Given the length of this thread, and the lack of focus, I can
see why it is confusing...
In any design there are cost vs benefit considerations and the
point here is that if the cooling system is operational (as
opposed to an experimental design)
Right. Of course what we are talking about *is* an experimental
design. One of a kind, first ever... This is *not* a drop in
pre-engineered and beta tested system.
Hence the point here is not what you apparently have addressed.
That makes most of your comments invalid for what we have been
looking at, even though they are indeed correct for what *you*
are talking about.
then CPU temp provides the
most benefit (for a cost that's already included in virtually
all modern motherboards) with little, if anything, gained by
You missed the fact that we are discussing a system where the
*intention* is to reduce noise. That's the target: minimum
noise.
Which is to say, the goal posts aren't the same as it would
be for an overclocker or someone who simply wants to see a
water cooled cpu system functioning.
adding the expense to also monitor water temp. If the water is
Note that the "expense" to also monitor water temperature,
assuming that the system monitors *any* temperature, is minimal.
Temperature probes cost a couple dollars. It might actually
cost 3-4 more bucks to mount a probe (for example in a water
jacket). Overall, it isn't a significant expense.
'over temp' then the CPU will be too so CPU monitoring catches
the problem. If the water is within spec, however, you still
don't know if the CPU temp is. It 'should be' but monitoring the
CPU ensures it while water temp does not.
But monitoring the CPU won't help decide when and how to reduce
the fan noise associated with cooling the water! Since that was
the entire point of this design, we it makes no sense to leave
out that part of the design.
Using your logic of monitoring 'all the cooled parts' one could
NO, NO, *DO NOT* pin that crap on me. That was kony's statement,
which he then confuses because he doesn't realize how many
"cooled parts" there are! As I noted, the cpu is *not* the
only "cooled part".
also demand a radiator sensor as it's a 'cooled part' too. And
Smart idea! (There is hope for you!)
Given that it cost only a couple bucks and some wire, it really
is a good idea to measure the water temperature going into and
coming out of any heat exchanger intended to change the water
temperature. That difference is what can be used to control the
fan noise with fine enough granularity to make a "quiet" system.
(Of course, since I'm a geeky kinda guy, I'll spend another $5
and measure the air temperature on each side of the heat
exchanger too, just because it makes a really interesting graph.)
while that, too, might be a useful diagnostic tool for debugging
the design it's an added cost to an operational system that,
Nobody has discussed adding that to an operational system which
is *not* trying to be super quiet. (Which is what I assume you
mean, because otherwise your statement doesn't make good sense.)
again, adds little to the task of detecting a general cooling
system failure.
Nobody has claimed that it has *anything* to do with "system
failure". Other than kuny's strawman claim that failure
detection systems were somehow intended to replace control
systems. In addition to not being able to separate the *two*
cooling systems (one for the cpu and one for the water), he also
could not separate failure monitoring from control monitoring.
CPU temperature is sufficient to determine if it's doing the
job, or not, as that's the purpose of it, and all those other
intermediate 'cooled parts': to keep the CPU cool.
That simply isn't true. That isn't true even for systems that
overclockers want! It takes a significant amount of design
effort to balance the cooling such that each specific part that
is cooled gets exactly what it needs, and no more or no less.
Have you looked at some of the more complex systems? Water
jackets on disk drives and in power supplies, as well as on hot
chips!
There is no way to simply calculate that, throw a system
together, and end up with a optimal configuration. The
variation is in *how* it is monitored, not if.
For example, if software design is not part of the project (and
in fact not even all programmers are able to write the software
required, much less hardware hackers who don't do serious
programming), some of the manual probes might be used, or just
simple software, and only a few parameters measured at a time
over many tests to eventually compile a full set of data. It
will work just as well. But by the same token if complex
software design is part of the project, it doesn't cost
significantly more to monitor one or a couple *dozen*
temperature probes all at once!
Now, another purpose past diagnostic that monitoring other parts
might serve is in detecting an impending failure 'earlier' so
there's more time to shut down but I don't know that's of much
benefit to a typical PC.
It serves the purpose of controlling the noise making apparatus,
the fans, with much finer granularity.
I am assuming you're talking about a constant flow design and
not trying some kind of closed loop flow rate control.
You did assume something right!
I suspect from what you have said that if you adjust to the
actual target here, you'll see some differences.
Of course, there are also other ways to reduce fan noise, and
one of those is to *not* use fans to cool the water. For
example, where I live the outside temperature is always low
enough to allow convection cooling if the heat exchanger is
large enough. Hence I could mount a large enough outdoor
radiator and probably not even use a fan to cool the water. A
relatively low speed (quiet) fan could then be the only fan
needed to keep the case temperature down.
Others have done things like bury a surge tank in the ground to
accomplish the same thing. It works well. (Heh, I could bury a
tank in the permafrost, and end up having to *heat* the incoming
coolant rather than cool it!)
In those cases the initial stages require metering, but once a
good data set is obtained there is little need to continue
monitoring (other than for curiosity). But the same is not
really true for any system using a fan to provide cooling for
the liquid as long as the objective is to control the fan for
minimum noise.
