Asetek LCLC Takes Liquid Cooling Mainstream

bigwophh writes "Liquid cooling a PC has traditionally been considered an extreme solution, pursued by enthusiasts trying to squeeze every last bit of performance from their systems. In recent years, however, liquid cooling has moved toward the mainstream, as evidenced by the number of manufacturers producing entry-level, all-in-one kits. These kits are usually easy to install and operate, but at the expense of performance. Asetek's aptly named LCLC (Low Cost Liquid Cooling) may resemble other liquid cooling setups, but it offers a number of features that set it apart. For one, the LCLC is a totally sealed system that comes pre-assembled. Secondly, plastic tubing and a non-toxic, non-flammable liquid are used to overcome evaporation issues, eliminating the need to refill the system. And to further simplify the LCLC, its pump and water block are integrated into a single unit. Considering its relative simplicity, silence, and low cost, the Asetek LCLC performs quite well, besting traditional air coolers by a large margin in some tests."

Comments

Liquid Cooling already mainstream

[Anonymous Coward]

Heck, I'm typing this on an out-of-the-box ~4 year old liquid-cooled Power Mac G5....

Re:Liquid Cooling already mainstream

[MightyYar]

Let me know when Asustek sells as many kits as Apple sells computers.

Re:Liquid Cooling already mainstream

[DAldredge]

Asustek makes some computers for apple.

Re:Liquid Cooling already mainstream

[MightyYar]

Asustek makes some computers for apple. Indeed they do, but I made a typo. The manufacturer of this kit is Asetek.

Re:Liquid Cooling already mainstream

[Anonymous Coward]

Asetek != Asus

Asetek makes vapor phase change coolers, Asus makes motherboards and graphics cards. Neither Asus nor Apple makes commercial phase cooling or liquid cooling gear.

You managed to troll the wrong industry entirely!

Re:Liquid Cooling already mainstream

[MightyYar]

I mis-typed. Sue me.

Also, I wasn't the original troll. I was responding to a guy that seemed to think that a niche hardware maker was more mainstream than Apple.

Re:Liquid Cooling already mainstream

[bill_mcgonigle]

What's strange is the TFS extols the virtues of the new part that sounds just like the AC Delco part that Apple used.

Re:Liquid Cooling already mainstream

[infosix]

If the "quality test" in the number of units sold then you should change your Mac for a Pc !

Re:Liquid Cooling already mainstream

[MightyYar]

If I were to define "mainstream" in terms of quality, then Windows would be a niche product :)

For those of you who don't like stop & go traf

[bobdotorg]

A one page link:

http://www.hothardware.com/printarticle.aspx?articleid=1128

Liquid cooling for datacentres?

[mrogers]

I'm surprised liquid cooling is still seen as a fringe/hobbyist technique, with water (or oil) having a much higher heat capacity than air I would have thought liquid cooling would make sense for datacentres - instead of huge electricity bills for A/C you could just plumb each rack into the building's water system (via a heat exchanger of course, I don't really want to drink anything that's passed through a server rack). Does anyone know if this has been tried, and if so why it didn't work?

Re:Liquid cooling for datacentres?

[jfim]

As far as I know, that's what project Blackbox uses for cooling. Note the blurb where it specifies the water connectivity requirements.

Re:Liquid cooling for datacentres?

[notgm]

do you really want plumbers called in when your site is down?

Re:Liquid cooling for datacentres?

[ZeroExistenZ]

i would have thought liquid cooling would make sense for datacentres - instead of huge electricity bills for A/C you could just plumb each rack into the building's water system

There are a few things that come to mind:

• - A datacenter might have different clients renting a cage, owning their own servers you can't enforce the use of watercooling. AC will have to be present and running in any case.
• - Water + electricity is a risk. With tight SLA's, you don't want to fry your server with your extra investments in its redundant failover hardware altogether.
• - Available server hardware isn't typically watercooled. Who's going to convince the client hacking a watercooled system on your most critical hardware is a good decision? For defects, a support contract with the hardware vendor is typical. If you mod it, soak it, you're out of warranty and can't fall back on your external SLA.
• - electricity "bills" aren't an issue, you have so much amps you can run on each cage if you rent you keep under it or you'll have to rent another cage (notice an advantage for the datacenter here?) It's always part of the calculated cost, it's a non-issue really for datacenters or for you when you want to rent a part of the datacenter.

Re:Liquid cooling for datacentres?

[greyhueofdoubt]

Because air has some undeniable advantages over water:

-Free (both source and disposal)
-Non-conductive
-Non-corrosive
-Lightweight
-Will not undergo phase change under typical or emergency server conditions (think water>steam)
-Cooling air does not need to be kept separate from breathing air, unlike water, which must be kept completely separate from potable water

Imagine the worst-case scenario concerning a coolant failure WRT water vs air:
-Water: flood server room/short-circuit moboard or power backplane/cooling block must be replaced (labor)
-Air: Cause processor to scale down clock speed

I don't think water/oil cooling is ready for mainstream data farm applications quite yet. I also think that future processors will use technology that isn't nearly as hot and wasteful as what we use now, making water cooling a moot point.

-b

Re:Liquid cooling for datacentres?

[ibeleo]

IBM recently released the p575 which states "Cooling requirements Chilled inlet water supply/return required for all systems". They also have a kit to turn the rear door of a rack cabinet into a Heat Exchanger. So there is a move in that direction

p575 spec http://www-03.ibm.com/systems/power/hardware/575/specs.html
Rear door exchanger http://www-132.ibm.com/webapp/wcs/stores/servlet/ProductDisplay?catalogId=-840&storeId=1&langId=-1&dualCurrId=73&categoryId=4611686018425028106&productId=4611686018425023461

Re:Liquid cooling for datacentres?

[eagl]

The ONLY THING water cooling does is (potentially) provide a larger surface area to disperse the heat. So totally wrong/ignorant... Is this a troll? Water cooling does a lot more than that.

1. Can be a LOT quieter than normal air cooling.
2. Allows for heat removal with a much smaller heat exchange unit on the heat source.
3. Allows for heat transfer to a location less affected be the excess heat being dumped (such as outside a case) instead of just dumping the heat in the immediate vicinity of either the item being cooled or near other components affected by heat.

