10-15-2007, 11:30 PM

So, I came across something rather interesting while trying to run Rember on my MBP:


Note the CPU diode temperature. It's possible that this is a glitch in hardware monitor (though smc fan control reads the same), but it's also possible (and more likely, in my opinion) that the diode is close to the area on the cpu die that deals with memory access, thus the large temperature disparity.

I think this is backed up by the fact that I have never, ever, EVER seen the bottom or top of this machine get anything like as hot as it did. I actually burned my hand on the bottom case (it's still a bit sore several hours later).

Interestingly, it'd jump up to almost 130*C (the screencap says 123, I saw as high as 127), then it'd drop down, then jump back up again. I don't know if the processor was throttling or just what. I really need to install XP on this thing and run the Intel Thermal Analysis Tool.

Anyone here willing to put their MBP on the line and see if they can confirm that my 127*C reading isn't abnormal while running Rember? Or anyone that's got bootcamp installed that's willing to run the Intel TAT and see what it hits on that?

I think this bears some investigation... If the CPU in this thing is really hitting almost 130*C due to executing a specific application, then Apple needs to do something about it. There is absolutely no excuse-none-for a computer's cooling system to be so totally inadequate that a program can load a chip (be it the CPU, the GPU, the northbridge or the RAM or anything else) to the point where the chip actually exceeds its maximum operating temperature specification.

//edit: I should add a note here that testing this out, if the cpu is actually exceeding 100*C, could damage and/or kill your MacBook Pro.

10-15-2007, 11:35 PM

i know this may sound like a stupid question, but i dont recall seeing you mention anything about it. did your fans turn on at all?

10-15-2007, 11:37 PM

The fans in the MBP run all the time; I did not see any significant RPM increase, or at least I didn't notice any, no. I forgot to mention it, but I should have.

However, that in and of itself does not necessarily indicate that the CPU did not hit well over 100*C, nor does it necessarily indicate that the fans are operating abnormally for a MBP. All the evidence that I've seen points to a serious deficiency in the thermal interface between the CPU, GPU, and northbridge and the heatsink on these machines; (evidence in this case being the massive disparity between the heatpipe/heatsink temperatures and the chip temperatures under load) my personal belief is that it's due to a very poorly engineered mounting system that effectively forces the thermal grease to act as the primary interface between the CPU and the heatsink in some (though-due to variations in the height of the chips-not all) machines rather than as an 'assistant', if you will, which is its normal role. (I should note, though, that other potential causes for the temperature disparities include poor sensor placement and/or calibration)

And since the sensors that control fan speed (as far as I know) are mounted on the heatpipe, and the heatpipe stays much cooler than the CPU, it's entirely possible that they wouldn't speed up significantly even as the CPU got very, very hot. I notice that whenever I run the CPU under high loads, the fans don't speed up. I notice that with two other MBPs as well (the one I had to exchange for a different reason, and my grandmother's).

It is entirely possible that the CPU is not getting that hot and that it's a glitch. I have no way of accurately measuring it without voiding my warranty, and I'm not that dedicated to the pursuit of knowledge. But, I think, based on the history of heat issues with the MBP, that it's more likely that this is an actual problem. I could, however, be wrong. If anyone wants to send me their MBP and let me take it apart, I'll investigate further

10-16-2007, 12:44 AM

well, I've taken for granted so far that I've upgraded the MBP's RAM with OCZ lifetime warranty stuff, but no harm in testing it out to make sure that it is not semi duff

I'll post back later on what temps it hit.

similar results to you...

hit max of 126 C, 1 core was hovering around 40% the other around 80%, temp was up and down like a yoyo, but the most disturbing thing was that the fans never got anywhere near max rpm, just under 3500rpm at 126 C

been sat at 124-126 C for a while now doing Block Sequential, the actual temp of the case by touch seems the same as maxing both cores out converting video etc which for me is around 83 C, couple that with the fans not going mental, I suspect that it is a glitch, but I think Apple should be made aware, just to make sure. I'd test it with bootcamp, but have not got round to reinstalling WXP on it yet.

no burning smells or anything :p but I think I'll stop the test now, until it is clarified what is actually going on.

