An ultimate beginners’ guide to PC water cooling
What’s the difference between a guy with a hobby and an enthusiast? The way I feel it, an enthusiast is always okay to sacrifice more time and money for something in which (s)he revels. An enthusiast is someone who has crossed the line between the “reasonable” (from a person-with-a-hobby’s point of view) commitment to a hobby, and almost an unhealthy addiction to it. If you find it hard (and annoying) to explain why the heck you’ve bought the 10th guitar head and why the previous nine were not enough — forget it, you’re an enthusiast buddy, and only those of your kind will genuinely comprehend your way.
With that being said, I have a confession to make — I am a PC enthusiast. Most likely this started from video games that I have been sincerely loving since I was a kid. I was both enjoying big single-player titles and thriving in competitive multiplayer games, even participating in local championships in my town. Anything that would run on my budget aging PC my parent bought me in the high school would do.
Time went on, and I started to make small amounts of money myself (summertime jobs mostly), saving a part of it for my hobby. I remember very vividly that feeling of joy and happiness when I held my first gaming mouse, Logitech MX518, which replaced my previous $10 device, probably from Genius. At that very moment, I felt like a professional e- sports player, no less. As I grew older, my income increased and I was able to throw more money into this: large mouse mats that cover half of the desk, mechanical keyboards, monitors with increased refresh rates, expensive headphones with high-quality standalone mics, top-tier graphics cards… And strangely enough, as I was spending more and more money, I started playing less at the same time. Technically, I was still in front of a monitor most of my time, but now watching YouTube hardware reviews more often than actually playing. Hardware and PCs themselves became my new passion, which doesn’t make sense to many people around. Who would want to spend so much on something you do not even put to good use? Well, an enthusiast.
“Now, wait a minute, isn’t this article about water cooling, why am I reading your life story dude?” Don’t worry, we will get down to this right away, but before we start I feel the urge to post this: liquid cooling is entirely an enthusiast thing. It isn’t to say that no one else is allowed to do that, but if you are not ready to spend time, spend money, get disappointed and spend even more — stick to traditional air cooling or pay someone who will build a loop for you: there are many talented PC modders who do amazing client builds. I want not to discourage you with these words, but to save you from a potential setback. And if you still ask yourself whether or not you want it, and the answer is still “Hell, yes!” — welcome to the club :)
I hate reading an article only to find out a couple of paragraphs later, that it’s nothing but a giant ad disguised as a useful material. For that reason, I have to warn you that throughout the article, I mention products by EK Waterblocks much more frequently than other vendors’ stuff. However, the only reason behind is that I primarily work with EKWB parts myself: it’s rather complicated to get water cooling components in the country I live in, and I have to buy everything online. Orders made in Slovenia (that’s where EK’s headquarters is) ship faster and cost less, compared to shipping from, say, “Performance PCs” in the US. I am not (unfortunately) sponsored by EKWB, not employed by or affiliated with them in any other way.
Before we start, I’d like to address some of these questions for those who haven’t made up their mind yet
“Do I need water cooling?”
If you put it this way then no, I doubt anyone needs it at all. One could barely imagine a real- life scenario in which you could say that liquid cooling is an absolute must. Sure, you can significantly decrease your hardware temperatures, but that does NOT mean you cannot be all right with air cooling. A brand-new Tesla X can be fantastic, but that does not imply you cannot keep driving your 2010’s Ford that is still in great shape.
The best go-to scenario for a water cooling is probably having a top-tier graphics card (or several of them). For instance, NVidia’s TitanX and 1080Ti cards are ultimate powerhouses with an ultimate TDP (thermal design power) of 250 Watts. The amount of heat released by those 250 Watts of energy is gargantuan, and you need to dissipate it somehow. Otherwise, as a card reaches its maximum allowed temperature, it attempts to cool itself by automatically decreasing its operating frequency. This process is called “thermal throttling,” and it is entirely normal. In practice, that only means a card starts operating at its normal clocks, giving up those extra frequencies provided by the Boost technology.
Yes, heat is one of two most notorious hardware killers (the second one is over-voltage), and if you find a right way to dissipate those 250 Watts of heat, you prolong an expensive graphics card’s lifespan. However, with relieving components from an overheating threat, you instantly introduce them to another one — leaks. Compared to heat that causes slow degradation over a long time, a liquid is an instant hardware killer that is far more dangerous. So nope, even with top-tier components that are HOT, water cooling is not a miraculous solution.
