First time custom water - hardline

Nebell

2[H]4U
Joined
Jul 20, 2015
Messages
2,383
Update 2016-04-10:

Aaaaand completed :)

warpty.jpg


2nbve3d.jpg


v7uziv.jpg



1r94za.jpg



1zh1iz8.jpg



Update 2016-03-17:
Here's the final loop idea. What do you guys think?
I think either front or back radiator is upside down. I'm not sure. Any suggestions?

2vmdm5j.jpg
 
Last edited:
1: no need to bend if you want to use angle adapters liberally. However it is a ton more expensive!

2: I use SP120 fans and they are awesome. I know if you want super quiet, 140mm based rads/fans are the way to go.

3: I've never dealt with parallel systems before, always seems like adding extra components that may fail eventually. I would suggest one big rad.

4: I use the Phanteks evolv, and I love the quality, don't trust thermaltake: my last case was a Tt case and the quality was sub-par.

5: PLAN PLAN PLAN. You need to ensure you plan every millimetre of your setup, especially with hard tubing. So long as you plan every single tiny piece, then you are fine to have any loop order you want. Go nuts.
 
reality is the Rad surface area is the Rad surface area. it can be split up however you want. on my setup it goes pump, 240 rad, CPU/GPUs, 240 rad, res. there may be slightly less resistance using 4 120s in parallel vs a single 480, but in the end its not going to make a difference.
 
My two cents are, rigid is expensive, but pretty. Rads- I agree with Nimisys; whatever fits / don't make it hard on yourself. Series rads add almost no restriction and work fine, plus, any time you put things in parallel I wonder about flow balance (least resistance and all). You can mix and match gear, but read up on it first. Case in point, a certain company had quick disconnects that look and work great, but their black chrome version seems to be **** and peels or dissolves where the distilled water comes in contact. But if you stay with the bigger names you should be good to go. Even my EK stuff has given no trouble (but it's plain copper).
 
Honestly don't remember, and can't find the thread now...

Push/pull means fan on one side pushes air into the rad, and the fan on the other side pulls it out~ push /push doesn't sound reasonable, unless there is another way for the air to exit the rad.
 
I don't know, it just feels like more fun separating and playing around with different setups. Which company are you talking about?

Also Nimisys, you don't have a reservoar? :p

I was thinking pump > reservoar 1 > 120mm 1 > gpu 1 > 120mm 2 > gpu 2 > reservoar 2 > 240mm > cpu > reservoar 1. Might be a lot of tubing though.

Also going double fans on each, not sure if I should go push/pull or push/push. I read somewhere that push/push have a little bit better temps.
Probably going to get Enthoo Primo then since it has a lot of space.

A} Loop order doesn't matter, so long as there's a pump after the reservoir.

2) You don't need 2 reservoirs.
 
I don't know, it just feels like more fun separating and playing around with different setups. Which company are you talking about?

Also Nimisys, you don't have a reservoar? :p

I was thinking pump > reservoar 1 > 120mm 1 > gpu 1 > 120mm 2 > gpu 2 > reservoar 2 > 240mm > cpu > reservoar 1. Might be a lot of tubing though.

Also going double fans on each, not sure if I should go push/pull or push/push. I read somewhere that push/push have a little bit better temps.
Probably going to get Enthoo Primo then since it has a lot of space.
I listed the red, its after the last rad.

What I don't get is why you would go pump to a res then into a rad or block.

In your case I would run pump, 120 rad, CPU, 120 rad, parallel gpus, combine into the 240 and finally into the reservoirs. You can let the pump draw from both and dump into reservoirs if you want to use two. Reservoirs serve two purposes, volume of coolant for thermal mass, and allow for thermal expansion. The more coolant you have, the more stable your temps will remain, however you still need some where for it to expand as it gets hotter, and with rigid tubing, you don't get tubing swell to do it for you.
 
Thats actually one thing that is essential: The res needs to be the LAST part of the loop. It destroys pressure, so you want it as the last stop before the pup pushes the liquid up to pressure again.
 
