 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Today, ladies and gentlemen, I am performing a scientific experiment.
Yes, I realise how shocking it is that a guy working for a lab company
would actually be doing science. ;-)
It's not exactly the LHC, of course. No, it's rather more mundane than that.
The experiment:
Take one laptop. Place it in the server room for 12 hours. The server
room is kept at an icy 19°C at all times. The huge air conditioning unit
on the wall does that - as well as blowing away any bits of paper not
securely weighted down.
Turn on the laptop, and invoke the diagnostics tool from the BIOS
screen. Abort the testing procedure, and let it just sit there on the PC
health screen.
This screen gives me various readings. Among them are entries entitled
"CPU thermistor", "ambient thermistor" and "SODIMM thermistor". It
appears that the display only updates every 60 seconds, and there is no
way to make it update any faster, which is a pity.
The "experiment" then consists of writing down the numbers displayed and
the wall clock time, until the laptop switches off.
For today's run, the readings are as follows:
Time | CPU | Ambient | SODIMM
-------+------+---------+--------
12:22 | 41°C | 19°C | 20°C
12:23 | 61°C | 21°C | 25°C
12:24 | 74°C | 22°C | 28°C
12:25 | 83°C | 23°C | 31°C
12:26 | 87°C | 24°C | 33°C
12:27 | 90°C | 25°C | 35°C
12:28 | 93°C | 27°C | 37°C
12:29 | 94°C | 27°C | 38°C
12:30 | 94°C | 28°C | 39°C
12:31 | 95°C | 29°C | 40°C
...
12:51 | 97°C | 36°C | 47°C
Pro-tip: The temperature in the server room IS NOT 36°C. According to
our officially calibrated laboratory thermometer, it's 18.7°C (± 0.2°C,
for that is the calibration tolerance of this particular instrument). Of
course, I doubt that this "ambient thermistor" is located right on the
outside of the laptop casing.
Eventually the laptop turns off, and when you next turn it on, the BIOS
event log shows a shutdown due to a "critical thermal event". With the
CPU running at nearly the temperature of boiling water, this seems
unsurprising.
Initially the fan isn't spinning at all. By the end of the test, it's
reported as spinning at 5000RPM, which I imagine is its maximum speed.
The really weird thing is this: The laptop doesn't feel hot. At the end
of the last test, the underside was slightly warm. But not warm enough
to boil water on. Similarly, the air being blasted out of the vents at
high speed feels at best very slightly warm.
I gather that air in motion always feels cooler than it really is. But
I've had laptops that you could dry your hands with. For that matter,
I've had laptops that could cook your lap! This laptop isn't ICE cold,
but it's barely what I'd describe as "warm".
So... WTF? o_O
Anyway, I've noticed a couple of times now that the shutdown happens
just as the memory test is running, so I'm going to try again a bit
later and see if I can reliably reproduce this behaviour...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Le 11/05/2012 14:24, Invisible a écrit :
> Today, ladies and gentlemen, I am performing a scientific experiment.
>
> Yes, I realise how shocking it is that a guy working for a lab company
> would actually be doing science. ;-)
>
> It's not exactly the LHC, of course. No, it's rather more mundane than
> that.
>
> The experiment:
>
> Turn on the laptop, and invoke the diagnostics tool from the BIOS
> screen. Abort the testing procedure, and let it just sit there on the PC
> health screen.
>
> This screen gives me various readings. Among them are entries entitled
> "CPU thermistor", "ambient thermistor" and "SODIMM thermistor". It
> appears that the display only updates every 60 seconds, and there is no
> way to make it update any faster, which is a pity.
>
> The "experiment" then consists of writing down the numbers displayed and
> the wall clock time, until the laptop switches off.
If you can get a hand on a thermal camera, you might want to observe
that laptop with it.(looking for a hotspot)
Also, have you put the laptop on a wooden table ?
Have you tried leaving it 1 cm above the table (with only two linear
supports underneath, one per side ) ?
The ambient sensors might be on the bottom side, and on a wooden table,
it can rise.
> Anyway, I've noticed a couple of times now that the shutdown happens
> just as the memory test is running, so I'm going to try again a bit
> later and see if I can reliably reproduce this behaviour...
If the exhaust is cool enough, what about the intake ? are they opened
enough ?
Can you run that laptop while opened to see the fan ?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> The "experiment" then consists of writing down the numbers displayed and
>> the wall clock time, until the laptop switches off.
>
> If you can get a hand on a thermal camera, you might want to observe
> that laptop with it.(looking for a hotspot)
I would *love* to be able to play with a thermal camera. 8-)
I spoke to a guy the other day. His job - get this - he gets paid to
look at high-power cutting lasers using a thermal camera. How neat is that?!
> Also, have you put the laptop on a wooden table ?
>
> Have you tried leaving it 1 cm above the table (with only two linear
> supports underneath, one per side ) ?
