Slashdot is powered by your submissions, so send in your scoop


Forgot your password?
Hardware Hacking Toys Upgrades Build Hardware

Full Immersion Cooling Comes To Desktop PCs 192

mr_sifter writes "After three years of research and around £100,000 of R&D costs, UK-based Armari has unveiled its XCP prototype. It's a full immersion liquid cooled PC which supports standard ATX components. Unlike conventional liquid cooled PCs, the components are all easy to swap in and out as they're swimming in liquid, rather than under waterblocks. It also looks amazing, pumping around 70KG of electrically inert cooling fluid (salvaged from an old Cray) around its military grade perspex shell."
This discussion has been archived. No new comments can be posted.

Full Immersion Cooling Comes To Desktop PCs

Comments Filter:
  • by pwnies ( 1034518 ) * <> on Wednesday August 27, 2008 @04:32PM (#24770107) Homepage Journal
    No offense, but this just seems like an elaborate waste of money. We've seen immersion pc's before ( [] ). Other than a fancy case and a waterfall, what makes this any different? Why is it worth £100,000 versus a fishbowl PC that'll set you back $200? Give us some decent benchmark results at least; as of now though, I see nothing really original other than a cool case mod here.
    • by lgw ( 121541 ) on Wednesday August 27, 2008 @04:40PM (#24770189) Journal

      Ah, but this is just step 1. In step 2 they add a trained octopus to each tank that will do your PC repairs for you. Then you'll really see the value! Just don't forget to feed your octupus, or it will come looking for food on it's own.

    • by b4upoo ( 166390 ) on Wednesday August 27, 2008 @04:41PM (#24770211)

      It can be a superior mode of building. A waterfall is not what is called for. Rather a radiator like device is sufficient. That puts the cooling fans outside the case for easy maintenance.Dust inside a PC as well as corrosion are warded off completely in such devices. If done right it is a superior build. If done wrong it can make a mess.

    • Re: (Score:3, Informative)

      by Chad Birch ( 1222564 )
      The summary said that the R&D costs were £100,000, not that it was the price of the PC.

      I know, actually read the summary, must be new here, etc.
    • Re: (Score:3, Informative)

      Um, it never says that the case costs £100,000. It says that bringing the case to completion took them that much in R&D. Making something is cheap. Working out how to make it right is expensive.
    • yawn (Score:3, Funny)

      no doubt.

      Wake me up when they put a pc in a high vacuum. You could even put the turbo pump in a different room.
      • by treeves ( 963993 )
        Oh, but radiative heat transfer (which doesn't need a medium) is proportional to temperature to the fourth power. Get that sucker over 1000K and you're all set!
    • Re: (Score:2, Interesting)

      I will add my voice to the flood. Yes, this seems to be a waste of resources. These days we need cpu which consume less resource because vast majority of us are buying laptop PC. I love desktop computers because they can be upgrade but I don't think I would built a machine that require an AC. I wish companies will reseach in building batteries that will last six months, and pc manufacture & software companies will spend their time building technology that will consume less energy. This is where the mar
    • And get much better performance. All you need to cool on a PC for top performance is the CPU and the northbridge, maybe some of the voltage regulators, and the GPU if you want to overclock that.
      • by billcopc ( 196330 ) <> on Wednesday August 27, 2008 @06:16PM (#24771273) Homepage

        $400 ? I'd love to see a link.

        True phase-change cooling usually costs a grand for the kit, then you still have to gut your chassis to fit the ginormous cooling colon^H^Humn. Plus it's noisy as hell. It would require substantial improvements in both areas before ever being considered for general use in PCs.

        This fluorinert jobby is probably whisper quiet, but I don't see anyone racing to order one. In a Cray, the liquid made sense because they were huge machines and it wasn't realistic to even try to cool them with air. Today's computers are reduced to a single board, with a few very localized heat sources.

        Having a big body of liquid will actually hinder the heat dissipation, because the liquid moves far slower than air, and your CPU is putting out 100+ watts of heat in a tiny area, or in my case 350 watts, turning the area near the CPU into a mini deep fryer - definitely not cool!

        Given how today's air coolers can run whisper quiet (at stock speeds and voltages), I just don't see where immersion cooling could possibly fit in the PC market. It doesn't work any better than a high-end air cooler (Ninja or TRUE120), doesn't overclock anywhere near as well as TEC+water setups or phase change, and costs 50 times more.

        • This fluorinert jobby is probably whisper quiet, but I don't see anyone racing to order one.

          A. They're not selling it yet
          B. This is the type of computer you sell to a company so they can put it in their lobby, or to a millionaire as an expensive toy.

