overheating help

On Mon, 9 Feb 2004 10:32:36 +0000, Mike Tomlinson
wrote:

I think w_tom is either a 'bot or an idiot savant. Once you get past
his hand-waving and attempt to pin him down on specifics, he very
rapidly goes quiet. If you google for his posts, he posts almost
exclusively to threads such as this about cooling (and in particular
threads which refer to lightning protection and/or power surges, in
which he recommends solutions that are appropriate to his location,
North America, but useless for UK/Europe.)

He pops up in the most unlikely places as soon as there is a mention of
the above subjects. w_tom, if he exists, might be running a scanbot for
mention of overheating, surges, etc, and then jumps in with both feet,
dispenses his 'wisdom' and leaves, unwilling to be challenged in any
way.

The man's a ****, and I killfiled him a long time ago.
--

Dave

, perhaps ?

I think w_tom is either a 'bot or an idiot savant. Once you get past
his hand-waving and attempt to pin him down on specifics, he very
rapidly goes quiet. If you google for his posts, he posts almost
exclusively to threads such as this about cooling (and in particular
threads which refer to lightning protection and/or power surges, in
which he recommends solutions that are appropriate to his location,
North America, but useless for UK/Europe.)

His favourite phrases are "show us the numbers" and "junk science" but
he never has any useful data to contribute himself.

He has also been asked repeatedly not to top-post.

In short, yes, he is a troll.

This has been my experience. I made the mistake of trying to hold a sensible
discussion with him about the use of thermal pastes versus heat pads, but it
quickly became obvious that he doesn't actually build PCs. He advised one
poster to try and run without either heat pad or paste because he feels the
heatsink manufacturers should provide mirror image lapped surfaces. He
doesn't think a 5c reduction in CPU temperature is material! Tell that to an
overclocker (like me). I did ask him to give us the benefit of his
experience by providing some real life figures based on his experience (as I
did with him), but surprise, surprise.....nothing.

I don't get caught up with him now.

Martin

w_tom wrote:
| Concepts are demonstrated by application notes from
| heatsink manufacturers. First bottleneck for thermal
| conductivity occurs at each medium transition. A transition
| from 'CPU to heatsink' creates thermal resistance. That
| thermal resistance increases if the transition is 'CPU to
| thermal compound to heatsink'. Another major interface
| involves the 'heatsink to air' interface.
|
| Major thermal bottlenecks occur at these interfaces.
| Pentium II and III "passively" cooled systems would reduce
| thermal resistance by placing power supply fan so that air
| flowed over CPU heatsink. Again, does not matter whether fan
| blows or sucks. Important parameter is a number called LFM -
| amount of air that passes across heatsink surface. As airflow
| increases, its cooling effects decrease exponentially meaning
| a large increase in airflow provides little increase in
| heatsink cooling.
|
| To minimize 'heatsink to air' interface, we now use HSF
| (heatsink fans). Massive airflow makes a minor increase in
| heat removal. Since a CPU fan provides maximum air flow
| around and across a heatsink, then no external airflow will
| contribute anything useful.
|
| Manufacturer provides one number to summarize this entire
| assembly - all three above paragraphs: degree C per watt. As
| an example, HSF of 0.5 degree C per watt is proposed. Finally
| some numbers.
|
| Now for those chassis fans. Using a single 80 mm fan, the
| typical system might see a 9 degree C difference between
| incoming and outgoing temperature. Two fans might result in a
| 5 degree C difference. Seven fans might result in a 1.3
| degree difference. Let's assume the CPU is located at hottest
| spot inside that chassis. Let's assume no heat is removed by
| metal chassis. IOW we worst case everything for a typically
| sized system. Even with only one fan, CPU will experience 9
| degrees C above room temperature air. A 75 watt CPU will be
| 25 degrees warmer than adjacent air. With one fan, CPU should
| be 34 degrees warmer than room temperature. A 70 degree F
| room should leave CPU at, worst case, 54 degree C using one
| fan or 47 degree C using seen fans. IOW CPU is 36 degrees C
| below maximum normal temperature with one chassis fan or 43
| degrees C below maximum normal temperature with seven fans.
| More fans provide virtually no improvement.
|
| The OP complains of a 70 degree C CPU temperature spike.
| Will more chassis fans solve this problem? Of course not as
| numbers demonstrate. Even when using a rather average 0.5
| degree C per watt heatsink, CPU should never be that hot.
| Solution is not found with more fans. Increasing one fan to
| seven provides no significant temperature reductions. The OP
| typically should suspect a problem somewhere in the CPU HSF
| assembly where bad thermal interfaces traditionally cause
| major temperature increases and are easily created.
|
| Was this 70 degree spike really a temperature increase?
| Unanswered were some important questions such as where
| temperature was being measured and by what? Also how long
| were these temperature spikes? For some reason, CPU is
| getting 15 degrees C hotter than even a worse case calculation
| - assuming the heatsink fan assembly was average. Reasons for
| an unexplained 15 degree C of heat must be identified. More
| chassis fans will not create a 15 degree C reduction, which
| leaves a problem within CPU HSF assembly or with how
| temperature is being measured.

