PC Watercooling Prototype Is Pumpless

Watercooling is usually more efficient than air cooling for the same volume of equipment, and — important for many people — it is generally quieter. However, you still have water pump noises to deal with. [Der8auer] got a Wieland prototype cooler that doesn’t use a pump. Instead, it relies on the thermosiphon effect. In simple terms, the heat moves water — possibly boiling it — upwards to a radiator. Once the water is cool, it falls down back to the heat exchanger again.

It looks like any other AIO, but the block is extremely flat compared to normal coolers, which have the pump on top of the plate. As you might expect, orientation matters, and you can’t have tight bends in the hoses. The system also has to be totally airtight to function properly. The test was meant to be against a commercial AIO unit with the same number of fans. However, there was a problem, and the final test was done with a larger radiator with one of its three fans removed.

The prototype performed fine and was quiet. It didn’t do as well as the commercial cooler, but it wasn’t bad. Of course, this is a prototype. Maybe a final product will do better. Around the ten-minute mark, the IR camera came out, and it didn’t show any major unexpected hot spots.

We’ve seen water-cooled printer hotends, and pumping is a problem there. We wondered if this technology might work there. The whole thing reminded us of heat pipes without the internal wick to move cold working fluid. We’ve even seen a water-cooled calculator.

23 thoughts on “PC Watercooling Prototype Is Pumpless

  1. I saw couple of these in the early 2000’s.

    Only difference is that they used the thermosiphon on the radiator side, and the radiator was vertical on the side of the case, because there you have a much longer distance for the hot water to rise – or sink – I don’t remember which way it went.

    1. Yeah. My first reaction to this was that the radiator should mount vertically for this to work. Also, when this system was implemented in old houses the piping had a much larger diameter than in newer systems with a circulating pump. Maybe try mounting the radiator vertically and use 15-20mm diameter tubing + a radiator that supports that.

      This system will also respond very slowly to cooling demands. I think it has a better chance of working with a constant heat supply.

      1. The radiator should be vertically higher than the CPU to get the best effect, and the return pipe from the radiator should be from its coolest end.

        You also don’t need to be airtight, a header tank would be fine too (and it would allow outgassing), just like in my central heating system at home.

      2. Respond? I would guess the response is zero, or perhaps “instant but limited” is more fitting.

        But with regard to siphoon effect, the placement of the radiator should I think have a limited effect – what you want is a pressure difference, and this is provided by two pillars of water. The height and temperature difference of these pillars will dictate the pressure difference, so if your radiator is part of the pillar it will temperature difference between the sides smaller.

        Putting the cpu in the bottom of the car should make a difference though!

      3. Steam heat is not the same as (convection driven) hot water radiators.

        A steam system has one pipe. Steam goes up, condensed water drains at the bottom of the same pipe.
        Convection has a return pipe, and usually a circulation pump.

      1. No, gravity is your only pump.

        You run a short vertical tube from the top of the cold plate to the top of the radiator. The water heats, loses density(thermally expands) and rises easily to enter the top of the radiator. Then, while it’s in the radiator, it loses heat, gains density, and falls through the radiator. You want a radiator with little internal resistance, as a thermosiphon is partially a momentum-driven design, and the falling water doesn’t have a lot of momentum.

        It’s the opposite of the thermosiphon water heaters you see on wikipedia et al, because those are designed to absorb heat and raise the working-fluid temperature. There’s more potential energy available in a heating system than a cooling system, so the cooling system has to be designed much more carefully to minimize momentum loss.

        It’s really easy to make a prototype out of some garden hose, a couple sections of copper pipe, and a kitchen stove. Put a coil of copper in a pot on the stove, to simulate the heat picked up from the cpu. Run a fairly large diameter piece of copper pipe vertically, with the top somewhat higher than the heat pickup coil and the bottom much lower than the bottom of the heat pickup coil.

        Connect the two with some appropriate rubber hose. The large diameter copper pipe and the bottom/cold water return need to be strictly low flow resistance; the overall loop needs to have much less drag than a pumped exchanger.

        Play with the heat, watch the flow rates. You’ll notice that if designed and laid out right, its pretty functional, but the cpu temperature won’t be as cold. It works better with more water in the loop, because that gives you a bigger energy store, and more mass, thus more consistent flow once it gets some momentum.

        It’s probably not really practical for a computer, but there are many situations where being able to passively maintain a piece of equipment’s temperature is interesting. There was an electrical converter that just had to be installed in the hot roof area in a metal shed, for unfixable reasons. But it was a single-slope roof, and the backside was to the north, with good shade and solid breezes. Some one inch pipes on 2×4 offsets on the back of the shed, a two-gallon tank at the top, a couple of linear feet of copper tube spot-welded to the case of the converter, and some hose. Worked wonders. Not only did the converter itself suddenly stay under its 75 degree celsius design rating, the shed itself was noticeably cooler.

        at one point, it was decided that “a proper radiator” was needed, rather than just some large, gently curved pipe, but that didn’t work too well. Most radiators are designed for positive-pressure pumping, and even very large (junkyard) car radiators have enough drag to interfere with building any flow momentum… at the same time, they’re not that much more effective at transferring heat with such a limited thermal differential.

