Basement (External) Water Cooling Project

[FJC]

My gaming system is water cooled and runs great - but I had a goal to make it nearly silent. I also wanted to get it to stop dumping tons of heat into my study, especially with summer coming soon. I decided to build a cooling "tower" in my basement (directly below the study), and move the pumps, radiators, and fans there.

It has been quite successful - here's my little build log. I did this primarily to move the heat out of the study and make the computer quieter, not necessarily for better temps, so I don't have any firm benchmarks to share yet (though temps seem as good or better - they are larger radiators after all). I'll work on getting those.

Build log/pictures:
https://imgur.com/a/QTwDXu1

 

[guitarslingerchris]

That's pretty awesome. I just built an ITX watercooled machine and did an external 480 radiator and I'm still contemplating moving the radiator into the room just on the other side of the wall for the same reason so my living room stays silent!

 

[FJC]

Thanks - I'm pretty pleased with how it came out, and amazed at how quiet it is. Every time I walk into the study now, my first thought is, "why isn't my computer running?"
Now I just need to work on the fan controller and cable management on the basement setup.

 

[Zarathustra[H]]

That is pretty damned nice.

If I can successfully route tubing down to my basement, I am considering doing something similar, but with water chillers down there, running the coolant just slightly below ambient, but above the dew point so that it doesn't result in condensation.

 

[Zarathustra[H]]

How did you find that pump dealt with the extra long tubing (and elevation) needed for this kind of loop? I'm considering the same EK dual D5 top for mine.

I'd imagine the added head pressure would make filling slower, but once the whole loop is full and you have gravity pulling both with and against your flow it mostly cancels out. Still a long loop with a larger amount of coolant being pushed around than is typical for PC water cooling, so I'd be curious about what kind of flow you are seeing.

 

[FJC]

How did you find that pump dealt with the extra long tubing (and elevation) needed for this kind of loop? I'm considering the same EK dual D5 top for mine.

These D5's are rated for something like 3.5 or 3.9 meters of head pressure (the specs weren't consistent), so I figured two would handle it. I actually got this set up and hooked up a "test loop" upstairs using a spare res, CPU block, and a spinny flow meter so that I could do a "proof of concept" that the pumps would actually move the coolant. They easily did, so I was confident I could hook it to my actual computer and have it work.

Right now I have the pumps running at about 3000rpm each. I think I can go lower, but I want to make sure that if one pump fails, the other can still move the coolant. These "vario" model pumps have a little dial on them to set the speed (as opposed to PWM), and have an RPM single-wire cable. Currently I'm unable to read the RPMs on the fan controller due to some unknown issue, but if I run a 15' rpm cable (that I made by splicing wire onto a 3-pin fan extension cable), my motherboard shows about 3000rpm when I have the pumps on about the '3' setting (they go 1 to 5 on the little dials).

 

[Zarathustra[H]]

How did you find that pump dealt with the extra long tubing (and elevation) needed for this kind of loop? I'm considering the same EK dual D5 top for mine.

I'd imagine the added head pressure would make filling slower, but once the whole loop is full and you have gravity pulling both with and against your flow it mostly cancels out. Still a long loop with a larger amount of coolant being pushed around than is typical for PC water cooling, so I'd be curious about what kind of flow you are seeing.

Actually never mind, you addressed that.

Also, that problem with the reservoir location is one I thought of and decided to design around in my build.

I may even go with a custom reservoir for my build, because I want large amounts of coolant (to make the rate of change of coolant temp slower so the chillers don't have to constantly switch on and off) I also want to insulate the reservoir and the tubing as I plan on running slightly sub-ambient.

I'm thinking one of these 2 gallon beverage dispensers (glass, so I don't have any leeching) with the spigot replaced with a barb fitting for the flow out, and a hole in the lid for the return tube.

Then I'll put the whole thing in some sort of box (maybe a plastic trash can with a lid?) with a hole drilled for the tubing, and fill around it with spray insulation to keep it insulated.