There are other reasons, but these alone are more than enough. Did you not know these, or were you just trolling?

Re:Liquid cooling for datacentres?

[MightyYar]

It's also cheaper to pump a small amount of water than a huge volume of air.

Re:Liquid cooling for datacentres?

[diablovision]

The Black Box is a complete watercooled data center in a shipping container.

Re:Liquid cooling for datacentres?

[phantomcircuit]

You could run the hot water for the building through a heat exchanger before you heat it up with a boiler. Cold Water -> Heat Exchanger -> Warm Water -> Boiler -> Hot Water Over all the energy used to go from Cold Water -> Warm Water is saved.

Re:Liquid cooling for datacentres?

[KillerBob]

-Non-corrosive

Air is one of the most corrosive substances there is. Specifically, the oxygen in the air is. It just takes time. Normally, a server won't be in operation long enough for this kind of corrosion to happen, especially if it uses gold-plated contacts, but it will happen.

Air is less corrosive. But depending on the liquid that's in use in a liquid cooling rig, it usually isn't corrosive or dangerous to a computer anyway. Liquid cooling rigs are usually an oil such as mineral oil or an alcohol like propanol, neither of which is particularly harmful to electronics.

Also... while it's a technicality, air *is* conductive. It just has a very high impedance. It *will* conduct electricity, and I'm pretty near certain you've seen it happen: it's called lightening.

Finally... if your server is running hot enough that mineral oil is boiling off, you've got more serious things to worry about than that. (its boiling point varies, based on the grade, between 260-330'C -- http://www.jtbaker.com/msds/englishhtml/M7700.htm )

Re:Liquid cooling for datacentres?

[Artuir]

I'm sure there's some engineering to be done to solve that problem for servers. You could make copper piping through the entire solution (this IS a server, no expense is spared, no need for flimsy tubing) that would be good for 70+ years. You don't keep servers operating 70 years typically. Well, at least *I* don't. Some of you guys might just to brag about how much uptime your Linux box has to your great grandchildren.

Linux bastards.

Re:Liquid cooling for datacentres?

[RiotingPacifist]

Given that they use AC because they cant be bothered to organise a proper air cooling system (pumping the hot air out of the back of server, instead of cooling all the air in the room, etc), its simply because its cheaper to use AC than actually organise anything.

You do have a good point though, use of a non conductive oil, that was cooled against water pipes, would mean the servers are just as safe as they are at the moment.

Re:Liquid cooling for datacentres?

[Artuir]

Yes, because they are consumer grade solutions, not enterprise grade. You can engineer a solution that uses a 100% solid copper piping/waterblock framework with absolutely no joints or gaps anywhere in the actual box. Design the thing so the weak points, or any point where water could possibly leak is on the outside of the system (obviously not above or below, might need a custom server cabinet for this) and you can have cooling solutions that will not harm server hardware due to leaking.

I'm not saying my idea is the best (or even decent) as obviously it makes components impossible to upgrade.. but it's one possible solution of many more, I'm sure.

Re:Liquid cooling for datacentres?

[bitflip]

You mean like this:

http://www.apc.com/products/infrastruxure/index.cfm

Re:Liquid cooling for datacentres?

[evilviper]

1. Is a result of the larger heat exchange area. And makes no difference in a data center.
2. No benefit for any practical application. Definitely makes no difference in a data center.
3. Does not affect the cooling costs of a data center in the slightest.

Nothing about water cooling will reduce the cooling and energy costs of a data center IN THE SLIGHTEST. You're doing a lot of magical thinking, with NO experience in the subject.

Re:Liquid cooling for datacentres?

[evilviper]

The same is equally true for an A/C unit, as it is for a liquid system.

Re:Liquid cooling for datacentres?

[Have Brain Will Rent]

As any scuba diver could have told him, water conducts heat far more efficiently than air. IIRC it is a factor of about 25.

Re:Liquid cooling for datacentres?

[Have Brain Will Rent]

Not only that but if you attach a Maxwell's Demon to the output you can get cooled water separated out from the hot and then feed that back in to the cooler while sending the separated hot water to the boiler!!!

Re:Liquid cooling for datacentres?

[Anonymous Coward]

Not sure what you mean, isn't that what sys admins are anyway?

Re:Liquid cooling for datacentres?

[Murphy Murph]

You could run the hot water for the building through a heat exchanger before you heat it up with a boiler. Cold Water -> Heat Exchanger -> Warm Water -> Boiler -> Hot Water Over all the energy used to go from Cold Water -> Warm Water is saved.

Unless your datacenter is collocating with a (large) laundromat, there just isn't that much demand for hot water at a datacenter. No laundry, no showers, little to no cooking.

Someone check my numbers.
Tap water = 7 degrees C.
Water heater hot water = 50 C?
CPU = 65 watt thermal output = 55889 calories / hour.
56,000 calories can raise one liter of water 56 degrees C in an hour, or raise 1.3 liters of 7 degree water to 50 degree water in an hour.
So every CPU needs 1.3 liters of water an hour, minimum (assuming we can super-concentrate all the heat and not just slightly heat larger volumes of water.)
The Google "datacenter in a trailer" rumors were saying 5000 CPUs per trailer.
That's a 6500 liters of water per hour output.
LARGE laundromat.

Re:Liquid cooling for datacentres?

[eagl]

My experience is with datacenters that are apparently not as generic as the ones you seem to be claiming experience with. Just because you do things one way and maybe always will, does not mean that another customer will have requirements that your cookie-cutter approach can satisfy.

Put a datacenter 300 ft underground, and see how far simple air cooling gets you. In that case, there MUST be a way to dump the heat that doesn't involve simply blowing air around. If it works for you, that's fine. But attempting to imply that whatever datacenter you are responsible for cooling or designing would fit every customer's needs is pretty narrow minded. Maybe you really believe what you're posting, but your level of experience outside a cookie-cutter place to stack computing power seems to be somewhat lacking.

Re:Liquid cooling for datacentres?

[pavera]

I don't know where you are hosting where "electricity bills" don't matter.