        • Whisper quiet? Can high-performance PC's be made utterly dead quiet? I'd think the military might be interested, and they do have a history of supporting proven but expensive and obsolete computers* (as this would likely be in a few years). They're also resilient toward restraints on materials that are non-green. Submarines, stealth equipment of all sorts, there is a place for high performance dead-quiet gear.

          *The aiming computers on US battleships are a good example. They're analog, old, expensive, an

        • by julesh ( 229690 )

          turning the area near the CPU into a mini deep fryer - definitely not cool!

          Hmm. Vegetable oil's a dielectric, you know.

          I'm thinking of an 8-core 5Ghz quad-crossfire 5-litre fryer. Should do the job. You'd just need to make sure you had a way to throttle back your overclocking before you reached the flashpoint.

          It seems like somebody's already on the right track []...

    • What part of "Dream Machine" don't you understand?

      Besides, the hundred thousand pounds is not the retail cost of a product... it is the research and development cost the manufacturer has paid to date.
    • The R&D is 100000 pounds, not the price of the product. That's not at all unreasonable.
    • Plus never mind that for that much money you could have a seriously kick arse PC with the most memory, cpu cores, and video cards one could cram into a computer and have money left over for another 10 years of bleeding edge upgrades.

  • "so the XCP is filled with FLUAHRGHPT." Huh?! What's that again? I can't hear what he is saying. What liquid did they use?
    • by nycguy ( 892403 ) on Wednesday August 27, 2008 @04:39PM (#24770177)
      They used flourinert [].
      • by inviolet ( 797804 ) <> on Wednesday August 27, 2008 @05:37PM (#24770857) Journal

        They used flourinert.

        ...which is why this prototype will never see production. They got their flourinert from an old supercomputer, and that's not a viable supply for fullscale production.

        That makes me wonder about their motivations for this PR stunt. Venture capital, anyone?

        More seriously, I wonder if transformer oil could be used for this sort of thing. Flourinert may be overkill... or maybe transformer oil has enough capacitance to cause problems for the extremely high frequencies used on PC motherboards. Anyone know?

        • by Sj0 ( 472011 )

          Why couldn't you just fill a tank with Ethanol and dunk your PC in there? It's a non-electrolyte, it has a relatively high specific heat (twice that of air, half that of water), it's non-conductive, it absorbs water, what's the problem?

          Besides the fact that it burns pretty well. That's obvious.

          • Re: (Score:2, Insightful)

            by tabrisnet ( 722816 )

            It boils too easily. I don't mean 'rolling boil' I just mean that it will evaporate very easily at room temperature even, let alone at 40degC. Once evaporated into a normal oxygen atmosphere, it is highly flammable (not that it isn't already).

            • by Sj0 ( 472011 )

              Wouldn't that only help things? You'd need to top up your alcohol reservoirs, but the change to vapour would remove heat from the liquid, right?

        • by nasor ( 690345 )
          Transformer oil is the usual choice of liquid for people who do this sort of thing themselves. It works well enough, although some people have reported that over time it can start to slowly dissolve some of the plastic parts on the boards. The main issue with it is that it isn't really a great heat conductor compared to most other liquids. But let's be honest, this isn't really about cooling as much as it's about having a computer swimming in liquid for its own sake.
    • by necro81 ( 917438 )
      Fluorinert []
  • Was I the only one who read it as " Full Immersion Coding Comes To Desktop PCs" ?

    I had a picture in my head of a waterproof system. Perhaps it's a metaphor for coding while drinking a microbrew....

  • by sabatorg ( 1279426 ) on Wednesday August 27, 2008 @04:38PM (#24770169)
    I am happy that I do not work for the geek squad anymore... can you imagine asking grandma to bring in her 300lbs pc?
  • Fluorinert (Score:4, Interesting)

    by dfn_deux ( 535506 ) <> on Wednesday August 27, 2008 @04:43PM (#24770233) Homepage
    Flourinert [] is readily available from 3M in a variety of different compositions. It is the only exotic portion of this type of project and it's cost is the main reason why we don't see more full immersion cooling. I don't know about the rest of Slashdot, but I'd prefer not to spend several hundred dollars per gallon on cooling liquid in exchange for saving myself a little hassle removing cooling blocks from a [more] traditional closed loop contained coolant system. Not a whole lot to be gained from going to full immersion. Also, IIRC, California recently added Flourinert to it's list of potentially cancer causing chemicals, which IMHO makes it less than ideal for a warm LED lit water fall in your living room or office...
    • Re:Fluorinert (Score:5, Informative)

      by dfn_deux ( 535506 ) <> on Wednesday August 27, 2008 @04:48PM (#24770319) Homepage
      Heh, looks like my recollection on the pricing is a bit off: 250ml for 555.35USD [] at one retailer and 3Gal for 2,450USD [] from another.
    • Re: (Score:2, Funny)

      by Anonymous Coward

      California would add grilled cheese to its list of potentially cancer-causing chemicals if it could. Probably already has, in fact.