The chillers will be one Hailea HC-250A (295W cooling) and one Hailea HC-500 (790W cooling)

The thought process is that none of these are big enough to handle a fully overclocked HDT system with big GPU's on their own, so the small one will be set with a target a degree above the big one, and will handle idle and low load situations. When the temp of the coolant goes above what the little one can handle, the big one will kick in.

Anyway, I won't hijack your thread any more than this. I'll start my own instead when I get around to starting this project.

 

[FJC]

Anyway, I won't hijack your thread any more than this. I'll start my own instead when I get around to starting this project.

No worries, I look forward to your thread, that sounds like quite a project!

 

[FJC]

Ok, a good hour of Unigine Heaven is completed. Temps are all stabilized after 30 minutes, but I went an hour to be sure. GPU hovered between 34/35C. CPU touched 58C for a split second, but once then spent most of the benchmark right at 41C on the package. All nice and chill.

 

[NoOther]

Please tell me the cooling tower in the basement isn't just sitting like that on a shelf with nothing protecting or keeping it in place. Almost seems like a disaster waiting to happen. Otherwise, this is a pretty good plan.

 

[FJC]

Please tell me the cooling tower in the basement isn't just sitting like that on a shelf with nothing protecting or keeping it in place. Almost seems like a disaster waiting to happen. Otherwise, this is a pretty good plan.

it's sitting on a very sturdy table in the basement. One of those heavy-duty folding banquet tables.

 

[Zarathustra[H]]

it's sitting on a very sturdy table in the basement. One of those heavy-duty folding banquet tables.

Yeah, it looks sturdy enough.

I don't understand why anyone would think it would just fall. Basents generally don't move on their own. Unless it is precariously balanced or something, it ought to be fine.

Maybe NoOther keeps wild animals trapped in his basement or something

 

[NoOther]

Yeah, it looks sturdy enough.

I don't understand why anyone would think it would just fall. Basents generally don't move on their own. Unless it is precariously balanced or something, it ought to be fine.

Maybe NoOther keeps wild animals trapped in his basement or something

No, but people occassionally go into basements, and people sometimes do stupid things, like trip and fall and hit the shelf it is sitting on so that it then falls over, breaks the line, water spills out, the computer overheats and now you have a mess and a dead computer...

But yeah, essentially wild animals...

 

[VanGoghComplex]

This is awesome. Excellent work! I'm so jealous of anyone with a 3D printer these days.

I was confused at first by the reservoir you still have in your case, but I bet it catches a lot of air for you during the bleed, huh? Plus you can top off your coolant from there instead of having to go downstairs.

Personally, I'd opt for a separate power supply downstairs and just set the fans full tilt. You could run the cooling system 24/7, independent of the PC. Not much need to throttle fans back on low load when you don't have to listen to them, right?

 

[Zarathustra[H]]

This is awesome. Excellent work! I'm so jealous of anyone with a 3D printer these days.

I was confused at first by the reservoir you still have in your case, but I bet it catches a lot of air for you during the bleed, huh? Plus you can top off your coolant from there instead of having to go downstairs.

Personally, I'd opt for a separate power supply downstairs and just set the fans full tilt. You could run the cooling system 24/7, independent of the PC. Not much need to throttle fans back on low load when you don't have to listen to them, right?

The problem with the reservoir downstairs is during fill.

As long as the res is closed it works fine. If you have to open it to top it off, all the coolant that is above the reservoir in the loop (upstairs) will come rushing down.

 

[VanGoghComplex]

The problem with the reservoir downstairs is during fill.

As long as the res is closed it works fine. If you have to open it to top it off, all the coolant that is above the reservoir in the loop (upstairs) will come rushing down.

Hm... With that much static head, I guess I can imagine it pulling enough to collapse the timing upstairs. Or if there were air to there to be "stretched."