I have systems hosted in 3 different DCs, 3 different companies. All of them raised their rates in the last year by 20-30% in one way or another. One DC includes the electricity in your flat monthly bill, the only incremental charge in that DC is bandwidth (IE you get 100GB of transfer, if you go over its some dollars per GB), they raised their flat rate 20%, citing higher electricity costs.

The other 2 DCs provide metered electricity to the cage, some amount is included in the cage rental, overages are billed incrementally. These 2 data centers have both increased their incremental charges by 100% in the last year, and increased their cage rental rates by 10-15% citing increased electricity costs. Now you can say "they're just increasing their margins" but I live within 25 miles of 2 of the facilities, I know my electric costs at my home have more than doubled in the last year, up almost 250% in the last 5, so no they aren't just marking things up unnecessarily, its all the same electric co.

All in all, this means an additional $5-600/mo in cost for our hosting. from $2000/mo to $2500-2600/mo depending on electricity and bandwidth usage (and a hint, we've only gone over on our bandwidth 1 time for a total charge of $12). I can only imagine if we were grown out (we plan in 3-5 years to have multiple racks in these three DCs and have budgeted in our plan ~75k/mo for hosting costs (based on the prices from a year ago). Well, a 20-30% increase in that turns into real money like 15-25k/mo increase. Being able to save that money would mean being able to hire 3-5 full time engineers at 60k/yr each. I'd much rather have the engineers than give that money to the electric company.

Re:Liquid cooling for datacentres?

[evanbd]

Also... while it's a technicality, air *is* conductive. It just has a very high impedance. It *will* conduct electricity, and I'm pretty near certain you've seen it happen: it's called lightening.

If you want to get all technical about it, you're basically wrong. The resistivity of air is exceedingly high. However, like all insulators, it has a breakdown strength, and at electric field strengths beyond that, the conduction mode changes. It's not simply a very high value resistor -- nonconducting air and conducting air are two very different states, which is the reason lightning happens. The air doesn't conduct, allowing the charge to build higher and higher, until the field is strong enough that breakdown begins.

For materials with resistivity as high as air in its normal state, it's not reasonable to call them conducting except under the most extreme conditions. Typical resistance values for air paths found in computers would be on the order of petaohms. While there is some sense in which a petaohm resistor conducts, the cases where that is relevant are so vanishingly rare that it is far more productive to the discussion to simply say it doesn't conduct.

This is one of those cases. Claiming that air is conductive is detrimental to the discussion at best.

Re:Liquid cooling for datacentres?

[kisielk]

It's being done though not on the system level but on the rack level. SGI's ICE platform has water-chilled doors: http://www.sgi.com/products/servers/altix/ice/features.html

This is a great bonus for high density HPC applications. Typically in a datacenter you are blowing air up from the raised floor in front of the servers. However, a good deal of it is taken up by the servers in the lower part of the rack, leaving the top servers running warmer than the lower servers. Supposedly the water chilled doors help a lot in this scenario.

Re:Liquid cooling for datacentres?

[algae]

I believe that Sun's Blackbox uses water cooling for refrigeration between racks, not as a method of cooling the server hardware directly. Like the sibling poster says, too much risk of leakage near the electrical bits. With however many gallons/sec Blackbox requires though, you can turn a lot of hot air back into cold air and just move it around in a circle.

Re:Liquid cooling for datacentres?

[adolf]

Ok, you've got a brain, so you should also know this:

Given a water cooled rig and an air cooled rig which operate at the same efficiency (in terms of Watts dissipated per Watt of cooling power), water cooling and air cooling perform just about identically as long as things remain inside of the case.

Move the water cooling system's radiator outside of the case, and things start to slant toward water cooling.

Observations:

1. They're equal in cooling capacity, but the air-cooled system is simpler and has fewer single points of failure (and is not wet).
2. Nobody ever seems to put the radiator outside of the box in mass-produced PC-based systems.

Conclusion:

Water cooling is a sham. Bigger/more efficient heatsinks, cooled by air, would bring a more substantial improvement in the quest for quiet, reliable, or rugged computers, than does the current crop of water cooling arrangements.

Re:Liquid cooling for datacentres?

[cheier]

Liquid cooling can affect the energy costs in a big way depending on how well integrated the system is. As an example, CoolIT systems had developed a server rack with an integrated liquid cooling system that they had shown off at CES this year. The rack essentially used hydraulic fittings to allow you to hot-swap systems from the chassis, while still keeping the cooling centralized.

They had essentially used the radiator from a Honda Accord, which they found to be able to dissipate between 25 and 35 KW of heat. With a system like this centralizing the area where heat is dumped, fluids can be piped out to a radiator sitting outside, so essentially, a large portion of the heat produced from a rack of computers, can be relocated outside of the data center.

Even without moving the heat outside, you can still save on cooling costs. Because you have the capacity to dissipate so much heat, less AC costs are required simply because you can used a forced air system to move the gobs of hot air out and outside air in. This could potentially save up to 30% in cooling costs alone, let alone if you were to just relocate the exchanger to the exterior of the building.

Re:Liquid cooling for datacentres?

[evilviper]

Put a datacenter 300 ft underground, and see how far simple air cooling gets you.

Do you have ANY experience with ANY air conditioning? You don't see to have the foggiest idea how they fundamentally work.

An A/C will work superbly underground, with less work on the building itself (ie. never mind the per-machine hook-ups).

Re:Liquid cooling for datacentres?

[evilviper]

Liquid cooling can affect the energy costs in a big way

No, it really can't.

With a system like this centralizing the area where heat is dumped, fluids can be piped out to a radiator sitting outside, so essentially, a large portion of the heat produced from a rack of computers, can be relocated outside of the data center.

You could similarly open up a data center, with just large fans blowing ambient air in and out.

With either method, it just wouldn't work. A $50,000 server rack is not your home PC. It's not okay to let it operate on 40C+ ambient temperatures. It's not specifically out of spec, but it's a very, very high temperature to start off with, which means you don't have a very large margin for error, and you have to be careful of the extra heat output by the server when loaded.

I already also mentioned that, with such a method, you need a MUCH larger radiator, and much higher throughput forced convection, since you're trying to cool it with warm air, rather than chilled datacenter air... The temperature difference is just too low. It quickly gets to the point that your non-optimal cooling solution is almost as expensive as just running an A/C, and allowing less airflow, smaller radiators, etc. Heat-pumps are continually increasing in efficiency, after all.