No-one is trying to argue against scientific fact, here. We all agree on
the mechanisms involved and recognise the problems of transferring heat
across a materials interface.
The aim is to try to address OP's problem as originally posted, and
bickering will not help that.
Contributing something which is actually helpfull and move things forward
will.
On that score - either put up or shut up.
Kevin.

Kevin Lawton wrote:

snip other useful suggestions

I would also suggest that the OP temporarily re-sites the PC in question
in a relatively cool and open situation and observes the temperatures
reported by the system. Consider this to be an experiment to eliminate (or
otherwise) the possible adverse environmental effects of the PC's current
position (in a corner where the air has nowhere to go). If figures are
required for this experiment, then I would suggest the PC be placed on a
table approximately 0.8 to 1.0 metre high in a room at between 18 and 20
degrees celsius.
Kevin.

Definitely

Has it been established whether the reported CPU temperature is even
accurate?
Maybe it's time to borrow an IR thermometer and see if the sensor or
software is accurately reporting the temperatures.
I'm assuming we don't want to be drilling holes in the heatsink for
thermocouples

If there is an easy place to measure the *actual* core voltage with a
good DMM, this may not be a bad next move either...

Lee

--
To reply use lee.blaver and ntlworld.com

Provided is exactly what the OP required: numbers that
demonstrate where a problem and solution must exist. Numbers
that demonstrate what was posted previously: chassis fans -
even seven chassis fans - don't solve his problem. One
chassis fan is more than sufficient.

If numbers and concepts are too complicated, then simply
reread the last paragraph - in a post that only attacks the
problem and misguided science (does not attack any person):
Was this 70 degree spike really a temperature increase?
Unanswered were some important questions such as where
temperature was being measured and by what? Also how long
were these temperature spikes? For some reason, CPU is
getting 15 degrees C hotter than even a worse case calculation
- assuming the heatsink fan assembly was average. Reasons for
an unexplained 15 degree C of heat must be identified. More
chassis fans will not create a 15 degree C reduction, which
leaves a problem within CPU HSF assembly or with how
temperature is being measured.

Possible that a temperature spike does not really exist
which is why the above questions were posted. However I
suspect the OP has long since left since so much ill informed
speculations were posted as solutions; such as "More Air" or
"More Fans". Numbers (fundamental theory), real world
measurements using a thermometer, and even basic experience
tempered by knowledge all say this problem is not solved by
more fans and other such wild speculation. Posted is what was
repeatedly requested when I said, "provide numbers" to support
that 'more fan' speculation. Numbers say, and again, that
more fans and larger air holes are not a solution. Stick to
facts and numbers. Instead of attacking this poster (as other
junk scientists must do), please demonstrate a flaw in those
numbers (as a logical and unemotional scientist does).
Bluntly attack those numbers - not people.

Posted is an attack on outright misinformation - which, for
some silly reasons, you regard as a personal attack. The
numbers are what junk scientists never provide. Finally, in a
last post, you attempted to provide numbers. Good. But the
relationship between CFM and internal chassis temperature was
only speculation and was not correct. Your reasoning required
more advanced math - ie calculus or fractal type
calculations. Also in error was CFM for one fan. Number that
is obtained in fan manufacturer's datasheets. Posted, in
response were corrected numbers. Example: a typical 80 mm fan
should result in a 9 degree C temperature difference - worst
case.

Why do you attack the poster rather than learn how to
perform these calculations? Many 'computer experts' have
identified themselves for doing same thing previously -
posting junk science reasoning. Please don't be like those
other junk scientists - foolishly attacking the messenger.
Reasoning for how much a fan may cool was a honest, beginner's
attempt. But performed without having learned underlying
concepts. At least you finally tried to provide numbers.
Good. You could have also asked to learn what you (apparently)
did not know rather than, again, personally attack this
poster.