How much would a custom loop cost me?
I would like to pair this question with “How much time will it take to build one?” and give one answer to both of them: “Only God knows”. Let me demonstrate the way the cookie crumbles with another story.
My first water cooling experience started with plans to assemble a custom loop in an NZXT S340 Elite case — a cozy and beautiful chassis many people love. I did a fair amount of investigation (or better, I thought I did) and concluded that I would manage to install three radiators (1x240 in front + 2x120 at the top and rear). Which I was unable to do even after drilling two holes for pass-through ports at the top of the case (I seriously considered mounting a radiator outside the case after numerous attempts to fit it inside), and which ultimately led to buying an entirely new case that was relatively larger. A new chassis — a new pump\res combo! I didn’t need to go with the smallest combo available anymore, so I ordered a bigger tube. Which did not fit, yay! I mean technically it did, but not in the way I wanted it.
Long story short, the 240-mm radiator was the only piece of hardware that survived long enough to see the final build. Other parts went into the closet or were sold to some random folks around the Internet (used liquid cooling parts are not exactly the easiest thing to sell, so be prepared to make huge discounts for people willing to buy them). In the end, I spent somewhere around € 1000 and five months to build a pretty simple loop with 1 CPU and 1 GPU.
Not only does the final build cost depend on how accurate your initial plan was, it also hugely depends on what parts you use. Same rule as with everything else in this world applies to water cooling — parts from well-known manufacturers (EK Waterblocks, Bitspower, Phanteks, Singularity Computers, etc.) cost more. If you are on a budget that you do not want to stretch a lot (although you are almost certainly doomed to do that once you start building) — you can save quite a few pennies by purchasing stuff from other vendors. For instance, Barrow: a Chinese manufacturer that features a wide range of water cooling parts with very decent quality. There is a video on JayzTwoCents YouTube channel (be sure to check out this guy, he has tons of useful content regarding water cooling) where he compares the EKWB’s Supremacy EVO block with the cheapest, the ugliest, almost DIY- looking waterblock from Amazon that is five times cheaper. And they performed virtually identically! Makes one wonder whether its worth to pay more at all.
The tricky part here is to know what components are okay to purchase from AliExpress for nothing, and which are more quality-dependant. We will get to this later when we start talking about each loop node separately.
Custom loops vs AIOs
AIO (all-in-one) solutions are aftermarket coolers with factory-assembled “closed” loops. Some of them allow moderate maintenance like re-filling the coolant that may evaporate over time, but in a nutshell, you just take them out of the box and install them. There are much more CPU AIOs than similar solutions for graphics cards because motherboards have unified CPU sockets, unlike GPU vendors that may go for any “non-reference” board design. At the same time, water-cooling a CPU makes much less sense than water-cooling a GPU since processors emit much less heat and a good old “tower” air cooler will do just fine (often, even better than a mediocre AIO).
Apart from that, AIO issues are:
- most of them use thinner rads and aluminum parts, which means AIOs are less efficient when compared to all-copper custom loops;
- aftermarket solutions are often impossible to maintain and\or repair;
- their unified design means you cannot adjust them to your specific chassis (e.g., shorten tubes a little).
On the bright side, AIOs are much cheaper and, sometimes, sexier. I’ve been using the NZXT Kraken x52 for a couple of months and boy was it beautiful!
Another good use-case for buying an AIO is building inside tiny mini-ITX cases that cannot accommodate large (read that as “effective”) air coolers.
To sum up, if all you need is a CPU cooling solution that is compact and easy to work with — AIOs are worth buying. Other than that, if you have a roomy case (especially manufactured with having a custom loop in mind), I would not recommend that. AIOs aim to be jacks of all trades, but realistically are masters of none.