Re:Liquid cooling for datacentres?

[Cancel-Or-Allow]

If you live in a desert climate, air cooling sucks and if the dust is not dealt with at regular intervals things fail quickly. First dust starts to accumulate on the fan blades (unevenly) putting it out of balance thus placing greater strain on its barrings. Meanwhile, intel's ingenious design of their retail cooling fan and heatsink ends up being clogged with dust. The ambient temperature inside the chassis begins to increase as the chassis fan and PSU fans have now ceased, leaving only the higher power cpu fan spinning faster in vain packing the dust more tightly in an increasingly overheating heatsink. With no airflow coming into the chassis, the first thing that usually goes is the HDD. And is usually the first time a service call is generated from a user.

In my server closets I buy my own chassis with an easy to remove air filter on the front of each chassis, about every month I take it to the sink rinse it out, shake off the excess water and put it back. The closets have their own independent AC and room monitor with alerting sophisticated enough if for some reason the AC quit Servers would begin treating it like a power failure and begin shutdowns.

But for the poor PCs sitting on the floor of 100s of users all working as digital air filters. Unless regular maintenance is done (the dust storm blow job with the air compressor) they crash within 6 months to 1 year.

I really like the idea of liquid cooling. Just leaving a rig that I'm setting up for some running 24/7 with 2 8800GTX, QX9770, 2 Seagate 1TB, 3, WD 150 Raptor, evga 680i mobo, Tagen 1100W PSU 2 19" Dell 2007WFP in portiat mode sandwiching their 30" model in the middle.
Compared to last year, my electricity bill increased by 40% for this month My Kill-A-Watt reader registered an average 600W, but the heat would activate the main HVAC in my house much more often. All in all to run this rig costs close to $100 per month in energy costs.
If I could easily get that heat outside of the home, then I would only be expending 600W for the computer and saving 3KW each time the HVAC kicks on.

Another thing, the Cooler Master CMStacker is a POS. At least in the desert, it is nothing more than a dust storm in a box. So far I am using a great quality filtered rackmount chassis, but he wants CM with pretty lights lights. I'll show him where to the canned air is in costco and how to make small dust storms on his front porch every month.

Sorry for not having much of a point, my ambien as begun to kick in and will be typing in my sleep soon.

Re:Liquid cooling for datacentres?

[nospam007]

It did work. IBM used it in the past. Check Google for "water-cooled ibm"

Re:Liquid cooling for datacentres?

[eagl]

You keep going on about experience, when it's obvious you're the one with the limited experience since you have not seen any actual applications that require any more thought than "stick some more fans on it and it'll be ok" or "well, just put another AC exchanger on the roof and it'll probably work fine". You still have to get the heat out. And that's the whole point about water cooling. Getting the heat out without relying on whooshing air around (whether it's air blowing on heatsinks or air in an climate control system).

But again, you've obviously never worked on any projects that can't be solved by sticking more fans on it and you don't have the imagination to think about any applications that might not be solvable by whooshing more air around, so we may as well end the discussion.

Re:Liquid cooling for datacentres?

[jack8609]

As someone who makes their living figuring out how to move heat from A to B (in avionics, not datacenters), this comment makes my head hurt for a number of reasons... First off, as others have pointed out, liquid cooling in data centers is a reality and folks like IBM have worked on liquid cooilng for decades. Due to many of the reasons already mentioned, everyone avoids liquid cooling as long as they can and a number of technologies have helped on this. For example, the transition from Bipolar to CMOS around the time I finished grad school put a lot of thermal engineers out of work for a while. However, liquid cooling is used in plenty of places - Cray has done it for a long time on their supercomputers (not on the latest one - at least not for local cooling), F-22 & F-35 have liquid cooling for their avionics (for weight reduction), nuclear reactors (using liquid metal), etc. Every thermal conference held in the last 5 years seems to have had at least one session on data center cooling and most of the work is on implementing some aspect of liqud cooling. The electricity required for data center cooling is now on the order of 30-40% of the total power (don't quote me on that - I'm actually thinking it is higher than that, but as I said I don't work in that market). Air cooling is great for simplicity, but it has limits that we are fast approaching. The simple methods for air cooling involve just dumping hot air into the room and once you do that and are using the A/C on twice as much warm air as the the hot air that you would otherwise be cooling without the mixing, your cooling power requirements shoot up considerably. Liquid cooling has two potential benefits (as well as the numerous challenges already described). As people have pointed out, the thermal conductivity of water is much higher than air (k = 0.6 vs. 0.027 W/m^2K). This is important because it higher local cooling (the convection coefficient is ~20x higher). More importantly, the volumetric heat capacity of water is massive compared to air. The temperature rise (in C)of cooling air is ~(8*W dissipated)/(lb/hr of cooling air). So when you are dissipating MW of power and trying to keep the electronics no more than 20-40C above ambient, you need (literally) tons of air. Water is about 4x better in specific heat and about 1000x more dense, so you reduce the volumetric flow rates by ~4000x and get more effective heat transfer at both the electronics and heat exchanger ends (so they can be smaller). Pumps can generally be much more efficient than fans and the amount of pumping power for a 100% device scales with the density * volumetric flow rate^(1.5) [assuming that I am doing the math in my head correctly...]. If either the price of electricity or the heat dissipation levels in data centers continue to go up (fairly safe bets...), you will see increasing use of liquid cooling in that application. Keeping things leak free, and other related maintenance issues, and dealing with legacy architecures seem to be the biggest hurdles.

Re:Liquid cooling for datacentres?

[BLKMGK]

This is actually pretty amusing as when I setup my home office I designed much the same thing! Sadly I didn't put it into place but indeed it would have worked quite well I'm sure. Radiator in the crawlspace, temp sensed electric cooling fan mounted on radiator - Ford Taurus fan most likely. Copper or PVC piping up through the floor into the office in a loop with a shutoff valve in the middle to regulate bypass flow. An agro pump to move the liquid or perhaps a small pool pump. Fittings on the pipe mounted to the wall to allow lossless connection and disconnection of cooling fittings for computers. Water blocks on computers are cheap enough and no pump would be needed nor radiator inside the case. Concerns were mostly surrounding filling it, purging it, and maintaining the radiator in the crawlspace which is tight. Google a bit and you'll find a guy who DID build a system like I had intended, only his "radiator" is his swimming pool - seriously!

