1) Where does almost all heat transfer from CPU to heatsink?
At the center. If any thermal compound appears in outer half
of a CPU, then a human has applied far too much thermal
compound. As Jon notes, the most minimal film is a best
application. Thermal compound only fills microscopic holes.
Additional thermal compound only obstructs heat transfer.
Most heat transfers where heatsink makes direct contact with
CPU. Any thermal compound obstructing larger areas of direct
'CPU to heatsink' contact only makes CPU warmer. Thermal
compound makes a single degree improvement when using properly
machined heatsinks AND compensates for inferior heatsinks that
are not even machined. Thermal compound makes microscopic
holes conductive - which is why a minimal film is required -
and only in center half of a CPU.
2) From an Intel engineering paper on this subject - even no
thermal compound is quite sufficient for CPU cooling. Again,
most heat is transferred where the 'change of medium' is
less. 'CPU to heatsink' is more thermally conductive than
'CPU to thermal compound to heatsink'. This becomes obvious
once arithmetic in simple thermal equations is calculated.
Unfortunately most advocates of thermal compound have never
calculated thermal conductivity. They just know - numbers be
damned. As Conors notes -
This Artic X vs Artic Y vs some other brand etc is all
overhyped rubbish
Why? Too many recommendations come without first doing the
numbers. No numbers means junk science reasoning. Notice how
many thermal compound recommendations come with junk science
reasoning - or no supporting data whatsoever. It's called
propaganda.
3) Some also think a heatsink must be lapped perfectly
flat. Again, not true. But myths about heatsinks are so
common. Heatsinks, properly machined, maximize contact
between CPU and heatsink where the heat transfers. Proper
pressure - in pounds per square inch - must be applied which
defines how a heatsink is manufactured. Some heatsink
manufacturer engineering notes even chart PSI as it varied
across the heatsink face. But then this information is only
provided by serious heatsink manufacturers - who also provide
numerical specs. Some even explain why too much pressure may
cause less thermal conductivity. But again, demonstrated is
what one learns from serious heatsink manufacturers -
manufacturers that provide specifications even on the
product's box or tube. Again, many heatsinks are not
machined. And so they also don't dare provide the 'degree C
per watt' parameter.
4) A properly machined heatsink requires no thermal
compound. But when so many don't even demand the numerical
specs called 'degrees C per watt', then a good CPU
manufacturer best recommends every amateur use thermal
compound. Many amateurs then assume that means thermal
compound is always required - and they need not first learn
the numbers. Again, even an Intel engineering paper compared
heatsinks with various thermal compounds and no thermal
compound. Even with no thermal compound, thermal conductivity
was quite sufficient.
5) Some still use the Home Improvement concept of mass
application - "more power". Any thermal compound that leaks
out onto CPU pins is degrading CPU signals. It does not just
cause electrical conductivity. It also causes leakage due to
higher capacitance - which also degrades CPU signals.
Excessive compound may or may not cause intermittent CPU
operation. Just because one CPU had too much compound and
works does not prove anything. It is the same reasoning that
killed seven Challenger astronauts. Every thermal compound -
even low conductivity types - form undesirable capacitors
between CPU pins. Thermal compound must not get on any CPU
pins.
If any thermal compound leaks out between CPU and heatsink,
then the human is clearly a threat to reliable computer
operation. At a minimum, he did not read directions. More
typically, does not understand the concept or purpose of
thermal compound. Any thermal compound on CPU pins is totally
unacceptable - decreases CPU reliability. Compound is only
applied in center half where heat transfers.
6) How to determine if a heatsink is properly machined and
applied? Simple experiment. Apply heatsink with no thermal
compound. Measure CPU temperature on CPU - not on heatsink.
Then reapply heatsink, this time using thermal compound. If
CPU temperature drops more than 10 degrees, then return to
newsgroups to report a defective heatsink manufacturer.
Therein lies the purpose of thermal compound - to lower CPU
temperature by single digit degrees. If too much is applied,
then thermal compound spreads to outer half of CPU - where no
thermal compound should be. Too much thermal compound is even
worse thermal conductivity than no thermal compound.
7) Did Martin run 'the numbers' before posting a warning?
Numbers were posted previously. Numbers are what myth
purveyors never provide. If one cannot show a fallacy in
those numbers, then one has no idea what and how effective
thermal compound really is. Too many computer assemblers just
blindly worship thermal compound without first learning how or
why it works - and especially the numbers. Too many only
apply thermal compound because others say they should; reasons
for thermal compound be damned.
In the meantime, because so many don't even know how to
identify a minimally acceptable heatsink, then CPU
manufacturers would rather the amateur apply thermal compound
- so manufacturer does not hear about the amateur's mistakes.
There is no appreciable difference between most thermal
compounds - no matter how many myths (without numbers) are
posted in the newsgroups. Especially overhyped is Arctic
Silver which lasts just as long as cheaper brands from serious
heatsink and thermal compound providers - because too many
recommend without even first learning the perspective called
numbers.
