On my laptop the normal temperature is around 60 - 70 C, with max of 85 C. Usually meltdown happens near 100 C, but probably won't happen for the aforementioned reasons.

Thanks for your replies everyone. I'm beginning to feel much better about the whole thing now.

I have and Pentium 4 2.8 on a D946GZIS motherboard.

Yeah you want 60-70C tops on high load. ~30C on idle is good.

I think everybody here is wrong because semiconductors prefer to be colder. One problem to make it colder is the heatsink needs to be very, very efficient and the ambient temperature needs to be lower than the desire temperature of the processor.

90 degrees C is too hot for a processor and I do not care what anybody says. That is hot! That is close to boiling water. Try putting a working processor over a pot of boiling water. It will not last long. Sure the BIOS can be off of monitoring the temperature, but again 90 degrees C is TOO HOT!

I recommend use compress air and blow out all the dust, re-order the cables to provide more efficient air flow, and have at least 2 intake to 1 out-take fans. Do no use a vacuum to clean the insides of the computer.

Both AMD (since K8) and Intel processors only uses a soft set thermal shutdown, so the thermal shutdown can be disabled at will in the BIOS. If the thermal shutdown is set too high, it will never catch the processor in time before overheating. Read the maximum thermal specs of the processor and set the thermal shutdown lower than that.

If the computer is from Dell, HP/Compaq, Lenovo, Gateway, Acer, eMachines, the BIOS can be set wrong for thermal shutdown or the BIOS can be buggy and it will never monitor temperature accurately to shutdown the processor from overheating.

Heat damages processors. One day the processor could be working, the other day it will not work. Digital electronics do not warn you when they are not going to work anymore. If the processor seems to be working after it has been through torture of going over its maximum temperature limit, some parts of the processor may not work efficiently or will provide corrupted data.

Processors do not melt when they get hot. They become like a pressure cooker. When the (gas) pressure gets high enough, it pops. Transistors and resistors are different. They smoke and later catch on fire. A special fire extinguisher for electronics should be used to put out the fire.

Back to Linux, lm-sensors can provide incorrect readings, but can be easily be changed if you know the hardware being used. Diodes and BJT transistors are the best temperature sensors and thermistors are the worst to use for temperature sensors. Diodes and transistors are the cheapest over all. lm-sensors has two settings for a temperature sensor. One is diode and the other is thermistor. Thermistors needs fancy math to make them linear to make them close to accurate while diodes are linear by nature, diodes just needs some fine tuning to provide accurate readings through the range.

Would the overheating cause damage to other components like the memory, PSU and hard disks for example?

I'm guessing you have a prescott(nicknamed presshots on ocing forums) and they get very hot. The best advice is to either get a heatpipe cooler or make sure your case has plenty of ventilation. An extra extake fan or two would be cheaper and easier to install than a heatsink. Also, make sure there is plenty of breathing room around your CPU.

All PC components can be damaged by heat. My understanding is that hard disks are the most sensitive to heat.

An overheated processor will not cause damage to other components UNLESS the processor fails.

If it fails, then the devil is in the details; exactly HOW does it fail? Yes, upon failure it can carry away other logic including memory and components on the mobo. Hard drives are sufficiently removed electrically that they shouldn't experience any damage.

How often does the failing processor damage the mobo? Often enough. My experience is that a mobo failure goes with a processor failure about 15% of the time.

97C is very hot. Probably you are OK if the processor still works. You won't get away with that too many times though.

Thank you for your replies everybody. I'll have to let things run their course I guess. Fortunately I'm building a new system soon anyway, so if anything does go wrong it shouldn't be too disruptive.

Thanks again, peace out. :)