In the end the SO veto'd my system and the work\mess would've been a hassle but it was QUITE doable. I've built systems smaller than this and was doing it 10 years ago with Peltiers to boot in order to overclock at near freezing temps. The primary advantage I found to this was not just that it cooled so well but that it allowed me to move the heat exchange *away* from my computer. When my home office was WAY small temps climbed enough that the air exiting would trigger my thermostat - in Winter the house was cold, Summer it was freezing - except in the office.

When I water cooled my hottest computer the radiator was placed outside my window with a small fan - temps IN my office dropped dramatically as did the dB level and the CPU ran COLDER than ever. Ambient temp outside is often low enough compared to a blazing hot CPU that the temp drop is awesome. 90F outside is no biggie when your CPU inside is pushing 50+C - as my quad is next to me right now. Lots of industrial tech exists to chill water far cooler than ambient temp outside too and some datacenters likely already use it for their A/C today.

These days I water cool only a little having lost hardware to water leaks over the years - twice to be exact. I do have a system on my desk waiting to be resuscitated that ran for 2 years nearly trouble free - yes normal tubing allows evap BTW but this one was retired because it's older not from coolant loss. I am a firm believer in using water, I will use it again, but it IS more trouble and each new damned video card requires a different block than the last at notable cost. The temp drops you achieve are amazing - especially on today's hot video cards. New systems built like the one in this story and the new Swiftech stuff make water easier for laypeople who don't want maintenance.

Anyway, what I have learned with my own systems I believe can be used in a datacenter to advantage. The primary thing I've learned being that the heat exchange can be efficiently made "elsewhere" and not someplace where exchange from one machine effects another. I'd also imagine it might be nice to be able to walk into a datacenter and not sweat hearing protection as much!

Re:Liquid cooling for datacentres?

[BLKMGK]

The biggest issue with running a datacenter on 40C ambient air with big fans blowing it in and out the doors is that air cooling is so inefficient that the cooled components would overheat as they pick up so little temp drop from AIR. 40C WATER cooling on the other hand would bring those CPU and HDD temps down a good bit.

You're failing to understand just how much better water transfers heat vs air.

Re:Liquid cooling for datacentres?

[BLKMGK]

You haven't actually ever built and run a water cooled rig have you?

Re:Liquid cooling for datacentres?

[Paul server guy]

Y'all are basically idiots.

I just came from NASA Ames research center, (Talk about heavy supercomputing!) and they are heavily water-cooled. Right now they have coolers on each of the processor blocks, and radiators on the backs of the cabinets, but are quickly moving to directly chilling the water.
They use quality hoses and fittings, no leakage.
The efficiency is so much higher than air, and it makes the operating environment much nicer. (They have people in there regularly swapping out drives tapes, whatever.)

Of COURSE water cooling is what you want to use for any high-performance computing. It's purely a matter of efficiency. (And you can use the hot water elsewhere.)

Re:Liquid cooling for datacentres?

[evilviper]

It's almost amusing to see you continually backpedaling, after every factually incorrect statement...

Good luck with your subterranean server farm.

Re:Liquid cooling for datacentres?

[sarcells]

Liquid cooling hit the mainstream, mainframes in fact, 'way back in the early 1980s with the IBM 3090 mainframe. What we now call a water block, IBM called the Thermal Conduction Module. According to an article in Scientific American at the time, it combined a water block and chip packaging. The metal of the TCM directly contacted the chip substrates.

Does anyone know about any water-cooled supercomputers?

Re:Liquid cooling for datacentres?

[adolf]

No, I have not.

Would the basic rules of thermodynamics change if I had?

Re:Liquid cooling for datacentres?

[kitgerrits]

I actually have a rack of watercooled equipment sitting in a datacenter.
air-cooling was not an option because the air-cooling system was maxed-out for that floor, whilst there was plenty of floorspace left.
(blame it on the silly cooling requirements of bladeservers)

Re:Liquid cooling for datacentres?

[BLKMGK]

What would change is that you would actually have a better understanding as to what you were talking about vs simply making broad assumptions with ZERO experience. You start by assuming that air and water systems work at the same efficiency and that mounting the radiator inside the case somehow puts them there. This is not true, had you actually worked with any of this you would know how well it works even with a radiator in the case.

But hey, all of the various big guys moving to water to cool their large systems must be stupid and you know better since you grasp thermodynamics better than they do. And better than all of the folks who are documenting serious temp drops and higher overclocks using water cooling. They must all be going to this trouble for no real gain right?

Don't bother to post back with more assumptions on the thermodynamics involved, experience trumps theory.

Re:Liquid cooling for datacentres?

[eth1]

I too have often wondered why more datacenters don't use water. (aside from the fact that you'd have to build the place from the ground up for it, probably)

I think a standardized interface could help immensely:
- each rack-mount server has a cold water input on one side, and a hot water output on the other.
- the rack has a cold water rail up one side and a hot return down the other.
- under the floor, each rack is plugged into hot and cold "bus" pipes, which feed into one of those industrial waterfall coolers like you see on big a/c systems

Leaks would be fun, though!

Re:Liquid cooling for datacentres?

[phantomcircuit]

Put a datacenter in near city center and you could sell the warm water to steam producers.

At least in the US most major city centers use steam to heat their buildings.

Re:Liquid cooling for datacentres?

[phantomcircuit]

That's a good point. Using the heat from a data center to heat the air is a good idea.

Re:Liquid cooling for datacentres?

[adolf]

My statements are true. If you do not understand them, then please ask for clarification. If you'd like to refute them, feel free to use your anecdotal evidence to do so.

But all you're doing is waving your hands around and talking about "various big boys," as if mere the notion of it growing in popularity obviously means that it is better, while insinuating that the total concept of dissipating waste heat into air is impossible to grasp without actually experiencing a water cooling rig first.