    • Water has the highest heat capacity of any common liquid. It takes more energy to raise the temp of a given amount of water one degree than for any other substance. High heat capacity is one reason it is so efficient for power generation and cooling.

      For example, one BTU of energy raises one pound of water one degree F. In metric units, the specific heat of water is about 4185 J/kg/K (15C). Whereas the specific heat of Flourinert is about 1049 J/kg/K, or 24% of water. OTOH, Flourinert is about twice as
      • by Sj0 ( 472011 )

        I'm not sure density would be a good trait. The problem is, the denser your fluid, the more difficult it will be for convection currents to carry the heated fluid away.

        I was thinking of ethanol. It's got about half the heat capacity of water, but it's a non-electrolyte and non-conductive.

      • by nasor ( 690345 )
        Simply having the deionized water exposed to air will cause it to become conductive as CO2 dissolves into it and forms carbonic acid.
  • After the Sony rootkit fiasco [] why in the HELL would anyone name a computer product "XCP"???!!! []

    I'm not sure I'd want one. I don't care how quiet it is or how far I can overclock it. If they're dumb enough to screw up with its name, well...

    • Re: (Score:2, Funny)

      After the Sony rootkit fiasco [] why in the HELL would anyone name a computer product "XCP"???!!! []

      Because most people have never heard of the rootkit fiasco (let alone "XCP"). Additionally I don't think most people who have heard of the rootkit would be dumb enough to confuse a bad copy protection system with a liquid cooled PC. Besides, a lot of acronyms double up for different things, there's only so many TLAs available in this world.

  • Unrealistic (Score:5, Insightful)

    by HermMunster ( 972336 ) on Wednesday August 27, 2008 @04:46PM (#24770291)

    Is never happening ever for the average person and thus makes it just a novelty item. Their design is excessive and cumbersome, not to mention has excessive weaknesses such as cost to maintain, cost to purchase, time to maintain, etc.

    It was tough to decipher their speech as well. Word use and pronunciation were a bit distracting. It's tough when your target audience are distracted by your speech instead of focused on your product.

    • Re: (Score:3, Informative)

      by VeNoM0619 ( 1058216 )
      Apparently its approx $500 a gallon [] for the Fluorinert solution.

      Once again, an article that sparks my interest, then someone comes along and destroys it with reality...
    • Re:Unrealistic (Score:4, Insightful)

      by asc99c ( 938635 ) on Wednesday August 27, 2008 @05:48PM (#24770973)

      Oh come on - they're asked to show a dream PC and they've come up with a mini and modern Cray-2 - fits the bill perfectly. It's a concept PC - having some interesting ideas, not making people think yeah that's practical. I don't want a BMW with a flexible rubber 'skin' but I think it's a good concept.

      And the speech is just a English accent - a real one! (many British actors on American TV have to learn the English accent generally used on TV). I have similar difficulty understanding a Texas drawl.

    • by Tim C ( 15259 )

      It's tough when your target audience are distracted by your speech instead of focused on your product.

      That's true, but given that it was an English team showcasing their stuff on an English website, I'd say they targetted their audience appropriately. (Quite apart from the fact that the guy does not have an particular regional accent - if you think that's hard to understand, you should try a strong Glaswegian accent)

      Besides which, you don't hear us Brits whining about all the US accents we have to put up wi

  • What a load of bullshit. There's howto's all around the net on how to do this, and has been for a long time.
    • And besides - Given the cost of living in London (consistently one of the two highest in the world), and the rest of England it is not unreasonable to expect that 100,000 pounds would be only barely enough to employ a team of two-three entry/mid-level engineers for a year. Weird that they bothered to point that out at all.

  • . . . to link to an ad-filled TFA with bandwidth-consuming cheesy music. I hope their server fries.
  • So, you spend $10k on a top of the line rig like this, overclock your CPU to double or even triple normal... which, according to Moore's law means you may have gained 3 years of non-obsoleteness. Don't get me wrong, it's really cool, but is this really more economical (in terms of flops per dollar or some such) than buying a $3k machine, a $3k machine 3 years later, and a $3k machine three years after that?
    • Re:Practical use? (Score:4, Informative)

      by Fulcrum of Evil ( 560260 ) on Wednesday August 27, 2008 @05:17PM (#24770649)
      since when do computers get twice as fast every 18 months? That hasn't been true for a couple years.
      • by Ilgaz ( 86384 )

        They still get twice as fast but Mhz myth is kinda gone. In fact, there are some same speed Xeons performs almost 30% faster than the same Mhz model because of low nanometer, bigger cache etc. The "Quad Xeon" Apple pro one is a good example. 2 versions, same or similar Mhz, the later one beats first generation by 30%.