Reposted is a summary paragraph that focuses on the original
problem. Instead of attacking this poster, why not
concentrate on that paragraph? What do I keep going back to?
Addressing the OPs original problem by demanding numbers.
Calculations suggest a ballpark 15 degree C problem that
should not exist. A problem that obviously cannot be solve
with "more fans".

Kevin Lawton wrote:
No-one is trying to argue against scientific fact, here. We
all agree on the mechanisms involved and recognise the problems
of transferring heat across a materials interface.
The aim is to try to address OP's problem as originally posted,
and bickering will not help that.
Contributing something which is actually helpfull and move
things forward will.
On that score - either put up or shut up.

w_tom wrote:
| Provided is exactly what the OP required: numbers that
| demonstrate where a problem and solution must exist. Numbers
| that demonstrate what was posted previously: chassis fans -
| even seven chassis fans - don't solve his problem. One
| chassis fan is more than sufficient.
|
| If numbers and concepts are too complicated, then simply
| reread the last paragraph - in a post that only attacks the
| problem and misguided science (does not attack any person):
|| Was this 70 degree spike really a temperature increase?
|| Unanswered were some important questions such as where
|| temperature was being measured and by what? Also how long
|| were these temperature spikes? For some reason, CPU is
|| getting 15 degrees C hotter than even a worse case calculation
|| - assuming the heatsink fan assembly was average. Reasons for
|| an unexplained 15 degree C of heat must be identified. More
|| chassis fans will not create a 15 degree C reduction, which
|| leaves a problem within CPU HSF assembly or with how
|| temperature is being measured.
|
| Possible that a temperature spike does not really exist
| which is why the above questions were posted. However I
| suspect the OP has long since left since so much ill informed
| speculations were posted as solutions; such as "More Air" or
| "More Fans". Numbers (fundamental theory), real world
| measurements using a thermometer, and even basic experience
| tempered by knowledge all say this problem is not solved by
| more fans and other such wild speculation. Posted is what was
| repeatedly requested when I said, "provide numbers" to support
| that 'more fan' speculation. Numbers say, and again, that
| more fans and larger air holes are not a solution. Stick to
| facts and numbers. Instead of attacking this poster (as other
| junk scientists must do), please demonstrate a flaw in those
| numbers (as a logical and unemotional scientist does).
| Bluntly attack those numbers - not people.
|
| Posted is an attack on outright misinformation - which, for
| some silly reasons, you regard as a personal attack. The
| numbers are what junk scientists never provide. Finally, in a
| last post, you attempted to provide numbers. Good. But the
| relationship between CFM and internal chassis temperature was
| only speculation and was not correct. Your reasoning required
| more advanced math - ie calculus or fractal type
| calculations. Also in error was CFM for one fan. Number that
| is obtained in fan manufacturer's datasheets. Posted, in
| response were corrected numbers. Example: a typical 80 mm fan
| should result in a 9 degree C temperature difference - worst
| case.
|
| Why do you attack the poster rather than learn how to
| perform these calculations? Many 'computer experts' have
| identified themselves for doing same thing previously -
| posting junk science reasoning. Please don't be like those
| other junk scientists - foolishly attacking the messenger.
| Reasoning for how much a fan may cool was a honest, beginner's
| attempt. But performed without having learned underlying
| concepts. At least you finally tried to provide numbers.
| Good. You could have also asked to learn what you (apparently)
| did not know rather than, again, personally attack this
| poster.
|
| Reposted is a summary paragraph that focuses on the original
| problem. Instead of attacking this poster, why not
| concentrate on that paragraph? What do I keep going back to?
| Addressing the OPs original problem by demanding numbers.
| Calculations suggest a ballpark 15 degree C problem that
| should not exist. A problem that obviously cannot be solve
| with "more fans".