Very very cool setup, though. I wish my house had an arrangement that made this a thing I could do. =)

 

[Zarathustra[H]]

Hm... With that much static head, I guess I can imagine it pulling enough to collapse the timing upstairs. Or if there were air to there to be "stretched."

Very very cool setup, though. I wish my house had an arrangement that made this a thing I could do. =)

Yeah, it happens in loops fully enclosed in cases too, but there is less head, and less loop volume so it doesn't have as much of an impact.

In a large loop like this you could have problems filling it, as you top of the reservoir, shut it and turn on the pump, and it goes down low.

Then you open the reservoir to fill it again, and the coolant you started pumping upstairs comes right down at you again. I guess it would be possible to fill it if you are very fast, or find some positive pressure way of pumping the coolant into the reservoir, but it will be much easier to just place the reservoir at the top of the loop.

 

[thesmokingman]

^^Ya fill it with a funnel. Here's a pic of my old system power flushing the old fluid out while filling with fresh fluid.

That is cool but its a lot of work for such a small load.

 

[FJC]

This is awesome. Excellent work! I'm so jealous of anyone with a 3D printer these days.

I was confused at first by the reservoir you still have in your case, but I bet it catches a lot of air for you during the bleed, huh? Plus you can top off your coolant from there instead of having to go downstairs.

Personally, I'd opt for a separate power supply downstairs and just set the fans full tilt. You could run the cooling system 24/7, independent of the PC. Not much need to throttle fans back on low load when you don't have to listen to them, right?

There's no reservoir downstairs in my final setup, just the one upstairs. It catches all the air, and I can bleed/top off from there.

I strongly considered just having a power supply downstairs, running the pumps/fans 24/7. That would work just fine. However, I'd have to be sure to remember to go downstairs and turn those pumps off BEFORE ever disconnecting the loop (for example, if I wanted to work on my PC). And conversely, I'd have to remember to turn the pumps back on prior to turning my PC back on. I tried to figure out how I could have a positive indicator whether the pumps were running or not from upstairs. So, I bought one of those little spinny flow indicators, and planned to incorporate that into the loop in my computer. I could glance at that to tell if the coolant is flowing or not before I turn on the computer. However, while doing my proof-of-concept tests I found that the darn little spinny flow indicator is noisy - it must be unbalanced, making a pulsing sound. As the goal was to make this system as quiet as possible, there was no way this was going in the loop. So I ditched it and went with the current plan, that the pumps are powered from the computer PSU. It's simple and it works, though I've been warned that my long molex cable might cause some signal noise issues.

 

[FJC]

No, but people occassionally go into basements, and people sometimes do stupid things, like trip and fall and hit the shelf it is sitting on so that it then falls over, breaks the line, water spills out, the computer overheats and now you have a mess and a dead computer...

But yeah, essentially wild animals...

Again, it's not on a shelf - it's on a very sturdy table, in a fairly blocked-off section of the basement with no foot traffic.

Certainly the whole system adds risk. Water cooling itself adds risk, and this compounds that risk with more connectors, long hose runs and such. I did use thick tubing (3/8th inch inner diameter, 5/8th inch outer). On the floor of the study where I cut the holes I installed a plate similar to the one inside the computer, and used bulkhead fittings so that there's no tendency for the tubing to try to fall into the basement (thus taking strain off the section upstairs).

 

[Zarathustra[H]]

There's no reservoir downstairs in my final setup, just the one upstairs. It catches all the air, and I can bleed/top off from there.