But rather than converse or debate, I do suspect that you're just mad because you think I pissed in your oatmeal. Whatever dude - enjoy your hobby. It's not personal.

Re:For those of you who don't like stop & go t

[Clay Pigeon -TPF-VS-]

Too bad they didnt compare it to a good air cooling solution like the thermalright ifx-14 or ultra-120.

think big - the heat is still there in the room

[gelfling]

A DC might have 20,000 servers. That heat has to go SOMEWHERE. If it's pumped into the ambient air just like an aircooled machine, you're still going to need large AC units to move that hot air out of the DC

Re:think big - the heat is still there in the room

[gelfling]

Yes I know what those are. But carrying the heat away from the servers and venting it to the room isn't going to help the overall need to cool everything. It may make each server slightly cooler but it's not going to alleviate the need for power or cooling in any data center overall.

Re:think big - the heat is still there in the room

[Dare nMc]

considering most CPU's can run at 90C but most air cooled systems will recommend something below 23C ambient. Efficiency of a cooling system maintaining 60C would be much greater, which with a much higher conductivity should be fine hence the cost savings of liquid cooling.

Also from the article, it doesn't work

[gelfling]

Even the article tries hard to tout its benefits but their own stats show its not worth it. Either it's a crappy implementation or its simply not relevant.

Re:Also from the article, it doesn't work

[eagl]

Even the article tries hard to tout its benefits but their own stats show its not worth it. Either it's a crappy implementation or its simply not relevant.

How so? They show that it's quieter and more effective than stock cooling, and significantly quieter than an aftermarket air cooling solution. What exactly are you looking for then? You gotta be more specific than just a completely unsupported criticism that doesn't even reflect the test results, let alone explain your personal criteria.

Here, try something like this next time:

It looks like a good/bad item because the performance was/was not what I'd expect from a water cooling system costing [insert price here]. You can get similar/better performance from [insert alternative product here] for less. Tradeoffs with the alternative are it's quieter/cheaper/louder/expensive but based on my own critera of [insert your own priorities here], I think this product is great/teh suck.

Give it a shot, you might like it.

Re:Also from the article, it doesn't work

[gelfling]

You don't have to be an entirely patronizing asshole. But I'm guessing you don't work in sales.

Ok so it's marginally quieter. As for its absolute cooling power it's on par with whatever air cooled unit you can get today with a lot less complexity. All in all that's pretty weak justification. If that's your definition of "it works well" then you clearly care about noise above all other criteria. There are probably better ways to make your PC quieter than this.

Re:Also from the article, it doesn't work

[eagl]

You don't have to be an entirely patronizing asshole. It's the internet, so actually I do (heh).

But you are still arguing from a position lacking in factual information. Water cooling can be almost completely silent, and can remain so even when cooling hardware that would otherwise require very loud fans for conventional air cooling.

This does not even address the additional cooling requirements seen in overclocking, small form factor, or otherwise special use equipment. A water cooled HTPC for example typically has to trade off performance for noise, as high performance video cards and cpus get far too hot for silent air cooling even if you don't use a small enclosure that looks like it belongs next to a TV or stereo. Water cooling, especially simple ones like the review subject, would let you use a top of the line cpu and fast video card in an HTPC without having to worry about how to get the heat out of the case in a quiet manner. That means the HTPC could be used not only for home theater applications, but also for PC gaming on your widescreen plasma, LCD, or projector.

And that's just one application. There are many others, including high density rackmount or server applications that cannot sound like a jet engine or must be able to dump the heat outside of the environment containing the computers. And as for costs, and complexity, a properly designed system should not be any more trouble than the air cooling requirement of replacing fans and cleaning dust out of the filters (or system if filters are not used).

I don't water cool because I don't need to. I don't run computer hardware that generates a heat load requiring loud air cooling, and I have no problem with the heat from the cpu getting mixed with the air inside my case before it gets sucked out by the exhaust fan. I also do not have the spare time or inclination to "get into watercooling" and a half-assed approach is a good way to dump a gallon of coolant into your case and onto the floor. But an easy to use system like the review system, at a nice price, is an option I will consider for the next time I build a system from scratch because it appears to be no more difficult to install than a conventional heatsink, is quieter, and cools at least as well as a high-end (and noisy) air cooling system. And it's sealed up so it is unlikely just slapping it in and forgetting about it will result in destruction of my computer and carpet.

Re:Also from the article, it doesn't work

[gelfling]

So it's just quieter. Again, if that's your main concern, then fine. There are probably less problematic ways to address that. I for one would not want to ever have to worry about hundreds of little water cooling systems in a data center each with the potential to break down and cause a catastrophic failure among many machines.

Back in the old TCM days you might have a dozen CEM complexes each with a TCM ganged into a single chiller pump system. That's a kind of failure rate that's manageable. But with 5000 server blades in a data center today the probability of something going horribly wrong from 5000 water coolers is unacceptable.

Re:Also from the article, it doesn't work

[eagl]

In a datacenter using blade servers, I'd expect some sort of hybrid heat exchange system would be more useful than pure water cooling. I strongly disagree that the only benefit is lower noise, but also remember that we're not just talking about datacenter applications either. All sorts of applications (such as the htpc setup I described) could get not only lower noise but also allow higher performance due to the better managed thermal load.

And that is all water cooling does - allow a better and more managable way to move heat from one area to another. Saying that the only benefit of managing heat transfer is lower noise is a bit like saying the only benefit of owning a car is because you can choose the color.

"mainstream"

[iroll]

I thought the G5 Power Mac took liquid cooling mainstream in 2004.

I guess this is one of those phrases, like "the Year of Linux on Desktop," that we'll hear ad infinitum.

Re:"mainstream"

[dwater]

Is that only one machine?

If so, I don't see how that could be considered mainstream. Perhaps I misunderstand the term, but to me it means it is used on many different computers, not just one.

Perhaps 'mainstream' is valid in this case because the one model sold a lot? I don't think that fits with my uderstanding of the word, but it's at least debatable, I suppose.