Okay. I am not disagreeing with your concept that empirical facts -
numbers - will help us to determine the cause of the OP's problem.
Unfortunately, the only numerical data so far supplied from the OP relates
to the CPU and m,/board reported temperatures. I realise that these might
not be wholly accurate due to the mechanism involved in measuring them.
Attempting to understand the problem from the point of view of materials
science and fluid mechanics is fine so long as you can suggest ways the OP
can provide the measurements you require. We have yet to see you explain to
the OP how to perform these measurements.
We seem to agree that, provided there is some sort of airflow through the
PC case, a correctly assembled HSF and CPU should run within a reasonable
temperature ange under normal conditions. I believe that this leaves us with
four alternatives:
1) That the deployment or assembley of the HSF onto the CPU is somehow at
fault. Either the HSF is not working correctly, maybe due to the fan not
spinning or a severe dust build-up, or heat is not being transferred from
CPU to heat-sink.
2) That the fan on the heat-sink is repeatedly cycling the same air
through it which is heating up, ie: there is no significant airflow through
the PC case. No matter how good the HSF is, it cannot cool the CPU below the
temperature of the air passing through it.
3) That at the time of the reported seventy degree CPU temperature, the
CPU was subject to an abnormal load and became hotter as a result. Examples
of what might cause this include, but are not limited to, overclocking
and/or increased core voltage.
4) That the measurement of the reported CPU temperature is at fault and
the CPU is not actually running at seventy degrees.
If yopu actually disagree with the above then please state so and why.
For some reason you keep using 'seven fans' as an attempt to dispute
anything I have suggested, though at no time have I advocated using a large
number of chassis fans in this case.
Actually, I have stated two points on this subject:
1) That the HSF will not be able to effectively maintain a reasonable CPU
temperature over a long period of continuous operation unless there is some
airflow. It will not work properly in a sealed box and some change of air is
required. To this end I have suggested that one fan should be more than
sufficient in most circumstances. From the perspective of getting some air
to flow, the fan in the PSU could be sufficient if it were to operate
continuously.
2) It is strongly desirable to ensure that the air entering the CPU case
is filtered in order to avoid a build-up of dust and dirt which will impair
colling efficiency and contribute towards overheating. Obviously in order
for this to work the air has to be encouraged to flow through the filter and
not enter the PC case through other openings. I did not suggest that a
seventy degree temperature spike was likely to be caused by dust alone.
If you do actually disagree with either of the above statements, then
please say why you feel they are not accurate.
The figures I gave were purely hypothetical and used to illustrate a
point.
I also stated that using an intake fan would ensure that the air entering
the case would pass through the filter.
My points are based on experience gained from working on PCs and other
computer systems over a long period of time. I am in no way attempting to
dispute or ignore the underlying science but I acknowledge the difficulties
the OP might encounter in providing accurate measurements of temperature and
airflow. Straightforward observation can be a usefull diagnostic toll under
such circumstances, and I have therefore asked the OP to make certain
observations and report back. If, for example, the OP obseves that the HSF
is choked full of dust and dirt it does not require a precise measurement in
order know that cleaning that dust and dirt from the HSF would be a good
thing to do.
Kevin.

do_not_spam_me wrote:
| "Kevin Lawton" wrote in message
| ...
|
||| Cutting out grills for better cooling is a waste of time,
||
|| Oh no it isn't !
|| Quite often case grills are just a few concentric circles of
|| small holes. Their total area might amount to less than two
|| square inches. An three-inch fan has an area of around six
|| square inches, and axial fans an't deliver much pressure.
|| The net result is that the little holes will both reduce
|| airflow by increasing air resistance and crate turbulence.
|
|| 1) Try running a case fan in free air without a grille.
|| Notice how quite it is.
|
| You are absolutely correct that it's quite -- quite quite, in fact.

Okay, my brain was going faster than my fingers could type and I made a
slight spelling mistake.
I was showing that a case fan run in free air without a grille is quiet,
quite quiet in fact.

|| 2) Try blowing with your mouth into free air, and then
|| at or through an empty PC case fan grille. Notice how much
|| noise the air makes as it impinges on grille. Notice also
|| how not all of the air you are blowing passes through
|| the grille, but bounces off the metal instead.
|
| It's good that you're not one of those people who obsesses about
| cooling for the sake of cooling and who performs proper systematic
| testing and doesn't use ad hoc methods that are subject to great bias.

In this case I was illustrating a point that the PC case fan grilles
increase noise by causing turbulance as they disturb the passage of air from
the fan. Precise empirical measurement was not absolutely necessary in order
to demonstrate the principle, which is that when a fan has to push (or pul)
air through the little holes of a typical PC case fan grille it can generate
noticable noise.

Are you disagreeing with me, or just taking the **** ?

Two things here...

