Slashdot Mirror
Silent Pump for Water-Cooled PCs
Wycliffe writes "New Scientist has an article about a silent pump for water-cooled PCs.
The system, developed by a Californian start-up company, aims to silently solve the problem that the faster chips get, the hotter they become."
Californian start-up company, aims to silently solve the problem that the faster chips get, the hotter they become
I know what you mean but... they aren't solving the problem. They are developing a way around the problem. Solving the problem would be to break the laws of thermodynamics and develop a chip that gets cooler as a function of time.
While they say it pushes 200mL/min, they don't say how much power it requires to do so... peltiers are incredibly effective, but suck an obscene amount of power to do so.
If this new pump requires 75W or more, then you're unlikely to win in the long run - you'll just need a bigger PSU (and bigger, noisier fan in it) to get the job done.
Anyone have any more detailed links?
Namaste
nt
"Even if all the technical details are ironed out, I think it will be five years at least before fans are replaced. They are still the cheapest option."
/. crowd. It's the geek "cool"ness factor (groan).
However, that's not the point with the
However, others are cautious about the idea. "I don't like mixing water and electricity," says Paul Lee, at QuietPC in North Yorkshire, England, a company that specialises in PC noise elimination. "Even if all the technical details are ironed out, I think it will be five years at least before fans are replaced. They are still the cheapest option."
I really got a kick out of this statement. This is an example of how insecure people are and how most people view change as a bad thing. If Paul embraced the technology, perhaps his company would win the OEM contract with Intel. Imagine the money! But not with that thinking.
I think it's clearly about time that computers move into the liquid cooled stage. Look at what it did for automobiles. Anyone here own an old air cooled porche is a big city?
Using a stiring engine for heat recovery would likely be counter productive as your cold sink is too small to help drive the engine, this is the cooling problem you are trying to solve in the first place with the chip not cooling enough.
The drawback is to get the efficiency at that small of a scale, the cost will be fairly high. The engines can be made small enough, but for a price probably far too high for the general market. Say about $2,500 each. Then you have to shield the rest of the computer from the NeFeB linear motor/oscillator magnets used to drive the engine in reverse. Another thing is the drawback that thay are probably shock sensitive. So a good whack could cost a fortune to the average pocket book.
Once you comprimise the efficiency to the point of about say 8%, you might as well just give up on the stirling idea and go with a Peltier device since they are solid state and fairly cheap (20$~100$) and already off the shelf.
The high efficiency capabilities of the stirling engine are about the only reason you would want to use one in the first place, so the option is really not available it would seem for a cost effective and mass produced market.
However, others are cautious about the idea. "I don't like mixing water and electricity," says Paul Lee, at QuietPC in North Yorkshire, England, a company that specialises in PC noise elimination.
Translation:
However, others would like the idea to go away. "Let me think up a reason to discourage this," says Paul Lee, at QuietPC in North Yorkshire, England, a company whose business model depends on PCs being noisy.
While a noble sentiment, I just don't see a market for it. Can you imagine AMD trying to pitch thier new, low-power 3GHz CPU against the newest Intel 5.5GHz (190W heat dissipation)?
"But that extra 2.5GHz doesn't really make a difference. Our chip will be saving you money on your cooling bill!"
Sorry, but MHz/GHz sell CPUs. That's not going to change anytime soon.
I am very tired of idiots who post about being worried about water near electricity, rust, you name it. I think it shows how basic science education is being neglected. So I would like to make a few points.
Pure water is a poor conductor of electricity. That's why ordinary condensation caused by the lowering temperature of humid air rarely causes problems with electrical equipment. The conductivity of good quality DI or distilled water is actually not high enough to affect most digital systems. It wouldn't be good if it got into rotating components, but in the low-resistance 5-12VDC environment, a little damp is not a problem.
Furthermore, in the absence of dissolved air, water does not promote rust. Generations of science teachers have shown kids that iron nails in distilled water stay bright while those in aerated tap water rapidly rust.
The biggest problem I am aware of with water cooled electronics - and I have been involved on and off with liquid handling since 1980 - is inappropriate choice of materials. The common polyamide (nylon) pipe is water absorbent, as are some other widely used polymers. Cast metals are also often prone to porosity, pinholes and slag inclusions, which can be major sources of leaks. Pressure testing is a good idea.
Another problem with water cooling for electronics is inappropriate design of connections, resulting in too much mechanical load either on the connection itself, sealing faces of pumps, or the attachment of the cooling plate to the substrate.
Unlike automotive cooling, vibration is not usually a major problem.
This isn't a howto essay, but here are a few pointers.
DO NOT EVER use automotive components. They are designed for robustness and can handle high levels of sludge, and in any case are designed for use with glycol mixtures.
DO NOT create high pressures. The object should be to have wide flow paths working at low pressure differentials with minimum turbulence. I'm amazed when I read descriptions of heat removers describing them has having internal passages designed to promote turbulence - because turbulence is bad. You want, ideally, laminar flow across the hotspot so that it is in contact with a constantly changing flow of cold water.
Platinum cured silicone tubing is very good. It contains no cure residues that could make the water acid and it is very flexible. It needs to be carefully routed to prevent kinking but it puts low stresss on joints.
Flow back to header tanks should avoid bubbling to prevent aeration, and header tanks should be covered except for a minute hole (filtered) for pressure compensation. If you can keep spores and bugs out of the water, you will not grow algae.
Ideally all metal parts should be the same metal or at least metals of similar electrochemical potential.
Use lab grade DI water.
Use compression joints rather than just relying on the elasticity of the pipe.