What's new

[theeddie55]

It doesn't seem much different to the gigabyte kit i put in my computer 2 years ago http://www.cluboc.net/reviews/super_cooling/gigabyte/galaxy/index.htm the only difference being the pre built bit which could cause great difficulty if you want to do something sensible like mount the radiator on the outside. (Note: soon after i got mine they released a second version with a different pump and reservoir, and i can tell why, after 13 months, just out of waranty, my reservoir cracked)

Re:To nitpick the summary (no, did not rtfa)

[dwater]

I noticed this too. I wonder what they were trying to say.

...or is it just marketing crap?

Well, maybe.

[jd]

If a server came ready-built with fail-safe plumbing and cooling mechanisms, the answer would be yes. Water, oil, flourinert - these would all be excellent. Total immersive cooling would be more logical than piped cooling, as there are fewer parts that could fail and less possible damage from a failure. You could have a completely sealed compute unit that contained everything and was ready to go, so eliminating any special skills on the part of the data centre or any special requirements in the way of plumbing for that centre. It would, however, require that things like the PCI bus be external to the main unit, or the system would be unmaintainable. That increases fragility. For pre-specced servers that aren't going to require significant upgrades, this would be an obvious solution.

Re:To nitpick the summary (no, did not rtfa)

[ChrisMaple]

The article says that most water cooling systems use silicone tubing, which the author seems to think is not a plastic. I'm not an expert on plastics, but PVC seems like a poor choice to me. It's too likely to degrade over a decade or so and become brittle or fragile.

Mainstream

[Have Blue]

Didn't liquid cooling go mainstream when Apple used it in the last generation of Power Mac G5s?

Necessity is the mother of invention

[damburger]

This is kind of inevitable, and IMHO overdue. Monolithic heat sinks and fans the size of jet engine intakes have been a pain in the arse for top of the range gaming machines for years. Also, I don't know about anyone else, but the air cooling of my computer is a depressingly efficient mechanism for sucking dust and fluff into the computer and keeping it there.

Shuttle PCs have had this for years

[Animats]

Shuttle PCs have had a heat-pipe and heat exchanger liquid cooling system for years. This made possible their little "breadbox" systems.

Re:Shuttle PCs have had this for years

[BLKMGK]

Intel and AMD systems are also using heat pipes just like the Shuttle XPCs and have been for a year or three now. All of the best "air" heatsinks I am aware of use some liquid in them in the same fashion. Shuttle just managed to build it such that the radiator was a little further divorced from the heatsource is all.

Re:Shuttle PCs have had this for years

[I.M.O.G.]

The important difference here is that the heat exchanger in shuttle pcs uses heat pipes which thru the use of a pressurized fluid, utilize phase change to transport heat. Phase change systems include heat pipes and those using compressors. Liquid cooling with pcs includes pumps, tubing/piping, and a liquid - the fluid never changes to a gas. In this way, your point does not apply to the liquid cooling topic.

Ummmmm

[Have Brain Will Rent]

Wouldn't "is a totally sealed system" take care of "evaporation issues, eliminating the need to refill the system" without requiring "plastic tubing and a non-toxic, non-flammable liquid"???? I'm just saying....

Re:Ummmmm

[ncc74656]

If you had RTFA, you would've found that making a sealed system apparently isn't enough by itself. The silicone tubing used in most liquid-cooling rigs apparently is somewhat permeable, so water can seep through it and evaporate. Replacing silicone with vinyl fixes that, at the expense of slightly increased rigidity.

Re:Ummmmm

[kd4zqe]

This is very true. I just recently disassembled my system in favor of a Core2 Duo machine. I built the rig because my 1st gen P4 3.6GHz was a pain to air-cool efficiently. I noticed that about a year after assembling the system that the temps climbed rapidly moments after power up. I found that almost all my fluid had gone from the system.

What I thought was fluid was actually UV dye that had permeated the silicone tubing from the cooling solution. Additionally, when I stripped the system, all the tubing ends had swelled dramatically, presumably by the liquid accessing the non-heat fused cut ends of the tubing.

Also, in a rebut to the statements by the article, my system was a WaterChill system from Asetek, and included a CPU block and VGA block in addition to the pump and 120mm heat exchanger, and I found the cost to be quite reasonable at only about $250US. It was very easy to install, and made a nice evening project. Because I transport my system to LAN parties, I decided to reverse the radiator and route the fittings to the inside of the chassis. This took a little common sense, drilling, and planing, but I was very specific in my wants.

Without the cosmetic changes, this still was a very sensical kit to own. I'd recommend for ANYONE to try to build a water rig at least once. If Asetek is trying to move liquid cooling into a more mainstream arena, more power to them.

Re:To nitpick the summary (no, did not rtfa)

[glitch23]

As for plastic tubing, what else are you going to make it from? Metal?

Rubber?

Heat Pump?

[bill_mcgonigle]

With the caveat that thermodynamics scares and confuses me, if you have a bunch of heat coming out of the servers' water-coolers, couldn't you pipe that into a heat pump and recover some cooling energy or even electricity? I'm familiar with a local facility which operates its air conditioning systems on steam, though I forget the name of the technology at the moment.

Re:Heat Pump?

[flappinbooger]

With the caveat that thermodynamics scares and confuses me, if you have a bunch of heat coming out of the servers' water-coolers, couldn't you pipe that into a heat pump and recover some cooling energy or even electricity?
Yes. Now, THAT would be smart. Eliminate the cost of water heaters, augment winter HVAC bills, etc. Steam power plants use "waste" energy, the heat left over in the water after it runs the main turbines, to preheat the water going into the boiler. There's usually heat left over after THAT, and it is at a good temp for use in the power plant building itself. Any heat sent back out to the environment is wasted, and wasted energy = wasted $$.

Now, if it's enough wasted energy to warrant the cost of capturing it... Depends. That's what they pay engineers for, to figure that out.

This is, at least, what I remember from my thermo 1, thermo 2, heat transfer, and advanced thermo courses in college. MMmmmmm.... Thermo.

Re:Heat Pump?

[bill_mcgonigle]

Now, if it's enough wasted energy to warrant the cost of capturing it... Depends. That's what they pay engineers for, to figure that out.

Which means Google already probably knows the answer. ;) Those guys are ruthlessly (and brilliantly) cheap.