        BTW as everyone stares to memory bandwidth issues etc. now, people investing to "dead" SGI and rescuing it from chapter 11 are damn clever. Same goes for Cray too.

        • and the 'double speed every 18 months' thing is pretty much gone, so your computer is no longer obsolete in 5 years. Hell, it's fast enough for just about anything, so long as you don't run pointless bloatware, so why not sit tight for a few years.
      • Do you mean the entire computer or just the CPU? For many applications, the hard drive has been the limiting bottleneck for some time now. And of course, liquid cooling will do fuck all to improve response times or throughput.

    • by Panaflex ( 13191 )

      If you really need the FLOPS - you'd do it too. I'm working on a project now that requires 9k of custom hardware acceleration. I'm already spec'ing the next gen product to run on a $100 video card in the next year or two.

      I really doubt I would spec this product... though... it looks like a great addition to some evil genius collection though! Would look nice next to the sharks-n-lasers tank!!

    • There's a false argument that goes if you have a task to complete that will take 100 years on today's computers, wait. Since transistor density doubles every 18-months (allowing more transitors, more cores, more processing power), the computer you buy today will be unable to outperform a computer you buy 18 months from now. I.e. computer A will compute for 18 months. If you buy computer B, it can do in 9 months what the first computer did in 18. Then in can double that work in the next 9 months, complet
  • Not another... (Score:3, Insightful)

    by mpapet ( 761907 ) on Wednesday August 27, 2008 @05:06PM (#24770541) Homepage

    You desktop jockeys have no idea.

    Datacenter workers are far more aware of the demands and complexity of cooling.

    1. It's a commercial pursuit, which is meaningfully different than one-off's from the lab. They must have some customer in mind. If they don't, well, their investors will get burned.

    2. I can easily imagine a commercial application where, perhaps cooling needs overwhelm a building, this may come in as a cheap alternative.

    Get back to us when you've figured out how to cool a rack full of blade servers working near capacity. This may do it more elegantly than air.

    • Yes, that may be so, but it's so fucking *messy* that I'd hate to use it. Also do not underestimate the complexities of a housing a column of fluid that's 7' (2.10 meters) tall...

    • by geekoid ( 135745 )

      Seal the rack and fill it with mineral oil. Done.

  • Midel 7131 (Score:2, Interesting)

    by vimm ( 1300813 )
    rumored to be about 8$ per gallon.... This is just proof that we're in the last few years of VC funding for "amazing, innovative, and revolutionary computer design" instead of something that works.
  • I can see this for extremely dense packed server blades in a rack. Where today our problems are electrical and heat and not compute power. This would solve one of those problems at any rate.

    It's like the good old days of TCM mainframes with massive 400psi chiller pumps.

  • Big deal... (Score:3, Informative)

    by Pedrito ( 94783 ) on Wednesday August 27, 2008 @05:22PM (#24770713)

    This has been done before with fluorinert and mineral oil. In fact, there was a posting here on Slashdot [] back in 2000 where the guys did liquid nitrogen-cooled fluorinert. Definitely more cool-points (pun intended) for that.

    Fluorinert is definitely a better choice over mineral oil if you ever intend on being able to upgrade or fix the PC, since fluorinert evaporates without a residue, but it's a bit pricey.

  • by Mesa MIke ( 1193721 ) on Wednesday August 27, 2008 @05:29PM (#24770781) Homepage

    .. when we have fully baptized and oil-annointed CPU's.

  • Aside from it looking cool when new and polished, this will be an overpriced piece of junk in 3 years. Given the rate, my Wristwatch will have a stronger and faster CPU by then.

    • by uspsguy ( 541171 )

      Of course, the really pricey part - the liquid cooling system can easily be recycled into a new computer.

  • ORAC

    I wonder if it has his voice and attitude too?

  • []

    only $2450 for 3 gallons..... it wouldn't be exotic if the liquid was cheap.

  • I did a proof of concept on this for myself several years ago. I used inexpensive mineral oil. Distilled water is non conductive as well, but most likely would be easier to short out with contaminates.
    Although I never found out the heat dissipation efficiency of the oil, It did work.
  • by kungfoolery ( 1022787 ) <> on Wednesday August 27, 2008 @09:02PM (#24773521)
    That fluorinert-based cooling apparatus comes with a PC.
  • With improved power efficiency in chips lowering the heat generated, and better fans and case designs, we're already at the ideal place of a serious desktop computer being silent and cool-running.

    Recently I've built two computers following Ars Technica's guide for the Hot Rod. There's no noise at all in a quiet room, and when I periodically check on the temperature, it's lower than older computers--typically 30-35 degrees on the CPU, when older computers are in the 40-50 range. That's with four fans, thre

Some people manage by the book, even though they don't know who wrote the book or even what book.