Slashdot Mirror
New 3D CPU Water Cooling Method
captain igor writes "According to this story on Wired News, a new company launched by researchers from Stanford has come up with a way to layer a silicon network of tiny tubes on top of a microprocessor. The system then uses a solid-state motor (no moving parts!) to pipe cold water through the silicon network. According to the article, this system can handle 1000 watts (yes, a kilowatt) per square centimeter."
No. A motor is by definition "one that imparts motion". This device certainly qualifies.
http://dictionary.reference.com/search?q=motor
Seems a perfectly good use for the word 'motor.' If you're going to be a pedant, at least try to get it right.
No, actually, they're not because the G5 is excessively hot, nor are they for show. They are for maximizing the efficiency of the 9 (VERY low speed) fans in moving heat out of the system with minimal airflow
People assume that because the G5s have a extremely well-engineered cooling solution that the G5 is also extremely hot. It's simply not true, it's all about noise reduction.
Random and weird software I've written.
Say water goes in at 30 degrees C and comes out at 50 degrees C. According to the spectacular Google calculator, 1000 watts is 239 calories per second, and it takes 1 calorie to increase the temperature of 1 cc of water 1 degree C, so you'd have to move 239/20 or about 11 cubic centimeters of water through the cooler every second assuming a delta-v of 20 degrees C. Doesn't sound unattainable.
thad
I love Mondays. On a Monday, anything is possible.
The factor they always leave out is how much of a temperature rise one can tolerate at the heat sink. Let's assume that the incoming water will be no higher than 40C and the CPU can become no hotter than 60C - that's 20C rise.
1 kilowatt is 1000 joules per second, or 238 gram calories per second. Conveniently, a gram calorie is the energy needed to raise a gram of water one degree celcius. For water, one gram is also one milliliter. So, a single gram of water will be raised 238 degrees C in one second. We don't want it to be raised more than 20C, so we need to exchange water at a rate of 238/20 = 11.9 mL/sec.
Heat sinks aren't perfect - the outgoing water will always be colder than the CPU. Let's pretend that this sink is 50% efficient (the CPU rises to a temperature, relative to the incoming water, of twice that of the outgoing water). Ergo, we need 23.8 mL/sec.
How is this a problem?
OK, let's see here, first off, most stove top burners are closer to 2000 to 3000 Watts, not 1000.
s /c omparisonchart.html
i e= UTF-8&oe=UTF-8&hl=en&btnG=Google+Search&me ta=
http://www.consumersearch.com/www/kitchen/range
But regardless, your analogy is much too much work, let's just figure out how much water you can boil each second with 1KW of power.
This page:
http://www.infinitepower.org/calc_watts.htm
Says you can evaporate 0.0001172 gallons each second. According to Google, this works out to:
http://www.google.ca/search?q=0.0001172+gallon&
Or about 0.44 mL
So as long as you keep more than 1/2 a millilitre/second flowing over your square centimetre you won't be boiling. Of course to be safe you'd probably want a lot more than this.
That is a pretty good off the cuff Spherical Cow analogy. I would suspect that the network of silicon tubes uses something akin to counter current flow to achive higher rates of cooling.
Also, with a small network of tubes the relative surface area of the water to the heat would be higher than a teaspoon on a stove. While this probably means that the water would vaporize more quickly this might not be a bad thing. There was (is?) a company that produced PC cases that contained a compressor and supercooled liquid systems that operated on the vapour change principle. (I cannot for the life of me remember the company - they used to supercool the first Athlons.) When water changes into water vapour it absorbs more energy undergoing the phase change than when it is in liquid water alone. Also, water vapour can be moved about more quickly than water as it is less dense. Therefore, I suspect that if this is not a typo in the article, that liquid water is pumped into the chip, and it undergoes a phase change to steam which can only exit the chip in one direction - thus adding to the pressure in the system. The evaporation of the water would cool AND help push the coolant throughout the cycle.
So that is my guess - countercurrent flow to maximize the amount of heat removed from the chip, and the phase change of water to absorb more head and provide power to moving the water about the system.
Try to hack my 31337 firewall!
They claim the potential to move 1kW through this surface, but they don't mention the conditions.
If you make it really cold on one side, and really hot on the other this could happen by itself.
Think of your cooler, it doens't leak heat much on a cold day, but on a hot day it will warm up much quicker.
Change your temperature difference, the heat flow rate will change.
On your boiling water, take steady state water evaporation vs energy input. Your 1kW Burner isn't going to be boiling thousands of teaspoons per second. You have to heat it up to the boiling point, then apply the energy to vapourize it.
Energy=Power*time = Mass * Heat of vapourization +Mass * Temperature Change * Heat Capacity
More people should take physics.
The engineering problem is getting Thot to be the microprocessor temperature, not the exiting cooling water temperature - this would give you much better efficiency, but at the possible cost of cooling power.
It's not wasting time, I'm educating myself.
You're ignoring the convective heat transfer coefficient for water.
The heat transfer rate is a product of the temparature difference between the wall and the free-stream fluid, the surface area, and the convective heat transfer coefficient.
MJC
Not so, you could link the nanotubes to larger "arterial" and "venal" tubes to move the heat off-chip totally, then have another heat exchanger where the primary cooling circuit is cooled by a second water circuit, which because there's more room off-chip could be a flow of tap water in and water passed to a drain on the way out. This should prove pretty effective.
"'I pass the test,' she said. 'I will diminish, and go into the West, and remain Galadriel.'"
- JRR Tolkien.
And since we're being pedantic, everyone else in this thread has neglected that the energy to raise 1 cc of water 1 deg C varies based on the water's initial temperature, and is only 1 calorie at 15 deg C, or 4 deg C, or the average between 0 and 100 deg C. It's not that big a difference, though.
Litigious bastards