I strongly considered just having a power supply downstairs, running the pumps/fans 24/7. That would work just fine. However, I'd have to be sure to remember to go downstairs and turn those pumps off BEFORE ever disconnecting the loop (for example, if I wanted to work on my PC). And conversely, I'd have to remember to turn the pumps back on prior to turning my PC back on. I tried to figure out how I could have a positive indicator whether the pumps were running or not from upstairs. So, I bought one of those little spinny flow indicators, and planned to incorporate that into the loop in my computer. I could glance at that to tell if the coolant is flowing or not before I turn on the computer. However, while doing my proof-of-concept tests I found that the darn little spinny flow indicator is noisy - it must be unbalanced, making a pulsing sound. As the goal was to make this system as quiet as possible, there was no way this was going in the loop. So I ditched it and went with the current plan, that the pumps are powered from the computer PSU. It's simple and it works, though I've been warned that my long molex cable might cause some signal noise issues.

My plan when I build my system is to have a relay in the basement, triggered by 12v power from my desktop power supply.

When the PC turns on, in triggers the relay which in turn switches on the 120v AC wall power to everything int he basement. I imagined picking up a 12v wall wart to power the pump.

The cool thing about this approach is that you could also use to to have a big box fan or something like that switch on, blowing in the direction of your cooling tower when the PC turns on.

I had found a premade wall outlet relay a while back when I started scheming about this build, but now I can't find it. I could always wire my own with a power strip and a separate 120v relay though.

 

[Zarathustra[H]]

My plan when I build my system is to have a relay in the basement, triggered by 12v power from my desktop power supply.

When the PC turns on, in triggers the relay which in turn switches on the 120v AC wall power to everything int he basement. I imagined picking up a 12v wall wart to power the pump.

The cool thing about this approach is that you could also use to to have a big box fan or something like that switch on, blowing in the direction of your cooling tower when the PC turns on.

I had found a premade wall outlet relay a while back when I started scheming about this build, but now I can't find it. I could always wire my own with a power strip and a separate 120v relay though.

Actually, here we go.

This should do the trick.

 

[Zarathustra[H]]

Actually, here we go.

This should do the trick.

And then use something like this to power the pump.

The beauty of bringing the pump off of the power supply is twofold. Firstly less load on the PSU, meaning less heat and potentially even less PSU fab noise upstairsairs.

Secondly, this allows us to use a 24vdc adapter instead of 12v on the pump (D5's are rated 8-24vdc input power) which might even get more oomph out of a single pump. (not sure about this though, haven't seen data)

 

[FJC]

Actually, here we go.

This should do the trick.

That's very cool. I've thought about seeing if I can take a Raspberry Pi and program my own fan controller/pump monitor, that could actually text my phone if something is out of order.

 

[guitarslingerchris]

And then use something like this to power the pump.

The beauty of bringing the pump off of the power supply is twofold. Firstly less load on the PSU, meaning less heat and potentially even less PSU fab noise upstairsairs.

Secondly, this allows us to use a 24vdc adapter instead of 12v on the pump (D5's are rated 8-24vdc input power) which might even get more oomph out of a single pump. (not sure about this though, haven't seen data)

My understanding is most D5s stop gaining flow rate after 13.1v but if you have the D5 strong it can push harder all the way to 24v or at least much further.

 

[Zarathustra[H]]

And then use something like this to power the pump.

The beauty of bringing the pump off of the power supply is twofold. Firstly less load on the PSU, meaning less heat and potentially even less PSU fab noise upstairsairs.

Secondly, this allows us to use a 24vdc adapter instead of 12v on the pump (D5's are rated 8-24vdc input power) which might even get more oomph out of a single pump. (not sure about this though, haven't seen data)

Meh, here is some testing.

Looks like small increases in pressure up to 13.1vdc, and then it flatlines after that, so little to no benefit of running a D5 at 24v

Can't hurt though. I'll take the minor gains at 13.1v

 

[Zarathustra[H]]

My understanding is most D5s stop gaining flow rate after 13.1v but if you have the D5 strong it can push harder all the way to 24v or at least much further.

Yeah I found that 13.1v test data right after posting.

Interesting. I had never heard of the D5 strong. I will have to do some googling.

 

[NoOther]

Again, it's not on a shelf - it's on a very sturdy table, in a fairly blocked-off section of the basement with no foot traffic.

