kony <spam@spam.com> wrote:
>On Tue, 12 Jul 2005 18:04:49 -0800, floyd@apaflo.com (Floyd
>L. Davidson) wrote:
>
>>You can literally calculate the amount of heat that can be
>>stored in any given sized reservoir of water, and from that you
>>can calculate just exactly how long it will take to raise the
>>temperature of your reservoir to any given point at which you
>>estimate that the cpu will not get sufficient cooling. (And,
>>once you get it installed you can measure that very easily.)
>
>Not true because even if one tried to prevent it, heat would
>be shed so long as ambient is less than water temp.

The question is, however, how fast? The answer: not very.

>>If there is no air movement across the radiator, the reservoir
>>is not going to shed any of the heat ...
>
>One has nothing to do with the other.

I am considering "the reservoir" to be the entire coolant
container, including the radiator, pump, and the hoses too.

Essentially the only way that heat will be dissipated is by
moving air across the "radiator"... which might be a separate
heat exchanger or might be incorporated into the physical
construction of the reservoir. (The URL the OP posted is in
fact a combination reservoir/radiator, all in one unit, though
there may indeed be another heat exchanger too.)

The point is that the water temperature of the reservoir,
lacking some system to remove heat, is *not* going to dissipate
the heat it gets from the CPU. It is going to store it.

>>(or, at least very little
>>of it). Hence it is not a matter of if it will rise to a
>>temperature where your cpu is not getting sufficient cooling,
>>but just a question of how long. (Note that it is just a bigger
>>heat sink than is a regular air cooled heat sink. The air
>>cooled device is so small that with no fan it only takes seconds
>>to overheat the cpu.
>
>Hardly, unless "seconds" means the better part of a minute
>or the system was so poorly set up that it was barely under
>the stable-temp margin while fan cooled.

Some CPU's will overheat in less than a hand full of seconds.
Others may take a minute or two. I don't recall if the CPU was
specified??

>>With a water reservoir, it takes minutes
>>or even hours if it is large enough... but it still overheats.)
>
>No, the reservior has little impact on it unless it's HUGE,

That simply isn't true. As I noted, you can calculate exactly
how much heat any given sized reservoir can absorb for a
selected temperature rise.

The *only* way that a water cooling system will work is if the
heat removed from the reservoir is equal to the heat removed
from the CPU. The CPU of course has to have heat exchanged at a
fairly fast rate, while the reservoir size determines how fast
heat has to be removed.

>and even then, only to the extent that a huge reserviour has
>high water-surface contact. It's not practical to plan for
>a huge reservoir, if someone has THAT much free space they'd
>be better off just adding a 2nd radiator.

Depends on what the point of water cooling is. If it is lower
temperatures, you are correct (and the reservoir need not have
any capacity larger than a very small surge tank). But if the
design target is lower noise, then reservoir capacity is
significant.

>>Because of the above, it would seem there is no point in trying
>>to start off with no fans at all, unless you have some other
>>arrangement to remove heat from the reservoir. It sounds as if
>>you are talking about a relatively small reservoir and a system
>>entirely enclosed in the computer case, which suggests that it
>>will perhaps take many minutes to heat up, and many hours to
>>cool off again. Which is probably not an acceptable recycle
>>time.
>
>A system should NEVER be planned based on "recycle time".

Well, that just depends on what the design target is! With
noise reduction, it *can* be based on recycle time.

>>The problem is that you need a way to *know* what is going on!
>
>Actually, that's 100% completely wrong.
>A properly engineered system has no need to inform the user
>of anything, no need for the user to adjust anything, etc.
>The only thing the user should ever need to know is that if
>the system shuts down, it did so before it overheated enough
>to cause damage.

The "user" as such doesn't need to know anything. But the
system that controls it *does*. The OP is talking about a
manual system, in which case *he* is the controller that needs
to know.

My whole point was to convince him that since he would need to
know a lot more, he might as well automate it and hand it all to
a controller.

>>That means temperature probes in cooling system.
>
>Ridiculous waste of time.

Only *if* you don't mind the noise of fans operating at maximum
(required) capacity. That is typical of systems designed for
overclocking, because the entire idea is to get minimum CPU
temperature. But, that is *not* what the OP wants. His purpose
minimum noise.

To have anything other than gross (on/off) control of the fan
requires temperature probes of the coolant.

>>The probe in
>>the cpu is fine, but it doesn't tell you what the status of the
>>reservoir is, and hence won't give you a clue about what the fan
>>needs to be doing.
>
>No, completely wrong. The fan control should be based on
>the hottest part if it's based on anything.

You are designing a minimum CPU temperature system with a fail
safe. That nice and is interesting, but is *not* what the OP
wants.

>Multiple parts
>with a relay control if it were to be more elaborate, but
>NEVER based on water temp. Suppose pump breaks, or water
>leaks, the water temp will not be as high because the
>components being cooled are retaining more heat.

That has *nothing* to do with fan control.

And fan control has *nothing* to do with shutting the system
down due to catastrophic failure. Note that I discussed some of
the functionality require for that too, farther down in my
article (and which has been snipped).

I didn't get far into that, but if the OP wants to discuss it
farther, the various requirements for failure protection are
interesting too. But the requirements depend on information the
OP has not provided yet.

>>Basically these devices allow you to have a fully metered
>>cooling system. It is possible to measure coolant temperature
>>in the reservoir, after the cpu waterblock, and after the
>>radiator, and air temperatures in the case and on each side of
>>the radiator, plus outside and room temperature too if you want!
>
>Which is all trivia. In a water cooled system, the only
>temps that should be weighed in determining fan control are
>those of the water-cooled components. Doing anything
>differently than this will only increase the margin of
>error.

Please realize that this is *not* a typical overclocker's system
design. He wants to reduce *noise*. Rethink what is allowed to
vary (the CPU temperature), and what is required to have a low
"margin of error" (fan noise!).

>However, if the chassis exahust fan is cooling the radiator,

He wanted to have a minimum number of fans, all running as slow
as possible.

>and that fan is also being controlled, then all passively
>cooled components much also be weighed in the determination
>of proper fan speed control, but still, NEVER the water
>temp. Water is absolutely the very last thing anyone should
>care about until it starts boiling- and if your water is
>boiling you have far larger problems than which fans are
>spinning.

That is just not true, given the design targets. (Not to
mention an absurd scenario, given the typical 70C shutdown
temperature, and the 50-55C target mentioned by the OP.)

--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@apaflo.com