Sign in with
Sign up | Sign in
Your question

102 degrees C? Why I doubt the measurements.

Tags:
Last response: in CPUs
Share
March 2, 2012 3:10:33 AM

Hi everybody. I'm really worried about this but find myself wanting to believe that there isn't actually a problem. If you wouldn't mind the long read, please help me to figure out what's going on.

If TL;DR, then please answer me this: If my Intel C2D CPU were overheating, I'd expect it to throttle and reduce its voltage and underclock as a built-in safety measure. As far as I can tell, this isn't happening, even when it went as high as 102 C, so I suspect (and hope) my measurements are wrong. Is there any way that I can know for sure? Is there a tool that will record the event of the CPU throttling?



Here's the story:

I've upgraded my Intel CPU a little (from a Core 2 Duo E6550 to an E8400), and now I find that my idle core temperatures are 50 degrees C. When I play games (I tested with Mass Effect and Dead Space because both perform really well all maxed out), the temperatures rapidly rise to around 90 degrees C, and then they basically plateau there, but it seems like over a long session that plateau slowly rises to, say, 91, 92, eventually 93, ten minutes later 94, etc.

My CPU has a temp case rating of 72.4 degrees. My understanding is that it's okay to surpass that but an extended period of like 24 hours running at above that temp will cause damage. The program RealTemp holds my temperatures against 100 degrees as the maximum temperature junction, which I think is the point where the superconductors will melt or something.

Supposedly, 5 degrees below max junction the CPU is supposed to throttle and underclock itself in an attempt to cool down. If the CPU reaches the max junction, the computer is supposed to just shut off as a built-in Intel CPU safety function.

The thing is, there are never any performance hitches or freezing or anything in the games that run well while the temperatures are hiking into the mid-90s Celsius. Can I take this to indicate that the measurements are incorrect?

I ran a stress test thing where the CPU is made to calculate prime numbers and works at 100% load, and one of the cores went as high as 102 degrees C; the PC did not shut down, and as far as I could tell (I wasn't running a game or anything and couldn't judge moment-to-moment performance) there weren't any consequences to the heat. Isn't that odd? I mean, if my cores really hit 96 and 102 degrees C, wouldn't the computer have at least frozen up? But nothing happened. This test lasted about five minutes, and those temperatures remained at that peak for around a full minute, spending most of the rest of the time around 88-95 (one core is always a little hotter than the other).

I really wish I had taken readings before I changed any hardware. When I originally built this thing, the heat sink was a problem, and I was extremely concerned with the temperatures I was getting. I remember reseating it then, but I don't recall whether it made a difference; whatever happened, I've run for five years without issue.

When I installed the new CPU today with that same fiddly stock Intel heat sink, I went back and reseated it twice when I saw the temps. I'm concerned that I may have failed to seat the heat sink properly despite repeated attempts because the pegs do not snap through the motherboard very nicely; I think one corner may be a little loose. But I felt like it was really sturdily on there when I tried to move it. My other worry is that I may have used too much thermal paste. I kind of spread a liberal amount on there using a plastic card (yes, I cleaned all old paste off first).

But maybe everything is exactly the same as it has been for the last five years. I'd really like to believe that everything is fine here despite the readings. I can't afford even a $60 heat sink right now, so there's nothing I can really do about this; even if I went back to my old processor, I'd still have to put the same heat sink back on top and would probably only make that worse if there is really something wrong there. Besides, I'm out of paste. And I've taken next week off from work to play Mass Effect constantly.

I've tried to find stories online of people who have had high CPU heat measurements with none of the problems you would expect to accompany them, but it's kind of a difficult, nuanced search. I'm hoping that someone here might have enough experience or insight to make an educated guess regarding the truth of my situation. Do you think I have a real problem here? Should I drop everything, or should I keep on keepin' on?




Incidentally, both Assassin's Creed and Batman: Arkham Asylum completely crash the computer (it reboots) just before they get to their title screens, but I think that this is a compatibility issue with the Radeon HD4850 that I put in and not due to the CPU.
a b à CPUs
March 2, 2012 3:26:05 AM

re seat the heat sync, make sure you use a small pea sized ammnt of thermal paste, 100C is way too hot for a CPU, and the Tjunction is 100C, making the TjMax 105C, so you just below it
March 2, 2012 3:43:11 AM

But, again, wouldn't I see performance changes at such high temperatures if the measurements were accurate? Wouldn't it throttle, and then wouldn't I see a drop in temperature in my SpeedFan chart?