>
> The ambient sensors might be on the bottom side, and on a wooden table,
> it can rise.
Yeah, first time was on my desk. Second time has balanced on a stack of
books with free air around the edges. I'm not completely sure where the
fan intake is though, it might still have been covered. And of course,
the body of the laptop would have been actually touching the books it
was resting on...
> If the exhaust is cool enough, what about the intake ? are they opened
> enough ?
>
> Can you run that laptop while opened to see the fan ?
I had to open the case to take the HD out and retrieve some files. I can
see the CPU heat-sink and heat-pipe. All appear to be in order. I can't
see the fan though. None of the visible vent openings look clogged. (The
laptop sits on a desk in a docking station and almost never leaves the
building. So... totally worth buying a laptop then!) I did try blasting
it with some compressed air, but that had no effect.
I might try running it with the case opened, and a temperature probe
pressed against the CPU heat-sink. (Not sure if that will cause
electrical problems though...)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Fri, 11 May 2012 15:12:52 +0100, Invisible wrote:
> I might try running it with the case opened, and a temperature probe
> pressed against the CPU heat-sink. (Not sure if that will cause
> electrical problems though...)
It shouldn't, as the heat-sink is attached to the top of the CPU using a
non-conductive adhesive. If it conducted electricity, that would add to
its heat load (most likely) and cause the temp to go up.
Quite possibly the laptop does not have a thermal sensor that's
calibrated to +/- 0.2C. It's possible the sensor has failed in some way.
Jim
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Time | CPU | Ambient | SODIMM
> -------+------+---------+--------
> 12:22 | 41°C | 19°C | 20°C
> 12:23 | 61°C | 21°C | 25°C
> 12:24 | 74°C | 22°C | 28°C
> 12:25 | 83°C | 23°C | 31°C
> 12:26 | 87°C | 24°C | 33°C
> 12:27 | 90°C | 25°C | 35°C
> 12:28 | 93°C | 27°C | 37°C
> 12:29 | 94°C | 27°C | 38°C
> 12:30 | 94°C | 28°C | 39°C
> 12:31 | 95°C | 29°C | 40°C
> ...
> 12:51 | 97°C | 36°C | 47°C
OK, so I just signed out our officially certified calibrated laboratory
thermometer and took some measurements.
The ambient temperature on the floor of our server room is 19.6°C. I
wedged the end of the probe into the exhaust vent of the laptop and
fired it up. The highest temperature recorded was 34.7°C. (At this
moment, the laptop was claiming a CPU temperature of 98°C.)
The ambient temperature in our meeting room (which has a nice, large
table) is 22.6°C. I opened up the case and pressed the tip of the probe
against the CPU heat-sink. I'm not 100% sure about how reliable this is.
Before I started the test, the laptop was still warm from the previous
step; I got an initial temperature reading of 25.9°C for the heat-sink,
slowly decreasing as I watched it.
I then fired the laptop up, still balanced on its side, with the exhaust
vent facing upward. The reading quickly rose to about 53°C, and then
wavered up and down from their. Sometimes it fell as low as 52°C,
sometimes it reached as much as 55.5°C. And all the time, the laptop
claims the CPU is at between 92°C and 98°C.
I did find that if I wedge the probe into a certain spot, I could get
the reading to go up as high as 62.8°C.
So, what do we think, folks? Is pressing the tip of a long metal probe
against a metal heat-sink a reliable way to get a temperature reading?
Is it really likely that the CPU could be at almost 100°C while the
surface of its heat-sink is at only 60°C or so? Is it normal for a
laptop heat-sink to get that hot? Should the exhaust be hotter than 35°C
when the heat-sink is so warm? Will Dell give a damn?
Actually, scrub that last one - I already know the answer...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 22/05/2012 10:37 AM, Invisible wrote:
> So, what do we think, folks? Is pressing the tip of a long metal probe
> against a metal heat-sink a reliable way to get a temperature reading?
I would use a thermocouple attached to the heatsink with thermal paste.
Once you have done that box the laptop up as the efficiency of its
cooling system will depend on the air flow.
BTW what is your calibrated laboratory thermometer which is a long metal
probe?
--
Regards
Stephen
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 22/05/2012 12:19 PM, Stephen wrote:
> On 22/05/2012 10:37 AM, Invisible wrote:
>> So, what do we think, folks? Is pressing the tip of a long metal probe
>> against a metal heat-sink a reliable way to get a temperature reading?
>
> I would use a thermocouple attached to the heatsink with thermal paste.
Yeah, I don't have any thermal paste.
> Once you have done that box the laptop up as the efficiency of its
> cooling system will depend on the air flow.
Well, that's why I tried it in our server room. The AC holds the room at
a constant temperature. That's also why I used the meeting room; it's
large enough that the presence of one warm human and one hot laptop is
unlikely to affect the overall temperature of such a huge volume of air
with free circulation.