Certainly the whole system adds risk. Water cooling itself adds risk, and this compounds that risk with more connectors, long hose runs and such. I did use thick tubing (3/8th inch inner diameter, 5/8th inch outer). On the floor of the study where I cut the holes I installed a plate similar to the one inside the computer, and used bulkhead fittings so that there's no tendency for the tubing to try to fall into the basement (thus taking strain off the section upstairs).

Don't get me wrong, its very cool and I think a great idea. The one thing I think that could make it a little better would have been maybe mounting it on the wall, building a little cage around it, perhaps also shorten the run, though it seems fine as is. Something with the ability also to change out components if needed. But it's very cool man.

 

[thesmokingman]

And then use something like this to power the pump.

The beauty of bringing the pump off of the power supply is twofold. Firstly less load on the PSU, meaning less heat and potentially even less PSU fab noise upstairsairs.

Secondly, this allows us to use a 24vdc adapter instead of 12v on the pump (D5's are rated 8-24vdc input power) which might even get more oomph out of a single pump. (not sure about this though, haven't seen data)

You don't have to go to that trouble. Just run a lead from psu to psu, aka dual psu plug. It's literally just a jumper.

 

[Dayaks]

I put my whole tower in my basement and ran extension cables upstairs. Zero noise/heat.

The potential problem with OP's setup is if the coolant gets colder than your study you get condensation.

So if anyone copies this you need to ensure your coolant temp > tower ambient (or more accurately dew point.)

 

[Zarathustra[H]]

You don't have to go to that trouble. Just run a lead from psu to psu, aka dual psu plug. It's literally just a jumper.

Well, in my application (which is different from ops) I plan on having other things start up when the desktop powers on as well, so having a relay switching on mains power becomes very useful.

 

[Zarathustra[H]]

The potential problem with OP's setup is if the coolant gets colder than your study you get condensation.

That's not exactly how it works.

You can go sub-ambient without getting condensation, you just need to stay above the dewpoint.

The dewpoint is dependent on temperature and humidity. The important part here is that humidity is relative, which is why it is expressed as a percentage. The capacity of air to hold water increases as it gets hotter, so in isolation, I'd temperature goes up, the same vapor in the air results in a lower relative humidity and vice versa.

The dewpoint is the point at which the temperature drops enough that the air can no longer hold all the humidity that is in it (in other words, RH hits 100%) and thus you get condensation.

You can use a dewpoint calculator like this one to figure it out.

Example:

If it is 80°F and 60% RH out, the dew point is 64.8°F, so you just need to keep your coolant temperature above that (with a decent safety margin to avoid accidents).

This assumes - however - that you are operating at outdoor conditions. If your office is air conditioned, this lowers both the temperature and RH. Let's use 70°F and 40%RH as an example. Now your dewpoint is 44.6°F. That's much lower than I'd expect the coolant temperature to ever go.

These are things to be concerned about with actively chilled loops, but I doubt you'd have to worry about it with a passive basement radiator setup.

It is true underground temperatures tend to stay in the 50's year round regardless of weather, but in reality basements are hollowed out and leaky to the world around them, so they tend to not exhibit perfect underground temperatures.

My basement gets into the 90's in the summer and gets pretty damned cold in the winter.

Given the fact that most air condition in the summer, the basement may have cooler air, but probably is not going to result in sub-ambient loop temps, and even if it does low RH and lower temps due to air conditioning will likely prevent condensation.

In the winter the basement still likely won't result in sub-ambient loop temps, but if it does, winter air tends to be very dry, so the dewpoint is very very low.

It doesn't hurt to keep an eye on the loop temp just in case, but I'd argue that unless there is something extreme about the climate or basement OP is using, this is perfectly fine and very low risk.

 

[Dayaks]

That's not exactly how it works.

You can go sub-ambient without getting condensation, you just need to stay above the dewpoint.