10-16-2007, 01:05 AM

wytey: just thought I'd warn you before you did it (just edited my first post to reflect the warning)... if this thing is actually causing the CPUs to exceed 100*C, you could very well damage or outright kill your MBP. Just so you know.

10-16-2007, 01:26 AM

too late not in that respect lol

aye, MBP only just hit the 1 month mark, and also has AppleCare too, so didn't worry me that much, would have been a pain to reinstall everything, and set it all up how I want it.

going to put it onto a USB pen, and take it to an Apple Store later this month when I go get my MiniMac, and have them run it on their machines to see what they think of it.

10-16-2007, 02:08 PM

I just ran this, and it said my CPU got to 128C, however I immediatly did a Force Quit on Rember, and the temperature dropped to 68C in a matter of 3 seconds.

Are you sure the temperature reading isn't a software fluke in some manner? I didn't feel my MBP get any hotter than normal, and my fans never raised above 3k RPM.

10-16-2007, 03:59 PM

It's conceivable that it could be a software issue, but to be honest I think it's highly unlikely, because memtest/rember would have to interfere with the CPU diode measurement in some way, and I've never seen a program do that. I've seen Motherboard Monitor (old PC hardware monitoring program) return implausible-sometimes impossible-values for temperatures, but only when the wrong board or sensor configuration was selected.

I'm not saying it's impossible that it's a software glitch, but I really feel that it's unlikely. If anything was going to cause weird temperature readings, it'd be running more than one hardware monitoring program at a time, which would create a potential for both programs to try to access the same sensor at the same time, that could then cause bogus readings. But my temperatures are the same regardless of whether Hardware Monitor or SMC Fan Control are running alone, or both are running together. I can't prove it conclusively, but I really, really, REALLY don't think this is a software problem. It just doesn't add up for me.

It's possible for a CPU to experience localized heating where one part of the die is significantly hotter than another. This is particularly true when the CPU has very, very poor contact with the heatsink base (as it seems to in many MBPs).

And, though I'm not an electrical engineer by any means, from what I've been given to understand Intel (and AMD) tends to put the on-die temperature diode in whatever area of the cpu will get the hottest (that'd be the logical place to put it). Which means that while I suppose that there's a chance that it's a software glitch (though I can't fathom how), my gut feeling is that it's more likely that the reading is accurate.

Why? A couple reasons...

1. The MBP's fans will not necessarily speed up significantly with increased CPU load, meaning that it's not an accurate indicator of how hot the CPU is getting in all cases. As I mentioned, the sensors that control the fan speed appear to sense the heatpipe temperature, not the CPU temperature; this is supported by the fact that people have reported their fans running at full speed constantly if the thermistors on the heatpipe are not reconnected after disassembling the machine.
   
   Unfortunately, the thermistors on the heatpipes register temperatures much lower than the actual CPU/GPU/NB temperatures, likely thanks to poor contact between the CPU and the heatsink, thus the fans can sit at 2500rpms or whatever while the CPU cooks merrily away at 90*C.
2. The MBP has a known heat issue that, by all accounts, is traceable to an issue with the thermal interface between the CPU and the heatsink (and the amount of thermal grease Apple puts on these things is further evidence of a contact problem). Fullscreening the visualizer in iTunes will run both my current one and the one I had to exchange up past 90*C, and my grandmother's behaves similarly. Memtest is very, very hard on certain parts of the CPU, and this, combined with the thermal interface problem, could very well result in localized areas of the CPU die topping 100*C.

Ideally, someone with Windows XP installed on their MBP (I'm still trying to find any one of my XP CDs ) would download and run the Intel Thermal Analysis Tool and report how hot the CPU gets when the TAT is set to run both cores at 100% load. That program will produce more heat from the CPU than anything else I've come across-that's what Intel designed it to do. Supposedly (though I have no way of confirming it) it will make the processor dissipate at LEAST its TDP, if not slightly more. That'd give us an idea, I think, of how hot the CPU will get, realistically speaking.