Thats actually one thing that is essential: The res needs to be the LAST part of the loop. It destroys pressure, so you want it as the last stop before the pup pushes the liquid up to pressure again.


A watercooling loop is a circle. The res is always the last part of a loop, or it is always the first part of a loop, or it is always the middle depending on how you choose to write it down. So long as pump follows reservoir nothing else matters much.
 
A watercooling loop is a circle. The res is always the last part of a loop, or it is always the first part of a loop, or it is always the middle depending on how you choose to write it down. So long as pump follows reservoir nothing else matters much.

Yeah, I did not explain that well, Yes, the Res needs to be immediately before the pump. (I forgot that I'm probably the only one who thinks the pump is the 'start' of the loop.)
 
You could put the res on a T anywhere in the loop though, you just end up with less of the coolant circulating. Assuming the actual loop is bled and the res is at the highest point, the pump would remain primed. Like a T line for filling but with more volume. Not really sure why someone would want such a setup, but it would work
 
Can I add a second pump?
I guess there are more instructions on the net about how to connect things properly. This is gonna be so exciting :)

An extra pump is commonplace in super high-end setup, and EK make an awesome dual-D5 pump block that would essentially be good enough for ANY setup. I would HIGHLY reccommend a D5 pump, as the only DDC pump I've ever owned howled like a banshee from day-1. I run a single D5 with XSPC res and it provides enough pressure for my 3 blocks. Any more resistance and a second pump wouldn't go astray.
 
h

Yeah I was thinking about getting EKWB EK-XTOP Revo Dual D5 PWM Serial (incl. 2x pump) - Buy at CoolerKit.com
But two smaller pumps do give me a bit more customization. Perhaps two of these? EKWB EK-XTOP Revo D5 PWM - Plexi (incl. pump) - Buy at CoolerKit.com

Here's my ugly paint job:

301q3w0.jpg


Of course it won't be as straightforward as I will bend tubes here and there, but that's the plan so far, unless I messed something up. GPU reservoar will be larger. Any suggestions? Maybe I should connect those two reservoars just because? :p
EDIT: Since I've decided to go with Phanteks Enthoo Primo instead of Thermaltake Core P5, then I should have place for three 2x120mm radiators. Good choice?

The reason why I'm separating cpu and gpu systems is because it should make it cooler, slightly expensive, but cooler. Now I'm not sure if I should separate GPU systems as well, that might be going for too much.

EDIT2:
What about this setup?

5b65xv.jpg


Pic #1 makes a lot more sense, as the point of separating the loops is to have a temperature differential between the two systems. Essentially, the GPUs' loop's heat is not affected at all by the CPU temp and vis versa. Having the systems connected (at any point) would make it so that the whole system shares the same temperature. However, as what was said before, unless the reservoir is at the highest point in the loop, it's best to put it right before the pump. And you ARE planning on connecting the end of the loop(s) to the pump(s), right? Because the picture shows a line, not a loop!

Also, an overclocked CPU can put out upwards of 190-220w of heat, wheras two OC'd video cards can output over 800w(!) of heat, so separating the CPU from the loop isn't really making a dent in the GPU's heat output. It essentially means that when you aren't streasing the CPU much, but you are pushing the cards to the limit, the radiator you put on the CPU loop is essentially wasted space, and may be better put to use on the GPU side of things, if that makes sense.

Overall, I'm ALL for serial loops. I don't really see a huge benefit to parallel, but it isn't exactly 'bad', I just don't see the point.
 
Yeah, honestly I forgot about connecting things back to pump :)

But damn you complicated things for me now :)
But maybe I'm just overthinking it. 800w from two GPU's is a lot compared to CPU's 200w, but I mean, I'm not really out to lower every temp to lowest possible, I just want to make a good looking loop. And I don't mind sacrificing a bit higher temp for that.
Also when I tried to buy loop #2 it was €1200... without fans. While I was planning to spend a bit more on custom loop, that amount got me thinking that I could buy 6700k + high end mobo + more goodies for that amount :)
I'd rather not go above €1000. I still need to buy a new case and a PSU since I'm not sure about my 850W.
I agree with you about serial loops. Maybe I should go down that road.
Should I go like pump > res > 4x120mm rad > cpu > 4x120mm rad (or two 2x120mm rad) > gpu's > another rad > pump?