It is not strongly desired to filter air into PC cases
o Air filters introduce a static resistance of 15-45Pa (clean-dirty)
o DC Tubeaxial fans as fitted to PCs are not suited to filter use
--- glancing at P-Q curves will show how airflow falls with resistance
--- resistance of even 15Pa will significantly reduce airflow
--- resistance of just 30Pa will essentially collapse airflow
o Fan depth &/or size must be greatly increased
--- compensate by more cfm for a given static-pressure
--- unfortunately noise will increase correspondingly

PCs do not require a lot of cooling
o CPU+Graphics+RAM/HD = 105+55+45 = 200W max at 100%
o PSU at 80% efficiency draws 240W, dissipates 40W itself
o 300cfm cools 1500W, 50cfm will cool 250W
o If a PSU fan is 25cfm, case fan 25cfm, done

More cooling can be required due to resistance
o If you fit filters, then greater-fan depth or cfm is required to compensate
--- you need 50cfm, you fit filters of 30Pa resistance (60% clogged)
--- from the P-Q curves you read the cfm at 30Pa resistance
--- then scale fan size / depth / qty to get back to 50cfm (at 30Pa)
o If your intake or exhaust grills are punched metal, consider a change
--- punched case slots/holes are often just 45-55% free air
--- round-wire-grills are 89-91% free air = more cfm & less noise
o Your cfm is determined by the weakest link - the intake
--- not usually the fan punching
--- more usually the slot in the plastic front fascia - 3"x1"
--- 3"x1" slot = 1900mm^2 vs 2x80mm exhaust = 9000mm^2

All are really quite minor changes, and good cases don't need them.

CPU cooler:
o CPU coolers are measured in degrees-Centrigrade-per-Watt (oC/W)
o The oC/W must be low enough to keep the CPU in reasonable limits

If the case is removing air sufficiently - the CPU will only be recirculating
its own air at the known design spec (re quoted oC/W). If the case is not
removing air sufficiently, the CPU cooler will recirculate more of the air it
has just heated and so the realised oC/W will fall re lower delta-T.

It is worth noting that some CPU coolers are over-rated re spec, this is
more common for AMD processors than Intel - altho that will change as
the P4-Prescott comes in (P4-3.2N cooler 45-46oC, P4-3.2P using the
same cooler was 64-65oC; some report m/b components 100oC too).

The problem can occur with over-use of heatsink compound
o Two flat surfaces only make perfect contact at 3 points
o Heatsink compound is just to fill in that ultra-thin gap
o Spreading it like margarine does not help heat transfer

The problem can also occur with incorrectly or marginally seated heatsinks.

A very considerable problem is temperature reporting accuracy:
o Most thermistors are +/- 10% in accuracy & calibration
---- actual 50oC -- one board may read 45oC, another 55oC - 10oC gap
o It is not inconceivable for thermistor to be vastly out of spec
---- both in calibration & accuracy, to a combined figure of 20%
---- actual 50oC -- in-spec board reads 45oC, out-of-spec 60oC - 15oC gap

Rarely - but consistently - some very odd temperature readings are reported,
either overly low or overly high relative to the normal distribution of reports.

So several avenues of attack, but ones based on cause:effect, cost:benefit.
Most quite high-spec PCs require 50cfm of exhaust airflow, unless someone
has fitted fan filters or has the exhaust fans sucking thro a tiny intake slot. It
is better to fix the intake slot than stick a NASA windtunnel on the exhaust.

Roll on BTX, at least there will be a proper OPEN intake grill in the case front.
Apple got it right, then stuck too many fans in the thing. G4 Tower got the size
of heatsink right, the fan size right, but then stuck a 150cfm 57dB(A) fan in.
--
Dorothy Bradbury

Dorothy Bradbury wrote:

Roll on BTX, at least there will be a proper OPEN intake grill in the case front.
Apple got it right, then stuck too many fans in the thing. G4 Tower got the size
of heatsink right, the fan size right, but then stuck a 150cfm 57dB(A) fan in.

Interesting post, thank you.

It is possible to have too many fans. One optional case fan I added in
started making a lot of noise, so I removed it, and the effect on PC
temp was virtually nil.

It is possible to have too many fans. One optional case fan I added in
started making a lot of noise, so I removed it, and the effect on PC
temp was virtually nil.

Your airflow (cfm) thro the case was sufficient for the internal wattage,
so adding more cfm had no benefit for that environmental temperature.

In summer, under full load, you might see a difference since the delta-T
between the environmental air & heatsink is less. It's quite small though
since the heatsink is /still/ much hotter than the environmental air.

Biggest problem with PC cases is their often crude punched grill,
and the even cruder air intake at the front in the plastic fascia. You
have 1900mm^2 or less hole, and 9000mm^2 min from 2x 80mm.
Fitting a side-front grill can make a big difference to cfm, so you
can use quieter fans for the same or even lower temperature.

PCs will become a pain when we hit 200W/CPU, re heavy duty
dual-CPU graphics box, multi-HD, PSU - ie, 600W room heater.
--
Dorothy Bradbury