The dewpoint is dependent on temperature and humidity. The important part here is that humidity is relative, which is why it is expressed as a percentage. The capacity of air to hold water increases as it gets hotter, so in isolation, I'd temperature goes up, the same vapor in the air results in a lower relative humidity and vice versa.

The dewpoint is the point at which the temperature drops enough that the air can no longer hold all the humidity that is in it (in other words, RH hits 100%) and thus you get condensation.

You can use a dewpoint calculator like this one to figure it out.

Example:

If it is 80°F and 60% RH out, the dew point is 64.8°F, so you just need to keep your coolant temperature above that (with a decent safety margin to avoid accidents).

This assumes - however - that you are operating at outdoor conditions. If your office is air conditioned, this lowers both the temperature and RH. Let's use 70°F and 40%RH as an example. Now your dewpoint is 44.6°F. That's much lower than I'd expect the coolant temperature to ever go.

These are things to be concerned about with actively chilled loops, but I doubt you'd have to worry about it with a passive basement radiator setup.

It is true underground temperatures tend to stay in the 50's year round regardless of weather, but in reality basements are hollowed out and leaky to the world around them, so they tend to not exhibit perfect underground temperatures.

My basement gets into the 90's in the summer and gets pretty damned cold in the winter.

Given the fact that most air condition in the summer, the basement may have cooler air, but probably is not going to result in sub-ambient loop temps, and even if it does low RH and lower temps due to air conditioning will likely prevent condensation.

In the winter the basement still likely won't result in sub-ambient loop temps, but if it does, winter air tends to be very dry, so the dewpoint is very very low.

It doesn't hurt to keep an eye on the loop temp just in case, but I'd argue that unless there is something extreme about the climate or basement OP is using, this is perfectly fine and very low risk.

My basement would plus I use a 9x140mm rad.

I added dewpoint to my post but must have been while you were crafting yours.

Overall IMO it's a hell of a lot easier to put the tower and cooling in the same room and just run extensions plus condensation isn't a worry at all. You also don't have to worry about leaks drenching your living area, loop is easier to fill, ect.

 

[FJC]

Looks like I'm safe - we keep the A/C set at 74F with a target humidity below 46%. Even if the HVAC can't keep up with the humidity, I've never seen it go over 50% in the house. Per that calculator, the dew point is 54.2F. That's way cooler than the basement gets, even in winter. Typically my basement runs about 4 degrees cooler than upstairs.

Current humidity in my house - 38%. I do love my smart thermostat.

 

[FJC]

Overall IMO it's a hell of a lot easier to put the tower and cooling in the same room and just run extensions plus condensation isn't a worry at all. You also don't have to worry about leaks drenching your living area, loop is easier to fill, ect.

You're not wrong. I considered that as well - I could have just put the whole computer in the basement, and run long DisplayPort, USB, etc cables. But I do like just seeing my computer - all that RGB and tempered glass would go to waste in the basement. And I just sort of wanted the challenge of doing it this way. I like showing off the hardware to friends and family. My next build will be in one of the new Lian Li PC-O11 Dynamic cases, with glass on both the front and side. Gotta show off my nerd skillz, ya know.

 

[Zarathustra[H]]

I had forgotten just how pricy that EK dual D5 is. $110 just for the top. If you want it pre-assembled with two PWM EK D5 pumps, that's $275.

I guess redundancy is good, but if one is just going for the extra head pressure there seem to be much cheaper options, especially if out of earshot.

Koolance's 24v PMP-600 seems pretty damned kickass, and it only costs $95. (datasheet found here)

I can't find any reviews of it though, I guess that's because people willing to power their loops with 24v are probably rather rare.

An added bonus is that it appears to be air cooled, so it doesn't dump the pump heat into the water loop like the D5's do. This might be undesirable inside a case, but if the pump is free-standing in the basement, it would seem perfect.

I wonder what kind of pump that is under the Koolance branding. Eheim maybe?