Re: Going to one HD
Date: 25 Aug 2005 05:57:42 -0700
Jean-David Beyer wrote:
>>Jean-David Beyer wrote:
>>>Why would you want to duct the hot PSU exhaust at the CPU?
>>Mainly because it simplifies duct design. Air blown out of
>>a duct has some directionality, so it merely has to be
>>directed toward the heatsink. Air sucked into a duct comes
>>in equally from all directions, so it needs to be nearly
>>a perfect fit. Also, heatsinks tend to be more effective
>>when air is blown downward at it rather than sucked upward
>Then why does Intel suggest blowing cool air horizontally over the heat
Huh? That only makes sense if the heat sink is of a
"tower" heat pipe design.
Note that almost all normal heat sinks have a fan on top
of it, blowing down onto it. (When viewed with heat
sink on top and CPU underneath it.)
Have you ever seen a stock CPU heatsink/fan? Why do you
suppose they all have fans blowing into the heatsink?
>>On two of my workstations, I don't even have a duct! The
>>air blowing out of the flipped fan PSU is already directional
>>enough to sufficiently cool the CPU.
>But is it cool enough?
Yes. A processor is cool enough if it never suffers from
the effects of overheating. These effects are very
noticeable--either hanging the computer or throttling.
In the case of severe overheating (i.e. the heat sink is
improperly attached or there's no heat sink at all), then
the CPU can even be permanently damaged.
>>For a single fan computer, the CPU and PSU usually have to share
>>airflow in series, and that means compromises.
>Not the compromise of running hot air over the processor.
What not? There is absolutely no reason you NEED to run
a processor cooler than you HAVE to.
>Here is how Intel think cooling should be managed.
So what? Silent computing almost by definition means breaking
the so-called "rules".
>>It makes more sense to put the PSU before the CPU than
>>putting the CPU before the PSU. The CPU typically tolerates
>>higher operating temperatures, and also typically generates
>>more heat than the PSU. Also, CPUs tend to have built in
>>hardware to throttle or shut down automatically in case of
>>overheating. A PSU's thermal safety feature is usually
>>limited to thermistor fan control, if anything.
>But then get a slower processor that will put out less heat to begin with.
Why should I, when I can get my current processor to function
perfectly fine in my quiet/silent computer?
It's worth noting that slower does not always mean less heat.
Some modern CPU cores actually generate less heat than their
predecessors even though they're faster also. For example,
P4 Northwoods were cooler than slower Williamettes.
Also, there's an important distinction between how much heat
a processor generates and how much temperature it can operate in.
>>For quiet computing, thermistor fan control is a mixed
>>blessing. If the PSU is sucking hot air from inside the
>>case, the fan will usually ramp up and ruin quietness at
>>even modest loads. However, if the fan is flipped so that
>>the PSU is sucking cool air from outside the case, the
>>fan will only ramp up under very heavy loads (if ever).
>If the object is quietness over reliability, why bother with fans at all?
The objective is quietness AND reliability. The best way
to acheive this, it turns out, is with a design that uses
at least one fan. Acheiving adequate airflow in a truly
fanless design is a difficult challenge, whereas the airflow
from even one silently undervolted 80mm fan radically
improves temperatures to reliable levels.
You are blinded by dogmatic assumptions. Silent computing
is simply not about following established recommended
dogma. It's about actually UNDERSTANDING the interplay of
heat and temperature and airflow, and creating new
solutions using that understanding.