Slashdot Mirror
Power Water Cooling Kits
msolnik writes: "Toms Hardware has but together a head to head comparison of 4 different water cooling kits. Instead of buying each part seperately these kits come with everything needed. I would love to use water cooling but there is just something about having water inside of my case that makes me very uneasy. But for all you hardcore overclockers out there this may help you out a lot."
If you use destilled water it doesn't conduct and you should be fine, even IF it leaks.
True, but the problem remains that if that water leaks, it is most likely going to pick up deposits off of the motherboard and become conductive. Think of all of the dust and grime that collects inside of a case as time goes by due to the power supply fans and whichever other fans happen to be in there.
I do know of one case where someone tried to submerge their entire motherboard, power supply, and daughter cards, but they used mineral oil instead of water. This would be over at Dr. Ffreeze's website.
But in terms of voiding warranties, possibly destroying hardware, and overheating delicate computer systems, I never really had the urge to do it. I just can't understand the reasoning behind overclocking a processor just to squeeze a couple of extra megahertz out of it, when it's nearly impossible to tell the difference between MHz these days.
I don't know where you got the idea that overclockers only eek out a few MHz? Most people overclock only when they can spend far less money on a chip that can be overclocked to perform the same (or better) than the more expensive chips.
The celeron 300a was a classic. You could easily overclock it to 450MHz, and it would perform on par with those 450-MHz rated chips, for a fraction of the price. Some people went way beyond 450MHz with them. With water cooling, crazier numbers are possible.
Keep in mind that most of the chips that are popular for overclocking are actually the SAME exact chips as their overpriced, higher-rated brethren. The difference is that they didn't cut the mustard off the assembly line, so they were set to run at lower/safer MHz levels. The overclockers just risk moving them back up on par with the higher-rated chips.
"And like that
the water - cooler is (nominally) letting you use a quieter fan, because you're getting the effective cooling-area of the the larger heat-exchanger, which can be located outside the case
Linux is Linux, if One need clarify their dist: <Dist>/GNU Linux
bsds are of course just BSD
According to overclockers.com, THG tested these water coolers incorrectly. Apparently, THG tested the Swiftech unit w/o fans, relying on a "passive heat exchange", while on swiftech's website, the Swiftech Barebone Kit is CLEARLY shown with 2 120mm fans. Given that, it is amazing that the Swiftech finished so close to the Innovatek! It seems evident then, that Tom's conclusion is flawed. Swiftech makes great products (really the best high-end HSF maker, better than Alpha, but pricier), and I was dissappointed to see such a poor performance until I read about this testing anomaly at overclockers.com
eg: Many metals have a specific heat in the low hundreds. Water is in the mid-thousands. So, a one degree celcius temperature change in the water is in excess of a 10 degree change in the metal.
(This is why water-cooling is popular. Not because water is magic, but because finding a liquid with a lower freezing point and a comparable specific heat is a royal pain.)
Typically, what someone would do is add something to the water, to lower the freezing point. This almost invariably lowers the specific heat, too, but it's a decent compromise, usually. The water is then piped over the relevent components, dragging the heat with it. The water is then cooled by a second cooling system, often freon-based refrigeration. This takes the heat from the water, and (hopefully) gets rid of it fast enough.
The science of heat transfer is not impossibly complex, but it's not trivial, either. The heat and the specific heat are what are important. The temperature is merely a function of these.
(This is why liquid nitrogen sounds good, but isn't really used much. The specific heat is too low, so the temperature rises comparitively quickly. Useless for cooling, unless you have a LOT of liquid nitrogen and are piping it at a decent pace. That makes the pump more expensive, for a start, and would make it essential to use large reservoirs, which you'd need to keep refilling.)
Talking of cooling, there's a pelzier device over on one site linked to from extremecooling.org, which has a delta T of 140 degrees celcius. It's cheap, too. Dragging that kind of temperature off a chip could make cooling -seriously- fun. Again, you'd only need water cooling, to transport the heat, but you'd end up with a chip temperature about that of liquid nitrogen, without the expense, hastle or complications.
Oh, one other thing. Condensation is a killer, when supercooling. Usually, overclockers seem to just waterproof the relevent chip, but heat doesn't stay put. It dissipates. This means that heat will flow from hotter points to colder points. Which means that any water vapor in the air is not guaranteed to stay in the air.
Three possible solutions to this: First, strap on some kind of air conditioner for drying the air, so that there is no moisture in the case to condense. Might work, provided the case was otherwise air-tight. The second option is to not have any air in the case, in the first place. Again, make it air-tight, then set up a partial vaccuum inside. If there's no air, then there's nothing to hold the water. The third option has already been given a Slashdot article, and that's to flood the case with a non-conducting medium that does not mix with water, such as mineral oil. This will keep the water away from anything vital, and might actually help with the cooling effort, across the entire motherboard.
When dissipating the heat, however you build the case, don't forget to have a reflective surface between your dissipation system and the computer. Otherwise, half the heat goes right back into the case, and you're just baking the other chips.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
I'm not too sure I'd use the fan speed monitor that comes with the Innovatek... it looks like it could do some damage to a motherboard!
This image on page 3 of the article shows a 4-headed cable. Two connectors form a short disk drive power supply extension cord, with the two outer pins (+5v and +12v) tapped off to supply power to the fans. This trick allows the fan to run off of 12-5 = 7 volts (uh, they messed up the math, calling this 8 volts). These red and black wires go to what looks like a male connector to go to the fan. Then, a yellow wire comes from the fan and seems to go to a motherboard fan connector. That's the problem.
First, a little background. As I understand it, the speed feedback is essentially a switch that connects the yellow wire to ground two times per revolution. The motherboard must supply a little bit of power (in the form of a pull-up resistor) to actually get a signal out. This works well because, since the motherboard is supplying the power, it can make sure it doesn't supply too much voltage. If the chipset is 5 volts (impossibly rare nowdays), it'll supply 5 volts. If it's 3.3, it'll supply 3.3. This way, a fan can be used with any voltage chipset.
Ok, the problem is that the ground pin of the fan is connected to the 5 volt power supply. This is the trick used to get 7 volts. The fan will spin and connect the speed feedback wire to the 5 volt "ground". If you've got a chipset that uses 3.3 volts (most likely), you'll be feeding 5 volts into your 3.3 chip - a big no no that could burn out the speed input or the chip (which probably handles lots of other functions you wouldn't want to lose). The voltage on this pin will oscillate between 5 volts and 3.3 volts - not the 0 to 3.3 it expects. If you have a 5 volt chipset, then this pin will oscillate between 5v (shorted to "ground") and, uh, 5 volts (pulled up by the chipset)... so, while not doing any damage in that case, wouldn't work.
HIV Crosses Species Barrier... into Muppets