NEC has said they are working on a system designed to eliminate leaks - their curious reference to "resin" being, I suspect, a reference to epoxy or nylon components that are prone to leak slowly. The automotive industry has done it: liquid cooled auto engines now require hardly any maintenance of the cooling system. I'm sure that once the serious manufacturers get on the case and the amateurs start to fade into the background, liquid cooling for personal computers should come on rapidly, for all the same reason as automotives (more power in smaller space, more accurate temperature control, able to reach less accessible places, smaller block mass needed for heat exchangers). The technologies are all there, the need is obvious.
Panurge has posted for the last time. Thanks for the positive moderations.
The system, developed by Californian start-up company Cooligy, aims to silently solve the problem that the faster chips get, the hotter they become.
Water cooling only improves the heat dissipation by some constant factor, keeping the chips from melting down until they get yet faster and smaller. Next year, I guess we'll be hearing about liquid sodium cooling systems.
But note that exotic cooling systems don't actually make the the heat go away, they just move it to a different part of the room. This does nothing to solve the closet-full-of-computers problem. Your home is still going to get hotter than you want in summer. In fact, cooling systems themselves generate heat so any cooling that isn't actually needed to keep the processor from melting down actually makes the heat problem worse.
It's always nice when noise goes away though.
Back to the heat question, this kind of thing is great for pushing forward the frontier of single-cpu systems. But single cpu systems aren't where the action is, it's in SMP and clusters. For that reason, I'm a lot more interested in systems that generate fewer calories per mip in the first place.
Have you got your LWN subscription yet?
Something I've wondered....
Would increasing he size of the chip die help with heat dissipation? The die itself is about what, 1/2" square (approximately, and using an AMD Athlon as a reference)..
If we were to increase the die itself to the size of the `skirt' around it (that which holds all the pins)... perhaps a vented or flow-through design on the die also. I don't forsee that this would cause any major repercussions as far as motherboard size increase.
My other thought...
Aren't processors fast enough as it is? Maybe we'd do ourselves some real good by focusing on the other bottlenecks in the system and getting them up to snuff.
Besides.... we all know (whether we'll admit or not) that even the most modest computers in use spend most of thier time simply waiting around for the user to make them do something.
Or maybe focusing on efficiency- for reduced power consumption (and ultimately heat generation).
Or maybe, scrapping the whole thing and starting over (as has been talked about frequently in the x86 world). At least it's been said by people who understand hardware better than myself that the current state of x86 is a total mess.
I dunno. Maybe i'm not geek-worthy, or something.
do() || do_not();
Don't be a fool and get caught up in propoganda. I'm only posting this because I'm so sick of people spreading this crap.
.18 and .13 microns. Of course it uses 1/4 as much power as a classic Winchip, so would anything with that much shrinkage.
The Eden chip runs extremely cool because it doesn't DO ANYTHING. It is a brute-force, simple yet inefficient processor design.
It has no branch prediction, no out-of-order-execution, no register renaming, and a half-speed FPU. These are the exact same specs the Winchip had when Via bought it, they have simply shrunk the die to
This means you have to wait forever in CPU time to get anything done, which means you get real-world performance in the PII 300 range. Sure, each cycle wastes a tiny bit of power, but when you take 3x as long to do something, you use 3x as much real-time power.
This is what I am referring to when I say it's an inefficnent design. Sure, it's low-power, but you have to compare it to OTHER architectures to get a feel for how good it really is.
Consider that an Intel Pentium III Tualatin LV clocked at 733MHz would have only %50 higher max thermal power than the Eden chip, and you start to get the point. You could clock the Tualatain at 500MHz and match the Eden's max thermal power, and have significantly higher performance.
Consider that an 800MHz Pentium M would have ONLY 2 WATTS higher max thermal power, and it smacks you upside the forehead. Here we have a chip with roughly 2.5x the efficiency of the Eden ( processing power to power consumption ratio ), thanks to the fact that it has been DESIGNED FROM THE START to be efficient.
The Eden is only "low power" because it is inefficient, and it didn't sell well when it was sold as a normal desktop processor. The whole small form-factor ITX is the only thing the platform has going for it, and as soon as small systems with low-clocked Pentium Ms come into play, VIA's market will evaporate.
You could make the most inefficient core in the world run extremely cool, even say a Pentium IV, so long as you throttled the clock speed and shrunk the process. This is all VIA has done with the Winchip core used for Eden.
Incidentally, VIA finally released a new Winchip core, the Nemiah, with 6th-generation features like OOOE, branch-prediction, full-speed FPU.
GUESS WHAT? It performs better, but the power consumption is up too. Sadly, even Nemiah wasn't designed as efficiently as CPUs already out there.
EFFICIENT != LOW POWER.
Man is the animal that laughs.
And occasionally whores for Karma.
Last time I checked neither the Intel nor AMD plans included organic processors in 5 years. And they do plan 5 years ahead. But I'm just a human. Let me ask HAL here... Oh yeah. We haven't yet been able to even create a computer with the intelligence of a 2 year old child or a grey parrot or a chimpanzee. I forgot.
Well, anyway, I'm late for my inexpensive trip around the rings of Saturn. Oh yeah, even though we landed on the moon over 30 years ago, no human has traveled beyond it. Were you alive when the first human stepped on the moon? Wake me up in a couple of months when we have holodecks, intersteller space travel, replicators, transporters, time machines, and organic computers. I'm sure it's all just around the corner. And there's certainly no question that we'll all live to see it because the problem of immortality will surely be solved before 2005. Until then I think I'll just take a short nap.
Quite an experience to live in fear, isn't it? That's what it is to be a slave.