A water cooling loop follows the same principle an air cooler does: it takes heat from a hardware (e.g., a CPU die) and dissipates it into the air. Any loop consists of the following nodes:
- pump — pushes the liquid through your loop;
- reservoir — stores excessive liquid and feeds the pump with it;
- waterblock for CPU\GPU\memory\motherboard VRMs — uses a coldplate to make a direct contact with a hot hardware, transferring heat from this hardware to a heat spreader inside the block’s body;
- coolant — a liquid that flows through a waterblock’s heat spreader and captures its heat;
- radiator — cools the liquid down by forcing it to flow through narrow pipes with fins attached to them: fins create a huge dissipation surface area which ensures the liquid heat is rapidly taken away into the atmosphere;
- tubing — rigid or flexible tubing that connects all these parts into a single loop;
- fittings — connect tubing with other loop nodes and ensure the loop is sealed.
And a couple of auxiliary nodes that you may or may not have:
- fans — not those fans that admire you and bring flowers to your doorsteps, no. I’m speaking about fans that move the air through radiator fins to accelerate the heat dissipation. Although 99% of water cooling loops do have fans attached to radiators, these are not by mistake in the “auxiliary components” section. Technically radiators can dissipate heat all by themselves, especially if you have a substantial ultra-tower case that can accommodate super long and thick rads that create a massive surface area. However, realistically, you will most likely need fans for your build;
- drain valve — another secondary component that is still an absolute must: provides an ability to drain a loop without splatting a coolant all over the case;
- fill port — although you can always fill a loop through spare reservoir ports, using a dedicated fill port accessible from the outside of a case is just a more natural way to have the job done;
- flow indicator — a transparent box with a valve that rotates when liquid moves. Allows you to quickly identify how fast the liquid is moving inside a loop (and whether or not it moves at all). Somewhat useful in case your pump is set to low speeds, and the entire system remains dead silent when running;
- pass-through ports — rings with threading on both sides to connect with fittings. Used when you need to guide a tubing through an obstacle (e.g., through a PSU shroud into the case basement).
Now that looks complex, doesn’t it? Don’t worry; we shall break it all down below.
Pump and Reservoir
Back in the days, the entire liquid cooling department had significantly less attention from the public. It was a thing for a very small circle of people. There were much less ready-to- use parts you could buy, not to mention they would cost you a lot more than today, so people often went for some wicked DIY solutions for their loops. I’ve seen quite a few mentions of regular fish tank pumps employed for cooling PC hardware. Since those days water cooling has grown dramatically and you can finally leave your fish alone.
A pump is an electrical motor that rotates an impeller, which creates pressure that moves the liquid. The pump case connects with a top that has an inlet and outlet ports.
The liquid that flows inside a pump cools this pump down and, most importantly, lubricates it (something called “a wet rotor design”). For this reason, never run your pump dry; doing so may kill the pump in a matter of seconds.
In a couple of years, the initially large variety of available pumps has narrowed down to two renowned models from the same manufacturer: Laing D5 and Laing DDC. You may also meet the “Xylem” name instead of “Laing” but don’t be confused, those are different brand names for the same vendor. The following figure from the EKWB blog illustrates both models sitting side-by-side (D5 on the left, DDC on the right, custom EKWB tops mounted on both of them).
The D5 pump is larger, runs cooler and quieter, delivers a maximum flow rate of ~1500 liters per hour and provides a maximum head pressure of ~4 meters (a head pressure demonstrates how high a pump can push the liquid if the outlet port aims 90 degrees upwards).
The DDC pump, in turn, runs much hotter, has a bit lower flow rate (approx. 1000 liters per hour), but delivers a much higher head pressure (up to 7 meters). It also occupies significantly less space, which makes it a better choice for cozy cases. Also, the higher head pressure means DDC pumps are better to use when you have multiple water blocks and a lot of turns in your loop since those elements increase the flow resistance and make the liquid lose part of its initial pressure.
As mentioned before, a pump has a top housing (or “pump top”) which provides inlet and outlet ports. That’s where 3rd-party vendors come into play. See, you can buy a DDC or D5 pump from almost any liquid cooler seller: Alphacool, Swiftech, EKWB, etc. And all of them will have the same Laing (Xylem) pump at their core. The difference between these models is the pump top they use. Vendors claim that their after-market tops deliver increased performance compared to the factory top (e.g., EKWB states that their EK-XTOP DDC 3.2 PWM Elite pump increases the hydraulic performance up to 15%). If you wish, you can purchase a pump and a pump top separately, then assemble them manually, but this hardly makes any profit compared to buying a single pump+top unit. For simplicity, I will refer to a pump+top combo simply as “pump” from now on.