Like I said, I tried reseating the sink repeatedly. One time it was definitely a bad job and the temps just went up and up even idle from 65 to 70 to 80, but every other attempt has had the same result I see now with 50C idle and around 90 in games. If the sink weren't seated properly, wouldn't I be seeing much more instability like that one time?

I'm looking for someone to tell me how I can confirm that the temperature measurements are accurate. I'm looking for a way to see that Intel's CPU safety throttling is occurring when I hit high temps.
Related resources
a b à CPUs
March 2, 2012 3:51:32 AM

can you post a screenshot? that is waaayy to hot for a C2D chip
and also, can you list your entire system please?
March 2, 2012 4:18:53 AM

During a Sensor Test in RealTemp where the CPU was made to crunch prime numbers at 100% load, the temperature of one of the cores passed the "Distance to TJMax" count. It went from 2 to 1 to 0 to 127 (no negative value allowed, apparently). RealTemp seems to treat 100C as TJMax. Shouldn't the computer have shut down at 102C?
Anonymous
a b à CPUs
March 2, 2012 4:25:30 AM

i fried a Q6600 thinking it would shut down before it got damaged.

clean off old thermal paste, apply new SPARSELY ( i liked the thin line method for my C2D) reseat the heatsink and make sure any thermal protection in the BIOS is enabled, ok?
March 2, 2012 4:54:46 AM

Alright, I'm going to try it one more time with a small dollop only and being careful about the order in which I press down those stupid fiddly plastic heat sink corner pins. I'm amazed I haven't cracked the motherboard yet pressing down on those.
March 2, 2012 8:01:20 AM

I'm such a fool. I finally got the seating right, and I only used a pea-sized dollop in the center. It's completely fine now. Idle around 40C, Dead Space plays at about 55C. GPU gets hotter than CPU now.

Thank you both for steering me in the right direction. I only hope I didn't damage the CPU during the time I was being thick about it. I'll run some diagnostic tests overnight to check its performance.
Anonymous
a b à CPUs
March 2, 2012 2:37:39 PM

internisus said:
I'm such a fool. I finally got the seating right, and I only used a pea-sized dollop in the center. It's completely fine now. Idle around 40C, Dead Space plays at about 55C. GPU gets hotter than CPU now.

Thank you both for steering me in the right direction. I only hope I didn't damage the CPU during the time I was being thick about it. I'll run some diagnostic tests overnight to check its performance.


your fine. i have had issues with intel's heatsinks myself, especially those push pins.

if your gaming then whatever damage you *may* have done isn't going to affect you immediately.

if running tests do show errors, you could consider getting an Q95xx with a better heatsink or use about 2x that money and get a sandy platform.

happy gaming! :) 
March 2, 2012 7:46:23 PM

I appreciate that. I don't think that my motherboard would support the newer technology like Sandy Bridge, though, and my most immediate priority is to purchase a strong case and motherboard that will create a platform for gradual future improvement (it's been extremely limiting to work with a micro-ATX form board and small-form factor case). The future looks pretty good right now, though! Thanks again!
a b à CPUs
March 20, 2012 3:26:03 AM

Just to clear up a few things. Intel sets the thermal shutdown temperature approximately 25C to 30C higher than TJMax. An E8400 has a TJMax of approximately 100C so an official thermal shutdown will not occur until somewhere between 125C to 130C. Some CPUs will become unstable before this temperature, especially when overclocking and might randomly shutdown sooner than this.

If one of these CPUs hits a core temperature of 100C it will definitely start throttling but this happens so rapidly that when it first starts to happen, most users will barely notice performance degrading.

The temperature sensors that Intel uses in these CPUs can only store temperature information in 7 bits of data so that results in values from 127 to 0. When a sensor counts down to 0, thermal throttling begins. If the CPU continues to heat up beyond this point, these sensors wrap around and start counting down from 127 again. What you saw during your test was perfectly normal for these sensors. Think of it like an old car speedometer that would wrap around after you went by 120 mph and it would start counting up again from zero only this is the opposite. Intel Core 2 sensors count down and then they wrap around. Unfortunately Intel did not like RealTemp interpreting this data to show temperatures higher than 100C so they blocked this in the newer Core i CPUs. They count down to 0 and then that is it. You will have no idea how hot your core is after the sensor reaches zero but they still leave 25C to 30C of headroom after this point before a thermal shutdown will occur to protect the CPU.

Here's what an E8400 looks like after 3 hours of Prime95 with the CPU fan disconnected.



It ran just fine with 0 errors with the core temperature flat lined at the maximum. Intel thermal throttling works great as long as the heatsink hasn't fallen off and if it has, the CPU will reach the thermal shutdown temperature to protect itself from damage. No worries.
!