        The old cars that used thermosiphon effect to cool the engine without a pump used radiators with much less flow resistance than modern automotive radiators, and the water ran hotter than in a modern design. It lingered longer in the engine, moved slower into the radiator, moved slower through the radiator. This meant that it had more time to pick up heat, and more time to lose heat, so the temperature change in the water was potentially greater… but the temperature of the engine itself ended up higher, even under ideal conditions.

        Worse, the potentially higher temperature and the generally lower pressure tended to make it more prone to boiling, which didn’t just risk losing water, it also disrupted the flow’s momentum, creating a positive-heat feedback loop.

        A thermosiphon with insufficient water stops siphoning, and thus stops flowing. At that point, it becomes a lossy phase-transfer cooler… water boils, the steam leaves, taking heat with it, and stuff sort of works until you run out of water, at which point the temperatures spike IMMEDIATELY, and stuff breaks.

  2. I wonder if a small peltier might make this kind of setup work better. If your looking for the phase change of a medium as its main means of heat rejection. Than it would certainly fill that void, and could be tuned enough to prevent condensation issues what would normally be a problem.

  3. I saw a custom passive water cooled PC years ago that was basically a conical pot of water that fit directly on top of the CPU, with heat sink fins on the inside and outside. Basic convection – water gets heated by the CPU and rises, gets cooled by the fins and sinks.

    1. A thermosiphon is this, just with potentially more surface area, and, potentially, a much larger vertical drop to help encourage momentum-driven density-change circulation…. if sized right; you want surface area for transfer, but not surface area for drag/turbulence, it’s tricky.

    1. I would have guessed air cooling.

      The very first car engines had interesting features, like a pool of gasoline in a pan over the engine head to evaporate fuel for the intake, to serve as the carburetor. Hot engine makes gasoline boil, a simple butterfly valve mixes it with air.

  4. The comments above in regards to a vertical passive radiator to enhance the thermosiphon action, use of different liquids with phase change characteristics, and Wieland’s extensive expertise in cooling suggest this project is more about working within seriously stifling design constraints, than it is about getting it done. What would the computer case look like if optimized for passive cooling?

    1. With no constraints at all some direct heat path to the walls of of the airtight sealed enclosure the computer sits in at the bottom of a deep body of water…
      Plausibly placed anyway probably a very large vapour chamber/heat pipe thing feeding a really giant radiator with wide spaced fins for passive convection – in effect the entire case becomes the heat sink, and it wouldn’t be some tiny ITX sized case if you wanted to dissipate high end components… Or a similarly giant bucket of water with a crinkly edge to have lots of heat exchange area to the air still.

  5. A friend used to live in an old house that he squatted. Talking 20 years ago. It was a house that had to be destroyed but the owner did nothing with it for a while. When he lived there, the only heating he had was a wood stove that he put there himself. He made his own thermosiphon heating system so he could heat the upper floor of his house. He called it his roman heating system. He said he found the idea in an old book talking about roman plumbing and used that idea, using pipes that went into the stove. He was able to get a little bit of extra heat output for the uppper floor to make it a bit more comfortable.

    1. Every liquid cooling system benefits from greater thermal mass (larger water tank). For a thermosiphon, this isn’t just a good idea, it’s pretty much mandatory. You need a signficant thermal gradient, but you need the coldplate to not be subject to extreme temperature variations while waiting for momentum circulation to build up.

      In practice, if reasonably sized and positioned, you’d be looking at a 5-10 minute delay between significant changes in heat input and significant changes in flow rate, thus cooling.

      My experience wasn’t with a cpu, but it was a similar thermal budget (watts/BTUs, pick your poison) as a modern cpu+gpu. It would have worked fine as a computer cooling solution, if you fabricated cold plates with MUCH less backpressure. Basically no pressure (beyond a few vertical column-feet of water), and not much velocity, so even a 1 inch pipe cross section is almost too much flow drag. But enough vertical feet in the cooling part of the path will partially offset some flow-drag sins…

      For that project, the cooling system was in a windy, deeply shaded location, and was almost always under 40 degrees C. The coldplate temperature never exceeded 70 degrees C, even when dumping a kilowatt of heat (for several days at a time). But, it used almost fifteen meters of 1 inch copper pipe to manage that, and had a reservoir capacity of 15 liters (with generally at least 11 liters in it). That’s in addition to the fluid capacity of the loop itself.

      Thermosiphons are like sterlings. They can do great things using surprisingly little energy input, but they’re not at all compact, compared to an equally performing design.

  6. From an old Encyclopedia of the Automobile book by the Reader’s Digest there were radiators without fans or pumps, if memory serves. They relied on the radiator being higher than the engine so as to facilitate the flow of water. I think this works better if somehow the water has enough pipe to run through and dissipate heat and maybe it also can be fanless as well.

    Gut feeling says you’d need more real estate than a mini-ATX case, tho.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.