> BTW what is your calibrated laboratory thermometer which is a long metal
> probe?
It's made be Hanna Instruments (no, I don't know the exact model number)
and I'm using a type-K thermocouple. It was last externally calibrated
by a certified engineer about 8 months ago.
Unfortunately, as I say, the probe for the thermocouple is very long and
thin. It would probably work great for measuring the temperature of the
air in a room, or of a flask of liquid. (That's probably what the lab
guys want it for.) I'm not sure how appropriate it is for measuring the
surface temperature of a solid object.
For reference, I pressed the tip to my leg, and got a reading of about
33°C - not implausible for the outside temperature of a warm-blooded animal.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Unfortunately, as I say, the probe for the thermocouple is very long and
> thin. It would probably work great for measuring the temperature of the
> air in a room, or of a flask of liquid. (That's probably what the lab
> guys want it for.) I'm not sure how appropriate it is for measuring the
> surface temperature of a solid object.
You need some of these:
http://www.eurodidact.co.uk/images/nicrni-01.jpg
That's what we use, you can then just glue or tape several of them onto
whatever you want to log (and leave 1 or 2 dangling around for the air
temperature).
A delta of 35°C between the CPU and heatsink surface might be plausible,
although it entirely depends on how much heat the CPU is dissipating.
If it's dissipating 35W of heat, that gives a thermal resistance of
1°C/W which sounds about right. I woudl have thought the exhaust would
be hotter than 35°C, but if you're doing all this with the lid off then
that's probably affecting the exhaust temperature significantly.
Besides, isn't your job to send it back to Dell when it's broken, not
try and repair it yourself?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 22/05/2012 01:35 PM, scott wrote:
>> Unfortunately, as I say, the probe for the thermocouple is very long and
>> thin. It would probably work great for measuring the temperature of the
>> air in a room, or of a flask of liquid. (That's probably what the lab
>> guys want it for.) I'm not sure how appropriate it is for measuring the
>> surface temperature of a solid object.
>
> You need some of these:
>
> http://www.eurodidact.co.uk/images/nicrni-01.jpg
>
> That's what we use, you can then just glue or tape several of them onto
> whatever you want to log (and leave 1 or 2 dangling around for the air
> temperature).
Hmm, yes, that looks far more suitable.
> A delta of 35°C between the CPU and heatsink surface might be plausible,
> although it entirely depends on how much heat the CPU is dissipating. If
> it's dissipating 35W of heat, that gives a thermal resistance of 1°C/W
> which sounds about right. I woudl have thought the exhaust would be
> hotter than 35°C, but if you're doing all this with the lid off then
> that's probably affecting the exhaust temperature significantly.
The exhaust measurements were taken with the laptop fully assembled, and
placed in a normal operating position. (I.e., resting on a flat surface.
I note the air intake is on the underside. Since it was resting on a
/hard/ surface, it should still be getting good intake.) This was also
in a server room held at a chilly 19°C.
The CPU heat-sink measurements were taken in a room a few degrees
warmer, obviously with the case off, and with the laptop resting on its
side with the exhaust vent upwards. I'm sure that's not a configuration
the manufacturers will be expecting you to try.
> Besides, isn't your job to send it back to Dell when it's broken, not
> try and repair it yourself?
Yeah, but then I'd have to talk to that Asian guy who doesn't speak
English...
Basically I want to determine whether the hardware is actually doing
something abnormal before I send it back. Although running the fan
constantly and reporting a CPU temperature of nearly 100°C when the
laptop isn't even DOING anything sounds fairly abnormal to me...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Basically I want to determine whether the hardware is actually doing
> something abnormal before I send it back. Although running the fan
> constantly and reporting a CPU temperature of nearly 100°C when the
> laptop isn't even DOING anything sounds fairly abnormal to me...
IME (which might be out of date by now) the BIOS does actually use 100%
CPU. I guess it just never puts the CPU to sleep, and loops waiting for
a keypress.
Have you checked the heatsink is properly attached to the CPU? That
would be my guess.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 22/05/2012 04:30 PM, scott wrote:
> IME (which might be out of date by now) the BIOS does actually use 100%
> CPU. I guess it just never puts the CPU to sleep, and loops waiting for
> a keypress.
Yeah, I've seen various DOS and BIOS programs that use busy-waiting
rather than suspending the CPU. So that /might/ be the cause.
OTOH, under Windows the fan quickly ramps up to maximum speed also. This
suggests to me that the problem is simply that the CPU is overheating.
> Have you checked the heatsink is properly attached to the CPU? That
> would be my guess.
The purpose of all the measuring was to determine whether the fault is
the temperature sensors reporting an erroneously high temperature, or
the cooling system failing to cool the CPU.
The heat-sink /looks/ to be attached securely enough. There is also
nothing obviously wrong with the heat pipe or the fans. It's quite a new
laptop, and I can't see any sign of dust build-in in the air vents. It's
all quite puzzling...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
