Physics question regarding CPU temperature

[ambushed]

One extremely annoying aspect of asking tech support questions on tech fora (present forum excluded, of course) that I far too frequently encounter is when people utterly ignore my stated question(s) and instead “answer” what they arrogantly contend I should have asked! That’s just what’s happening again right now with regards to my questions elsewhere about the possibility of CPU temperature sensor and/or related motherboard malfunctioning.

You see, I cannot boot one of my computers because the BIOS hardware monitoring tells me my CPU temperature is above safe limits. And if the BIOS temp reports were accurate, that would be quite plausible.

But I simply cannot believe those measurements! Here’s the situation: The computer is in a room that stays about 19c/67F and I hadn’t run it in many days, so the CPU in question was at ambient temperature. Then I turned on the system, hit the key to enter the BIOS setup, and went to the hardware monitor page. There it reported the mobo temp was, as expected, at the ambient 19c/67F, but the CPU temp was reported to already be 70c/158F just seconds after powering on! Thereafter, the CPU temp readings increased at 0.5c every couple of seconds!

Here’s my first actual question, then: Is that even physically possible? A 51 degree centigrade jump in just 10 seconds without me actually aiming a blowtorch at the CPU?

That’s been happening now for a couple of weeks now. I wrote it up that way (in past tense) to emphasize that this happens even when the system hasn’t been powered up for days at a stretch.

Intuitively, those numbers seem utterly preposterous. But when I ask at other support boards about the possibility that some kind of malfunction is causing that literally incredible reading, everyone instead keeps giving me completely unsolicited “advice” about computer cooling, as if I’m some ignorant cretin who hadn’t already checked and double-checked and triple-checked all that. Please, please, please, dear dopers, don’t repeat that pattern here! I’ve been home-building computers for decades, and I know what I’m doing.

If those readings are, as I suspect, thermodynamically impossible, my best guess would be that the CPU chip itself (an Intel E6850) is malfunctioning, since the temp sensor is built into the chip. Has that kind of failure been known to happen? Or might it be a motherboard problem of some kind?

But primarily I want the physics question answered so I can cite your answer(s) in order to hopefully get others to focus on the possibilities of a malfunction rather than just keep telling me how to “fix” a non-existent cooling problem.

Thanks!

[srzss05]

@ambushed

But primarily I want the physics question answered so I can cite your answer(s) in order to hopefully get others to focus on the possibilities of a malfunction rather than just keep telling me how to “fix” a non-existent cooling problem.

You can’t possibly think “cites” to this message board is going to make other message boards focus on your stated problem, can you?

[jz78817]

Here’s my first actual question, then: Is that even physically possible? A 51 degree centigrade jump in just 10 seconds without me actually aiming a blowtorch at the CPU?

oh yes, with no heatsinking, that little sliver of silicon will heat up faster than you can blink. The pre-K8 AMD chips had no thermal safety, and if you booted with no heatsink they’d self-destruct in about six seconds.

ETA: think of it this way- the actual CPU chip is a thin piece of silicon about the size of your fingernail that has to dissipate about 50-90 watts of power. without the mass of the heatsink tightly coupled to it, it can easily jump 51 C in a couple of seconds. Intel chips have that heatspreader as a bit of a buffer, but it’s not much.[end edit]

If those readings are, as I suspect, thermodynamically impossible, my best guess would be that the CPU chip itself (an Intel E6850) is malfunctioning, since the temp sensor is built into the chip. Has that kind of failure been known to happen? Or might it be a motherboard problem of some kind?

it’s entirely possible the thermal diode(s) on the chip are wrong, or the BIOS isn’t reading it properly.

but I’m going to go ahead and do what you didn’t want me to do and talk about cooling. I’ve as much experience as you and I can confidently say that the push-pin heatsink mounting that Intel uses SUCKS ASS. it’s easy to get them just locked in enough to look like they’re seated, but one or more pins can pop loose tilting the heatsink off of the CPU package and causing overheats.

if you’ve truly rechecked the heatsink mounting- preferably by removing it and making sure the thermal compound has spread- then it’s not impossible that the thermal monitoring is screwed up.

[jz78817]

since I missed the edit window, I’ll add that if it is a problem with the thermal monitoring, I don’t know if there’s an easy way to solve it. you’d need at least another CPU or a motherboard to test and try to eliminate one of the two as a culprit.

