Blowing air out the side panel means you are sucking in pre-warmed air
over the CPU before exhausting it. Blowing air into the case directly
at the CPU means you get the greatest temperature differential. Since
the PSU is nearby, the air warmed by the CPU gets drawn out and
exhausted through the PSU rather than over the memory sticks or
chipset.
Blowing air in (or having no side fan at all) results in the
most pre-heated air flowing over the memory or chipset.
Granted, the air might be slightly cooler.
The key to reducing warm air flowing in circular or
counterproductive directions is to not interfer with the
time-tested and proven chassis airflow pattern from bottom
front to mid-top rear.
By reducing the bottom front intake rate, by use of a side
intake if not another method, there are lower velocity flow
everywhere except into the 'sink. The exhaust fans will
exhaust at same rate (providing front intake was sufficient)
but a short-loop is created, any air not exhausted into the
air most immedate to the exhaust fan will take a longer path
till exhausted, and slower.
Side intake is mostly Intel's attempt to ship cheaper
heatsinks with P4, particularly Prescott CPUs. A properly
set up system has no need for the side intake and is usually
as well off having it blocked completely. The CPU is not
the only part that needs cooled and the heatsink selected
for it should be selected so as to do the job acceptibly
without other parts running warmer as a result.
Remember that while a greater temperature differential, using
water-cooled systems, peltier setups, or whatever that you are simply
increasing the possibility of cooling your components as much as
possible. You can't cool them faster than the thermal interface
allows;
Untrue/non-applicable. The thermal interface doesn't allow
or disallow, it is only a decrease in efficiency of transfer
from the warmer to the cooler part. This warm vs cool is
the key element as mentioned below.
i.e., you could easily use excessive cooling beyond which the
heat cannot transfer across a metal plate any faster, ...
False. If the cooling isn't much better, the difference may
not be significant change in temp (only a little), and the
CPU may not have needed to run cooler at all, BUT it will
run cooler with a better cooling system and exact same
thermal interface, because the better cooling results in the
cooler of the two parts (warm vs cool), being cooler than it
would have been. Thus, the thermal compound with the same
thermal transfer efficiency, is seeing a higher temp
differential between the two parts which must necessarily
result in a cooler CPU (all else being equal, same thickness
of thermal grease, flatness of mating parts, contact area to
top of CPU, etc).
... so the extra
cooling is wasted.
It might easily be wasted money, time, even materials to
throw away the stock sink (and years(?) later this 2nd sink)
if there wasn't a specific *need* to get the CPU any cooler
(like in an extremely hot environment or high overclocking),
if it ran cool enough to be stable and have acceptible
lifespan, OR if the improvement in the cooling was minimal,
such a slight drop in temp that it varies more by ambient
room temp than anything else.
Similarly, for air cooling, buying a CPU fan that
is rated 50 CFM won't push that amount of air through the restrictive
air space between the heatsink fins (the fins cause turbulence and the
airflow has to take a 90-degree turn both of which equate to
resistance).
A higher free-air rated fan will not push the free-air
volume through a heatsink, that much is true. On the other
hand, given a similar fan (dimensionally matching the other
fan and heatsink and at least as thick) that has the higher
CFM from higher RPM, that will result in more airflow
through the 'sink. Perhaps not a lot more- these fans have
relatively low static pressure potential, but even a little
more, will directly correspond to a reduction in heatsink
temp, which as mentioned above, will ultimately cool the CPU
more.
Increasing airflow has a quickly diminishing return though,
Seldom is it worthwhile to have a fan faster than around
3000 RPM unless a special circumstance or very poor fan
(typically undersized or at least too thin for the
application). Often it is desirable to have even lower RPM,
it can be acceptible to have CPU slightly hotter to
significantly decrease noise if the margin allows for it.
You might find a 25 CFM fans cools just as well as a
much noisier 50 CFM fan.
If all else were equal, no, it is impossible. If the
diameter of fan or thickness, or even an entirely odd
proprietary fan design were used on same heatsink, then
these other factors would have to be considered as well, but
taking same exact fan one can vary the voltage to produce
different RPM -> airflow (airflow will increase so long as
the fan isn't terribly undersized and already at it's
practical maximum flow rate into the pressurized area in the
'sink) and see the temp change. It may not change a lot
but it is an inescapable scientific truth that to whatever
extent this air (which is cooler than the 'sink) flow is
increased, there will be a corresponding decrease in 'sink
fin temp, conduction away from the base of the 'sink to this
lower temp region of the fins, and through the thermal
compound to the CPU. The conduction efficiency does not
change with the fixed variables (grease, heatsink metal) but
across mediums (CPU to 'sink or 'sink to air) the higher
temp differential does always result in lower temp.
Naturally if the temp difference is minor, especially if
being reported by a motherboard-mounted sensor, it may be so
slight a change as to be indiscriminable. I am not
suggested anyone improve their cooling at all, unless it is
overheating.
In fact, I've seen CPU and case temperatures
drop in some cases by reversing the backpanel fan so it is an intake
fan rather than an exhaust fan (its normal position) but you need to
test in your own case.
This should never be done, it necessarily increases temps of
other parts (unless the case was otherwise unusual in it's
airflow before this fan was flipped over). Cooling a CPU is
not a game where lowest number wins. Anyone who has a
stable system and CPU that doesn't die prematurely, "wins",
so long as they deem the noise level acceptible. However,
while keeping the CPU cool enough, focus is then shifted to
the rest of the system - a CPU is more heat tolerant than
many parts.
.