Fail Of The Week: How Not To Watercool A PC

To those who choose to overclock their PCs, it’s often a “no expense spared” deal. Fancy heat sinks, complicated liquid cooling setups, and cool clear cases to show off all the expensive guts are all part of the charm. But not everyone’s pockets are deep enough for off-the-shelf parts, so experimentation with cheaper, alternatives, like using an automotive fuel pump to move the cooling liquid, seems like a good idea. In practice — not so much.

The first thing we thought of when we saw the title of [BoltzBrain]’s video was a long-ago warning from a mechanic to never run out of gas in a fuel-injected car. It turns out that the gasoline acts as a coolant and lubricant for the electric pump, and running the tank dry with the power still applied to the pump quickly burns it out. So while [BoltzBrain] expected to see corrosion on the brushes from his use of water as a working fluid, we expected to see seized bearings as the root cause failure. Looks like we were wrong: at about the 6:30 mark, you can see clear signs of corrosion on the copper wires connecting to the brushes. It almost looks like the Dremel tool cut the wire, but that green copper oxide is the giveaway. We suspect the bearings aren’t in great shape, either, but that’s probably secondary to the wires corroding.

Whatever the root cause, it’s an interesting tour inside a common part, and the level of engineering needed to build a brushed motor that runs bathed in a highly flammable fluid is pretty impressive. We liked the axial arrangement of the brushes and commutator especially. We wonder if fuel pumps could still serve as a PC cooler — perhaps changing to a dielectric fluid would do the trick.

50 thoughts on “Fail Of The Week: How Not To Watercool A PC

        1. ATF. very light oil without being too volatile. flammable, but you have to do it On Purpose and i’m pretty sure even the hottest PC isn’t going to reach flashpoint

          1. Diesel is less viscous and not too flammable or volatile.
            You can also find automotive electric water pumps that work well with ethylene-glycol based coolant, you’ll find them in some cars equipped with heaters (Webasto-style) or some early turbo-charged cars.

    1. Glycol or mineral oil is a better idea.

      But the fuel pump is still not a good idea, because the brushes and bearings are optimized for a MTBF of around 5,000-6,000 hours – which is the typical running hours of a car engine. It’s around 200,000 miles at an average speed of 37 mph.

      That means the pump would die anyways within a year of use in a PC that is more or less constantly on. There are 12 Volt squirrel cage pumps that are magnetically suspended and only have a single ceramic bearing in the middle around a stainless steel axle, and those too wear loose in about a year or two of use. While they continue to work, they will gradually start to make noise.

      Plus, dumping 60 Watts of heat from the pump to the cooling loop kinda defeats the point of liquid cooling. The pump makes about as much heat load as the CPU itself.

        1. i always thought water cooling as an unnecessary middleman. besides id rather put all that money id spend on fittings, blocks and radiators into better components. stock gpu coolers are pretty robust, but it is worth it to buy good high quality fans and a large cpu cooler with a lot of heat pipes and fin area.

          1. The point of water cooling is to distribute the heat over a larger radiator that can deal with it without excessive amounts of air flow. A regular heat sink is very limited in area, and adding more powerful fans just gets you diminishing returns and more noise.

            A liquid cooling loop is at a disadvantage from the start because it has more thermal interfaces to deal with, but it compensates with the ability to give much greater radiator surface area and better airflow over the surface thanks to more optimized radiator designs. In any case, they might give worse bottom line temperatures but offer better cooling capacity at high power.

            That said, most commercial kits do not achieve this, and are simply expensive placebo.

    1. There are brushless fuel pumps, but automotive suppliers like to absolutely sodomize their customer base. Anything above absolute minimum is expected to command stupid levels of markup.
      Swap $10 brushed motor for a $20 brushless one + $10 controller and $100 pump magically turns into $500 one.

      I mean everyone knows running the pump dry will fry it. Thermistors are almost free in the grand scheme of things, yet you dont see them integrated, cents saved + guaranteed dealer income down the line.

      1. Running the pump dry isn’t just about overheating it, but also the bearings run out of tolerance and the pump gets loose enough that it can’t sustain pressure. Once you’ve ran your tank empty and damaged the pump, next thing you’ll notice is the engine starts to stutter in acceleration because the fuel pressure won’t stay up.

        It makes no difference whether the pump is brushless – it’s just not designed to run dry.

        1. The “chip” is an ASIC that takes the hall sensor input and turns it into a two- or three-phase output sequence depending on the input pulse frequency. It is a controller, just not a programmable one.

    2. Yes, however, these pumps are cheap and they’re designed with a probability to fail after a certain time to increase the maintenance cost of older cars and force a replacement.

  1. Brush type motor in a pump with brushes in the fluid? That sounds bad. 7k rpm from a brushed motor? That sounds worse!

    If you water cool a PC or 3D printer extruder, it isn’t a matter of if, but when it’s going to leak. Spilled oil all over my desk and floor isn’t something I want to have to clean up, ever. I suppose you could set the whole thing in a big plastic tub to catch the spills that will eventually happen, but that would spoil the looks of the system which seems to be 90% of the reason for liquid cooling in the first place.

