Thanks.
Quote:
The horizontal run of the coolant doesn't matter (within reason, with a really long run you start to factor in pressure losses due to the run), what matters is whether you are *lifting* the water -- that means pressure is required. You are just moving water in this application, with a relatively short run and without lifiting it much (if at all).


Almost no lift at all. I didn't know at what point resistance of the hose walls would factor in, but I won't worry about it. I saw one setup that used corrugated hose -- would make bends nicely without crimping/collapsing -- but I thought that would both flex more under pressure and would present more sidewall resistance. I'm thinking of using the type of polyester cord-reinforced PVC tube used in the water system on my boat -- about a dollar per foot.

Quote:
If you are going to roll your own, I would recommend the quietest fish tank pump you can find. Get a pump for a larger fish tank, as it will probably have lines that are easily adapted to the size you will be using on your computer.


I looked at several of these, but they were all AC-powered. If I have some ambition of variable flow rate, not sure how I would govern them.

Quote:
When you are dealing with cooling, the thing you want to do is dissipate *power*. Knowing how many watts your CPU(s) consumes (when overclocked) can give you an upper limit of how much heat you need to dissipate (to be conservative, assume that all the electrical power is turned into heat). Convert the watts into something you can use to size your radiator pipe, like BTU/hour.


I won't overclock, but I have a clampmeter and a broken-out IEC power cord, so I may just put some fans on the CPUs initially, run a stress test and see how many amps I can chew up.

Quote:
What you want is a system that dissipates heat at at least the same rate (or faster) than the CPU generates it. In my humble opinion, since CPUs are generally just fine to +50C, there is no reason to use active cooling. If you can cool it to room temperature, that is plenty good enough. If you can keep the cooling block to room temperature, or even somewhat above, you're fine. This eliminates the need for active cooling and prevents any condensation/dew point issues.


Yes, I want as few moving/noisy parts as I can get away with, so if I can get by with just a pump using something like the baseboard radiator or a Zalman Resorator, that's the plan.

Even without active cooling, though, I was concerned about dew point/condensation issues. Average winter temperatiures here are pretty mild, but it can dip into 20s-30s Fahrenheit and I was concerned that if if the pump ran at a constant speed, I might start circulating coolant cold enough to create condensation. My thinking would be to slow circulation down a bit, let CPUs keep coolant warmer. There's probably a flaw in my logic there re: coolant then spends more time in the loop out of doors (getting colder).

Quote:
You *will*, however, need to move coolant through the system fast enough to keep the cooling block cool. You will also need to cool the coolant, which is where the radiator business comes in.

You pretty much have 3 options for cooling the coolant: passive convection, active convection (using a fan), and some kind of refrigeration. Its really a question of surface area. Those 3 options are arranged in decreasing surface area requirement. Completely passive is just fine, doesn't require any additional power, and is silent, but it is going to require a lot more surface area than if you blow a fan over it, or if you use some kind of refrigerant.


I am thinking that there are more reasons to go the heater core route with, say, a constant flow rate but a fan on the core that switches on and off depending on coolant temperature coming into the PC.

Quote:
So, for cheap, easy and silent, I would use a long length of tubing with fins on it (like the home heating radiator you suggested). To test things, stick your little CPU cooler on a bright, halogen light bulb (100W+) and measure the temperature of the coolant. You need enough radiator and fast enough flow rate to dissipate the heat.

So, that's my $.02. Cooling is a function of coolant flow rate and the size of the radiator. Of course, the heat transfer rate of the little thing you stick on the CPU is a factor, but I'm assuming this is a constant. From a practical point of view, it would be really easy to build a radiator that is "overkill" for the application. Then, the faster you pump the water, the closer to room temperature you'll be, limited only by the CPU cooler's ability to conduct heat away from the CPU.



If my concern about cold weather condensation is legit even with a passive system, the more I thought about it the more variables seemed to come into play -- CPU temp, coolant temp, pump speed -- if I wanted to control coolant flow in a passive system. Unless I started programming a PIC now to manage all of that, I figured I would build something that could have a really bad see-saw feedback effect -- speed pump up, slow pump down. Sooooo, I am thinking now of that heater core approach that would probably function passively during colder weather but with a big fan that would switch on when/if CPU temps hit a certain threshold.

One thing that surprises me a bit: I would have expected to see some of these water blocks have coolant temperature sensors on the block or the outlet. Haven't seen that yet.

Thanks again!
_________________________
Jim


'Tis the exceptional fellow who lies awake at night thinking of his successes.