I said this in another thread, and people got all kinds of pissed off at me, but it simply CANT be proven wrong: custom water cooling is NOT for performance.

"Stoopid, custom water is the HIGHEST performance!"

No, you see: you could easily buy a new Mainboard, RAM, and CPU and a high-end HSF for the same price as JUST custom water cooling your 3770 and cards. And that new cpu and higher ammount of RAM would give you MUCH more performance than the extra overclocking would provide for your current tech. It does not matter where your PC sits, buying newer, faster hardware will almost ALWAYS give you more performance than spending the same money on a custom water loop.

"But why do it then, HMMM!?"

Because, we want an amazing looking PC, we want a unique and beautiful piece of art sitting on our desks to show off to friends an Internet buddies. We want something that puts us in the 0.1% of users and most importantly: an experience of building something amazing that we can be proud of.

'Extra performance' is just the excuse we toss out there to justify our setups, because we don't want to sound like mushy girly-men by explaining the above.
 
My GPU's are different height. I was stupid enough to buy different ones :/
How much does that complicate things?

Amazingly. If the two cards are not reference models, then finding a compatible block will be REALLY difficult. however Alphacool will custom make one (for free, I've heared?) if you have a current, modern card that they don't have a block for.
 
I said this in another thread, and people got all kinds of pissed off at me, but it simply CANT be proven wrong: custom water cooling is NOT for performance.

"Stoopid, custom water is the HIGHEST performance!"

No, you see: you could easily buy a new Mainboard, RAM, and CPU and a high-end HSF for the same price as JUST custom water cooling your 3770 and cards. And that new cpu and higher ammount of RAM would give you MUCH more performance than the extra overclocking would provide for your current tech. It does not matter where your PC sits, buying newer, faster hardware will almost ALWAYS give you more performance than spending the same money on a custom water loop.

"But why do it then, HMMM!?"

Because, we want an amazing looking PC, we want a unique and beautiful piece of art sitting on our desks to show off to friends an Internet buddies. We want something that puts us in the 0.1% of users and most importantly: an experience of building something amazing that we can be proud of.

'Extra performance' is just the excuse we toss out there to justify our setups, because we don't want to sound like mushy girly-men by explaining the above.
Or maybe a custom loop is more powerful when discussing thermal management. Obviously faster hardware will always be faster hardware, but in the context of an OVERCLOCKING AND COOLING sub forum, discussing performance in terms of thermal management is not a hard conclusion to reach. Unless you want to be a nit picky bitch.
 
Or maybe a custom loop is more powerful when discussing thermal management. Obviously faster hardware will always be faster hardware, but in the context of an OVERCLOCKING AND COOLING sub forum, discussing performance in terms of thermal management is not a hard conclusion to reach. Unless you want to be a nit picky bitch.

If you're specifically talking about cooling performance, then custom loops are still a huge price for a small gain. With not much more invested, you can go phase-change and REALLY step up the cooling.

Price/performance: custom loops are completely useless. I run one, I LOVE my custom loop: it is one of the most beautiful systems I've ever built. But I don't lie to myself and say that every dollar spent on it was the best used for raw performance. I could have had a dual-CPU board and 128GB of RAM if I stuck to stock cooling. But like I said: water cooling is not for performance.
 
If you're specifically talking about cooling performance, that custom loops are still a huge price for a small gain. With nit much more invested, you can go phase-change and REALLY step up the cooling.

Price/performance: custom loops are completely useless. I run one, I LOVE my custom loop: it is one of the most beautiful systems I've ever built. But I don't lie to myself and say that every dollar spent on it was the best used for raw performance. I could have had a dual-CPU board and 128GB of RAM if I stuck to stock cooling. But like I said: water cooling is not for performance.

Depends on how you define "performance". A loop will keep your hardware running cooler while simultaneously being (much) quieter at just about any computational load you care for with the right fans at the right speed.
 
Depends on how you define "performance". A loop will keep your hardware running cooler while simultaneously being (much) quieter at just about any computational load you care for with the right fans at the right speed.