  2. It would have been reverse electrolysis of the exposed copper I reckon. Petrol would not facilitate electrolysis? Why does not an empty tank explode with this style of motor?

    1. It would, if you have a lot of ethanol in, which retains water, which facilitates corrosion.

      >Why does not an empty tank explode

      Because the concentration of fuel vapors to air is way too rich for ignition. It can’t sustain a flame.

        1. It is possible if your fuel tank runs empty and you have air venting from the outside, but in a car with only the single fuel tank, by the time the fuel bump is gulping air, the engine will stop and not provide any more power to the fuel pump, so this type of accident would be exceedingly rare.

  3. Haven’t seen other comparisons, so since the brushes looked ok (but other people’s MTBF/other comments apply)
    If the wire did burn through, I think there are a few things at play here.
    1. Operating a 12V pump at 5V. Should end up drawing more current
    2. Gasoline Viscosity: 0.509 cP Water Viscosity (at 20 degrees C and 30 degrees C) 1.0016 cP and 0.79722 cP
    So less than half the voltage and up to 2x the viscosity. Sounds like a recipe for high current draw!
    Note – Mineral oil at 40 degrees C (just a standard one) is 15.3 cP. Not a good idea….

    1. >Operating a 12V pump at 5V. Should end up drawing more current

      No. It will just run slower. Running a brushed DC motor at a lower voltage reduces the maximum current draw.

  4. Wait, y’all expected the bearings to seize? Why, because the water was a bit warmer than the gasoline it’s used to pumping? Well there’s still the boiling point keeping it down at the very least. That’s definitely not bearing-seizing temperatures. What kind of metal goes out of tolerance at 100 Celsius?

    1. I can’t believe that thing has brushes just soaked in the flammable fluid it’s pumping, though. That’s outrageously cheap. And speaking of cheap, paying $75 for this dog-ass part seems like robbery to me.

      1. There’s nothing to ignite if the pump is fully immersed in the fuel. Not enough oxygen to sustain combustion. it is not a bad design, just not the best possible one.

    2. The bearings can seize because the liquid you’re trying to pump is too thick and won’t form a proper film between the bearing at the high RPM the pump is running. It means there bearing surfaces will heat up excessively in trying to force the liquid into the small gap, the water will boil away, and leave the metals touching and grinding each other.

    1. That and pond pumps used to be the go-to when water cooling was strictly home brew… classic setup would be home milled (Or gouged out somehow) waterblock, pond pump and car heater core for radiator.

  5. I did water cooling on a Pentium 133… or maybe it was amd 600mhz slot…. and graphics card… and power supply… all diy cooling blocks with layers of aluminum using an aquarium pump for water circulation. The coolant tank was aquarium and 20 gallons of water warmed up pretty hot. it was great until the pump popped apart while i was out, the computer no longer turned on…

    Water cooling stories are great. way better than the pre packaged kits!!

  6. I have a Laing D5-Vario pump circulating hot water through a hydronic heating system. Brushless DC, 12V (it’ll take from 8 – 24 VDC), magnetic impeller on a ceramic bearing, variable flow. This place used to sell them, they claim it’s discontinued but you can get them from engineering, plumbing, or pumping supplies:

    This place claims to have 25 available – it’s supposedly an Aussie website but the description is German.

    Mine has a duty cycle about 50% – higher in Winter, lower in Summer. I replaced it after about 5 years – it wasn’t worn out but I suspect something in the control circuitry eventually got cooked. Running at close to max operating temperature probably had something to do with that. IIRC it’s rated up to 90 degrees C.

  7. Old I know, but still.

    What I’ve never been able to understand is the lengths people would go to mount all of these cooling blocks and then plumb them. And if it leaks… Ekkk.

    I would take a page from the early Crays, old military radar systems and some of the crazy hams(moon bounce comes to mind).
    There are plenty of liquids that are not conductive, mineral oil comes to mind.
    Make or buy a coolant tight container that is large enough to easily contain your device and will allow proper fittings.
    I would use a transmission cooler(as it’s made for a more viscous liquid than your average radiator), mount fans, add an appropriate pump and hoses. I would use rigid tubing or aircraft style hoses and the proper connectors. You don’t want to chance having one of the damn hoses come off(been there, done that), so hose clamps are out.

    Pop the PC into it’s new home, run power in and USB/Video out, and start filling it up…
    Note, the pump does not have to be very big at all. Just make sure that you leave the heat sinks mounted on the CPU(s). You can leave the fans too, but they won’t be doing much. :)
    Oh, only a SSD should be used here, unless you mount it OUTSIDE the container.

    I made something like this back when I used to build and tune amplifiers for moon bounce and meteor scatter. Lots of large WW resistors and about 10 gallons of mineral oil. I had a buddy of mine who was an airframe mechanic do all the fittings for me after I tried to use simple automotive hoses and clamps and made a BIG mess. It was a legal limit(1500 watts) dummy load. Worked for years.

    former N5VMF

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.