Uh...Johnny-Come-Lately

[FredThompson]

Welcome to 2008, OP. Sealed systems have been on the market for months. You can even find Cooler Master Aquagatte (on the market since 2007) in some of the larger retail stores.

Re:Uh...Johnny-Come-Lately

[Paul server guy]

Hell, I've been using a "Consumer grade" easy to use water cooling system in my desktop for over a year in the form of the Titan Robela (http://www.titan-cd.com/eng/watercase/robela.htm or http://www.inland-products.com/singleproduct.asp?search=accessories&partnum=03011)

I have the black Al faced one for longer PSs. It was extremely easy to set the water cooling up, and has kept my machine cool even with two extra blocks for the SLI cards and a chipset cooler. Yes it's not sealed, but then again, is that really a big deal? If it WAS sealed I couldn't have added my extra blocks, and this went together so simply, I doubt I would really have noticed the difference.

This is almost old news by now...

Re:To nitpick the summary (no, did not rtfa)

[Murphy Murph]

I'm not an expert on plastics, but PVC seems like a poor choice to me. It's too likely to degrade over a decade or so and become brittle or fragile.

Like the PVC drainpipes in modern houses?
Like the insulation on your home's wiring?

CFCs?

[bcmm]

What is this non-toxic, non-flammable liquid, given that it probably isn't allowed to be a CFC?

Re:CFCs?

[drt1817]

MMM makes a line of flournerts that have a low GWP but the problem is that they are low pressure and very expensive. These are used for immersion cooling of large power electronic systems, like on a locomotive. We've played around with R134a and it IMHO works better because it is higher pressure and you can set the junction temperature of the chip by adjusting the system pressure. Like the 3M fluids R134a is a dielectric so will not damage the electronics. Single phase water reaches a point where it becomes impractical to take away more heat, the only way to control it is via flow or run it into a chiller. 2 phase systems go beyond that, breaking covalent bonds offers big thermal rewards.

Patch the symtoms or wait for the cure?

[turing_m]

I can't help but think that this is a stop-gap measure. I used to read up about all the various methods of silencing a computer (with the intention to implement myself) but for consumer-grade applications I'd prefer to wait for a variant of Moore's Law to do its work - the propensity for performance per watt to keep increasing until it nears whatever limits are predicted by information theory.

At that stage there will be an option to cool with no moving parts for typical desktop/laptop applications, and it will be a superior solution in all aspects compared with any combination of cooling and sound minimization.

Another welcome change will be for the idle power consumption to drop, which it certainly can.

Efficiency is inherently hard - it's all about approaching zero loss, and losses seem to almost never present a target suitable for a single silver bullet approach. That means work - killing all the losses on your loss budget one by one, and figuring how to integrate all the changes so it still works. The good thing is that once you come up with a design that lops off as many heads of the efficiency hydra as possible, it's just a matter of mass production.

FLCL

[StCredZero]

They should've found a way to put F on the front of the acronym, so that it could be FLCL-Cooling.

Re:To nitpick the summary (no, did not rtfa)

[BLKMGK]

The clear plastic "bling" tubing is often medical grade and guess what? Over time liquid EVAPS from such systems. How? It actually manages to be absorbed by the tubing and slowly dissipate into the air, which is why this system uses a different kind of tubing and why they highlighted it's lack of evaporation issues. You haven't run a liquid cooled rig have you?

Oh and plain old water is a BAD idea in a liquid cooled PC. For one it tends to oxidize things like copper heatsinks over time and for another you get biological growth that QUICKLY kills cooling performance as the heatsinks are covered in slime. Figure two weeks on plain water before it all goes to hell. Those "bling" dyes you see are generally antibio agents too. Chlorine bleach is also effective but has other side effects, I for one prefer Water Wetter for a multitude of reasons.

Re:Overclocking, here we come!

[BLKMGK]

Swiftech makes a system you might be interested in that's also self contained. The pump sits right on top of the CPU and the heat exchanger fits where your 120mm exhaust fan is normally mounted. I'm not using it and would only consider it if I were cooling my vid card too but a friend is using it and REALLY likes it - trouble free install on his box.

Where to buy?

[Tyrion Moath]

Anybody actually find where you can buy this system? The article only says that they found one and the price for a minimal setup, but not where. I'm upgrading soon and this could be a good addition to some new hardware. Googling for "asetek liquid cooling system" only finds pages of news articles :(

Re:Where to buy?

[bobzchemist]

Unfortunately, Asetek does not sell to the retail market, and has no plans to ever sell to the retail market. This means that even if you do manage to get ahold of one, you will get zero support from the manufacturer if anything goes wrong. Not something I'd risk.

Liquid cooling is only going to take off

[goldcd]

when some standards have been defined and actually used. I'm sure one day we'll have an 'ATX+' power supply. As well as the plethora of wires hanging out the back of it, we'll have some loops of tubing with heat-exchangers on the end. Maybe standard ones for chipset, CPU and a couple of GPU ones. Buy a new graphics card and just snap on the right heatsink. It's never going to take off until the systems are all sealed (My mum is not going to buy a Dell with a bottle of 'UV Reactive' magic solvent). Sealed systems are never going to take off until it comes with the heatexchanger you need already there. And that's not going to happen until there's a standard 'fitting' for heat exhangers on the components. Sensible place to integrate cooling is in the PSU. I therefore foresee a future where your PSU is rated with a wattage and a 'cooling value' and a list of electical and cooling sockets. In the same way your components currently have to rate below the wattage of your PSU, your components will have to rate below the cooling 'power' of your PSU. Once a nice standard has been set for all the above, we can then start having desktop systems that aren't huge empty boxes.

Re:Liquid cooling is only going to take off

[Blastercorps]

While I completely agree with the need for standard fittings on the plumbing, is the high-voltage portion of your computer really the best place for the waterworks portion as well?

Not for the whole datacenter, maybe..

[itomato]

I can see a generation or two of blade-type applications returning to a CRAY-style apparatus:
http://upload.wikimedia.org/wikipedia/commons/e/e4/Cray-1-p1010221.jpg

You might not have an entire DC relying on a common non-air cooling implementation, but doing it for a complete rack-sized unit is feasible.

I'd personally like to see an entire rack siliconed up and flooded with mineral oil.