[Napier]

If the chip is 0.01 by 0.01 by 0.001m in size, that’s 1e-7 m^3 in volume. Most solids have a volumetric specific heat capacity of around 2e6 K/(m^3 K), so its nonspecific heat capacity would be 5 J/K. If you push 50 W (which is 50 J/s) of energy into the chip and it becomes heat that just increases the chip temperature, that would go up at 10 K/s (10 ºC/s). So, yes, sure it’s possible for it to heat that fast, on the basis of the energy and its bulk.

Whether you are actually seeing that would depend on whether all the temperature sensing works right.

Whether it is actually happening would depend on heat sinking.

You might want to look into chip cooling…

[ambushed]

Thank you, Napier, for directly answering my actual question.

It’s regrettable that I can’t use a thermodynamic argument to get people to accept that I’m not a complete and utter moron who did not just throw the system in question together several years ago without choosing a particularly well-reviewed high-end workstation-class motherboard renowned for its cooling capacity, did not bother to install a particularly well-reviewed and expensive water CPU cooling system, and did not bother to properly apply particularly particularly well-reviewed thermal paste in a case without four high-end cooling fans, but I guess there’s no helping that…

[Quartz]

Since the physics side has been answered, I will just say that I’ve had it happen to me. I thought I’d replaced the heatsink but I’d not replaced it correctly and it didn’t make contact with the CPU. Lots of beeping followed by an immediate shutdown.

If you’ve tried it a few times, you may have cooked the chip. Like I did. :smack:

BTW the Intel E6850 operates at up to 65W.

[iamnotbatman]

@Napier

If the chip is 0.01 by 0.01 by 0.001m in size, that’s 1e-7 m^3 in volume. Most solids have a volumetric specific heat capacity of around 2e6 K/(m^3 K), so its nonspecific heat capacity would be 5 J/K. If you push 50 W (which is 50 J/s) of energy into the chip and it becomes heat that just increases the chip temperature, that would go up at 10 K/s (10 ºC/s). So, yes, sure it’s possible for it to heat that fast, on the basis of the energy and its bulk.

You made an inversion by accident. 2e6 J/(m^3 K) * 1e-7 m^3 = 0.2 J/K. Pushing 50 W into that CPU will heat it at a rate of 50 J/s * 5 K/J = 250 K/s. So in a single second the CPU can easily increase to beyond 250 degrees C.

Answer to OP: it is indeed physically possible.

[jz78817]

It’s regrettable that I can’t use a thermodynamic argument to get people to accept that I’m not a complete and utter moron who did not just throw the system in question together several years ago without choosing a particularly well-reviewed high-end workstation-class motherboard renowned for its cooling capacity, did not bother to install a particularly well-reviewed and expensive water CPU cooling system, and did not bother to properly apply particularly particularly well-reviewed thermal paste in a case without four high-end cooling fans, but I guess there’s no helping that…

I don’t know why you’re getting so emotional about this, given 99.9% of the time a problem like this is installation related which is why people are harping on it. a failure of the temperature monitoring to work properly is not impossible, but it is rare.

have you checked just in case there’s a BIOS update for your board that might address an compatibility issue with the specific revision of CPU you have? that’s happened too, and hopefully you won’t get all offended that I asked you that.

ETA:

@iamnotbatman

Answer to OP: it is indeed physically possible.

to be clear, it’s possible assuming there’s no heatsink. The OP assures us that can’t be the case, so if there’s a heatsink properly attached then highly improbable.

[Quartz]

@jz78817

to be clear, it’s possible assuming there’s no heatsink. The OP assures us that can’t be the case, so if there’s a heatsink properly attached then highly improbable.

As I found out the hard way, the heatsink can appear to be correctly attached but in fact not be.

[jz78817]

@Quartz

As I found out the hard way, the heatsink can appear to be correctly attached but in fact not be.

which I (and presumably the people on other message boards) have been saying, but the OP is insistent that he is not so much of a moron to make that kind of mistake.

[Napier]

@iamnotbatman

You made an inversion by accident. 2e6 J/(m^3 K) * 1e-7 m^3 = 0.2 J/K. Pushing 50 W into that CPU will heat it at a rate of 50 J/s * 5 K/J = 250 K/s. So in a single second the CPU can easily increase to beyond 250 degrees C.

Answer to OP: it is indeed physically possible.

It’s regrettable that I can’t use a thermodynamic argument correctly to get people to accept that I’m not a complete and utter moron…