Noise is a good point. Custom loops are among the quietest setups you'll find, only really surpassed by true passive setups.

Noise (unless you are recording sound professionally) is an aesthetic thing, like the look of the system. It's one aspect to measure value, next to performance.
 
I don't know how far along in this build you are yet, but I'd caution against parallel loops like you have in both of the pictures of post #17.

Reason being - if you have any parallel path for flow, you aren't able to reliably control the split between the paths (*unless you invest in some flow meters and needle valves, which is adding a lot).

Sort of like you have in Pic #1 - if you really want to go parallel, each branch should have it's own pump and supply line (excepting the right-hand branch, where you have a split running from a single pump). They can all tie together from a common reservoir and combine into a common return header, but if any path doesn't have a dedicated pump behind it to push, you have no way to control or balance the flow, and it can lead to some unintended results with temperature control.

Bending tubing takes some practice, but is worthwhile to learn and makes for a very clean setup. Also fewer points for potential leaks.
 
I don't know how far along in this build you are yet, but I'd caution against parallel loops like you have in both of the pictures of post #17.

Reason being - if you have any parallel path for flow, you aren't able to reliably control the split between the paths (*unless you invest in some flow meters and needle valves, which is adding a lot).

Sort of like you have in Pic #1 - if you really want to go parallel, each branch should have it's own pump and supply line (excepting the right-hand branch, where you have a split running from a single pump). They can all tie together from a common reservoir and combine into a common return header, but if any path doesn't have a dedicated pump behind it to push, you have no way to control or balance the flow, and it can lead to some unintended results with temperature control.

Bending tubing takes some practice, but is worthwhile to learn and makes for a very clean setup. Also fewer points for potential leaks.

honestly this is blown out of proportion. I have run a fully parallel loop for almost 4 years now, CPU and two GPU's and never really had an issue. you get better flow rates than in serial, and in my setup, makes plumbing and servicing infinitely easier. it took me less than ten minutes to swap out my GPUs and have my loop running again. when one of my original GPUs died, I was able to remove it from the system without disturbing the rest of the components.
 
I don't see the point in setting up a parallel loop unless it is easier to plumb, and that seems like it would be pretty unlikely. If you don't set it up properly you can end up with worse flow through one leg of the loop depending on the relative restriction. In a serial loop you don't ever need to worry about this, all blocks get all the same flow rate. Also, there really is no reason to put a rad in front/behind every single component unless its easier than putting them anywhere else. Coolant temp within the loop reaches equilibrium and doesn't change much before/after individual components. The only thing that really matters in a WC loop is how much rad surface area you have. 360 is 360, whether its 1x360 or 240+120. Almost no other variable really makes a difference compared to this one, there has been years of debate about this stuff and testing kind of proves it.

I also think there is an issue with running multiple reservoirs, but I haven't confirmed it for myself. Basically, that dual reservoirs in a single loop should be run in parallel in order to make sure they fill properly/equally. I wound up doing that for my build (SMA8 Final Pics) but it does complicate things (behind the scenes: http://i.imgur.com/ChJ8uRa.jpg) There really is no reason to use two reservoirs other than for aesthetics unless you are running two totally separate loops.

Last word of warning/advice - Assuming both cards even have full cover blocks available - not having two identical cards is a major issue, especially if one of them is non-standard PCB height. If you're doing hard line it is even worse, if not downright impossible to do without making it incredibly complicated (i.e. plumbing each card individually). I would go so far as to recommend selling one of the two and acquiring another so that they match.
 
. Coolant temp within the loop reaches equilibrium and doesn't change much before/after individual components. The only thing that really matters in a WC loop is how much rad surface area you have. 360 is 360, whether its 1x360 or 240+120. Almost no other variable really makes a difference compared to this one, there has been years of debate about this stuff and testing kind of proves it.

I also think there is an issue with running multiple reservoirs, but I haven't confirmed it for myself. Basically, that dual reservoirs in a single loop should be run in parallel in order to make sure they fill properly/equally. I wound up doing that for my build (SMA8 Final Pics) but it does complicate things (behind the scenes: http://i.imgur.com/ChJ8uRa.jpg) There really is no reason to use two reservoirs other than for aesthetics unless you are running two totally separate loops.