It'd be even better if they scanned for artifacts in ATItool or ran rthdribl at the same time, as those programs will push the GPU pretty hard, and since the GPU and the CPU share a common heatsink... I'd be interested to see if that would generate results similar to Rember or not.

10-16-2007, 05:48 PM

If the CPU temp really was at 129c the CPU would blow. I am not calling you a liar but I really doubt that 129c an is accurate reading.

10-16-2007, 06:13 PM

Nah, no offense taken.

That has occurred to me as well. The only thing I can say is that I'm not 100% convinced that it would kill it instantly, or even over a period of a few minutes. It very well could, without any doubt at all. There's no question that it would not be good for it, to say the least.

But. It might not kill it instantly. I did an experiment a long time ago on a forum I used to frequent because some mental giant was insistent that the P4s were more stable at high temperatures than the Athlon XPs were. The experiment (which can be found here: http://icrontic.com/forum/showthread.php?t=14207 ) consisted, basically, of running a Pally 1700+ up to and somewhere beyond its maximum operating spec. The CPU suffered no ill effects from this, and indeed it continues to function just fine.

So yeah... I understand your point, and it's a very good one. And I'm torn as to which way to go with this... I'd be much happier if I could pull it apart without voiding the warranty and do some testing.

10-16-2007, 06:20 PM

Hi,
I noticed the problem you are talking about too. I'm quite happy not to be the only one...
I did some tests launching multiple super pi computations at the same time. Here are my observation, during the first one or two hours, no problems, the fans speed up to 6000 rpm during the tests, keeping the CPU temperature at 60°C. After this first period (3+ hours), the "bug" appears and even when the CPU reaches 87°C the fans stay at 2000 rpm (however I can still "manually" control the cpu via smc fan control). If I reboot the laptop the bug is still there...
To me the problem doesn't seem to be related to the hardware... It looks more like a software design problem. It sounds the soft doesn't take the "good" sensor into account, but a sensor near from the cpu (by near I mean for example on the HeatPipe):
at the beginning the laptop is cool thus when loading a heavy computation the delta between the higher temperature and the lower one is significant enough to trigger the fans. (even on other sensors near from the cpu due to heat conduction)
But after a while the average temperature of the laptop is much more important than at the beginning thus the temperature delta is not sufficient to trigger the fans. (the heat conduction is not important enough to increase significantly the temperature on sensors located in an area near from the CPU).

FYI, I've got a mbp C2D 2,2GHz 2Go.

10-16-2007, 06:30 PM

Toms hardware guide did a test once taking the Heat Sink off a AMD Palomino and P4. The AMD burst into Flames and was dead in maybe 2 seconds. The Intel kept on going and powered down to a low power state and was not in any way damaged.

Let us know if you find any more info out

to ogFromBrussels, What what the max temperature it got to? 87c? That is a far cry from 129c!!!

10-16-2007, 06:35 PM

I did the same thing (well, very similar anyhow) to one of my tbirds... installed a heatsink that had feet (Thermalright SLK800) with a shim installed. Turned system on, it started to POST, then froze, I said "what's that burning sme...oh, (censored)." Yeah.

On the up side it gave me an excuse to go out and buy a then-cutting-edge 2200+ and an Abit KX7 >_>

10-16-2007, 06:56 PM

Originally Posted by dtravis7

to ogFromBrussels, What what the max temperature it got to? 87c? That is a far cry from 129c!!!

around 87-90°C, which is way too high without any fan activation!!! The value 129 looks more like a defficient sensor when submitted to high temperature. I just did an other test, I turned on my fans at 6000 rpm for a while. The measured temperatures were then equal to the initial temperature. Then I rebooted my laptop and did some super pi test... the fans did not activate anymore... Hardware problem... ???

10-16-2007, 06:57 PM

slk800, what a beautiful beast it was too

@dtravis - yeah, I remember that lol think it was something to do with the AMD not reacting quick enough. been quite a while since I read about that heck or even been on Tom's site.

Will the Intel program run in Parallel?..... lets find out .... nope..... ah well... I'll put XP on now, no time like the present I guess

wish me luck.... as in that my MBP does not do a fireball imitation