Last word of warning/advice - Assuming both cards even have full cover blocks available - not having two identical cards is a major issue, especially if one of them is non-standard PCB height. If you're doing hard line it is even worse, if not downright impossible to do without making it incredibly complicated (i.e. plumbing each card individually). I would go so far as to recommend selling one of the two and acquiring another so that they match.
the only reason to run multiple reservoirs is total coolant volume. depending on space limitations, a single large tube may not fit, but two or even 3 small tubes may. while putting them in parallel would be preferred, again due to equilibrium, only one really needs to be in serial with the pump, the other(s) would still function off a single T-Line.
 
the only reason to run multiple reservoirs is total coolant volume. depending on space limitations, a single large tube may not fit, but two or even 3 small tubes may. while putting them in parallel would be preferred, again due to equilibrium, only one really needs to be in serial with the pump, the other(s) would still function off a single T-Line.
True, though total coolant volume really isn't particularly important. While greater reservoir volume makes filling the loop easier, it really only matters during filling. I remember my first custom cooling loop didn't even have a res, just a T-line. It took forever to bleed, but it was a very common practice in that era.
 
To sum up:

The bigger you Res, the slower your components will heat up.
  • The extra coolant will act as a liquid heatsink, The more material: the more energy is needed to increase its temperature. This is one of the BIG reasons why little AIOs aren't as effective.

The bigger your Rad, the lower your overall temps will be.
  • Rad surface area is important, the higher the surface area, the more heat you're able to suck out of the water and pump into the air. You get SLIGHTLY less effectiveness if you have a rad at the front blowing hot air into another rad at the top/back. This is a minor issue, though.

The bigger your pump, the quicker your system will reach equilibrium.
  • The faster pump speeds and higher pump pressure will ensure the hot liquid gets dumped into the Res faster, and goes through the Rad quicker, meaning the whole system will maintain a universal temperature more effectively.

Everything else is really visual and personal choice
 
Nicely summed up- anything outside this would be an outlier case, IMHO.
 
To sum up:

The bigger you Res, the slower your components will heat up.
  • The extra coolant will act as a liquid heatsink, The more material: the more energy is needed to increase its temperature. This is one of the BIG reasons why little AIOs aren't as effective.

The bigger your Rad, the lower your overall temps will be.
  • Rad surface area is important, the higher the surface area, the more heat you're able to suck out of the water and pump into the air. You get SLIGHTLY less effectiveness if you have a rad at the front blowing hot air into another rad at the top/back. This is a minor issue, though.

The bigger your pump, the quicker your system will reach equilibrium.
  • The faster pump speeds and higher pump pressure will ensure the hot liquid gets dumped into the Res faster, and goes through the Rad quicker, meaning the whole system will maintain a universal temperature more effectively.

Everything else is really visual and personal choice
Sorry, but in my experience this is not entirely accurate

1) The impact of res size is not worth considering. Yes, there is hysteresis in the loop that increases with coolant size, but unless you're going to have a 55 gal drum of coolant your loop will reach equilibrium within a gaming session no matter how much coolant you have. And don't forget, more coolant also means that your components will stay hotter longer after you've removed the load. If you have enough rad space, having enough coolant to fill the loop is all you need. Anything extra does not give you performance. This is NOT why AIO's are less effective. AIO's are less effective because they have crappy radiators, pumps, and waterblocks.

2) Agreed that area is important, but to clarify, its frontal area specifically. Rad thickness gives performance, but there are testing results that show that its not as important. I believe it was Martin's Liquid Lab that compared Alphacool 30, 45, 60mm rads to each other to show the impact. Get the thickest rad you can fit for best performance, but don't prioritize thickness over frontal area. (i.e. 240x30mm > 120x60mm)

3) High pump speeds are better, but its also not really worth worrying about unless you have a super complicated loop. A single D5 is enough for probably 90+% of all loops out there.
 
Back
Top