Reservoirs and combo units
If pump tops are not a big concern since you will most likely have them pre-installed, reservoirs have a bit more about them. First of all, again, you may have a pump and res as two separate units, or as a single combo unit.
Much as with pump tops, the choice is entirely yours as it has no impact whatsoever on the overall loop performance. Pump\res combos are more comfortable to install but require more vertical space. Separate units are better for more sophisticated loop designs where you can hide the pump in the case basement (under PSU shroud).
Regardless of your choice, remember that you must not run the pump dry, so you need to have a reservoir directly feeding your pump. Place a res must somewhere above the pump with its outlet connected to the pump’s inlet, and let the gravity do the trick.
Tip: Pumps have pre-defined inlet\outlet ports whereas reservoirs do not — if a res has multiple ports you can utilize any two of them. Just do not forget to plug the rest of them.
There is also a rectangular pump\res combo format that fits 5'25-inch case bays, which typically hold DVD\BluRay Drives. Following a modern trend, major case manufacturers keep ditching those 5'25'’ bays to favor an unrestricted airflow and a sleek appearance, so these pump\res combo units become less and less popular. But yes, these are still viable if you want them.
Many blogs and videos focus on differences between D5 and DDC versions, technical specs and so on, but say nothing about this, at first sight, simple aspect of a build. First, find a user’s guide that came with your case or Google for it. Find a section that should be called something like “water cooling compatibility” and check for pre-drilled holes your chassis has. For instance, my Phanteks Enthoo Evolv has a bunch of mounting holes in a motherboard tray and the bottom panel.
Inspect all mounting spots suggested by a manufacturer and search for other people’s build around the internet — this should give you a clue what units can fit in your case. You may need to drill holes manually and\or use tools like Dremel to cut through case parts that are in the way. There’s also a more straightforward solution that can save you from potentially ruining your case: pump brackets. Typically, you have a radiator on the case front or bottom side, fans facing outside to intake air and push it through the rad. Fan mounting holes on a radiator’s opposite side remain free. Brackets attach to these spare holes so that your pump \res combo is mounted directly to the radiator.
Radiators are the only loop parts responsible for disposing of the heat while remaining nodes merely transfer it, so it is quite relevant to pay attention to what you’re buying.
Forget everything you heard before: size DOES matter. The bigger your radiator is — the more surface area it provides — the better it dissipates the heat.
Radiator lengths are measured in the number of fans you can stick to one side. So you have “single,” “dual,” “triple” and “quad” rads to chose from (there are rads for five fans, but it’s pretty hard to find a case that can accommodate these monsters). Two most common fan sizes are 120x120mm and 140x140mm so you can calculate the radiator length. For instance, a duo rad designed for 140mm fans will be marked as a “280mm” rad. In return, if you see a “360mm radiator” in store, you should be able to read it as a “triple rad for 120mm fans”. The sexy white XSPC radiator at the figure below is a quad 480mm rad compatible with 120mm fans.
The second radiator size dimension we are interested in is thickness. Unlike length that is unified to fit fans, the rad thickness is entirely up to a manufacturer. Typically, you have sub-30mm “slim”, 30–35-ish “medium” and 40+ “thick” rads. You can find 80+ mm rads, but again, that is if you have a giant ultra-large case that has enough room for these bad boys.
A ubiquitous question is which size dimension has more impact on the loop performance, i.e., whether a slim but long radiator will outperform a thick but short one or vice versa. Remember this: increased radiator length gives your loop a much, much more significant boost than an increased thickness. Always go for length first, then thickness because the latter does not usually have that much of a difference. To a certain extent, however. Take some time to study this xtremerigs.net radiator review. Pay attention to 360mm radiators from EK Waterblocks: the difference between a medium and a thick models is negligible, just 5.7 Watts\10deltaT. But switching to the slim SE line gives you a massive performance drop of 25.9 Watts\10deltaT, which places SE radiators among the worst. So my advice would be going for medium-thick rads while getting as much length as your case allows you to.
Speaking of cases, most manufacturers do list supported radiator sizes in case specs.
Unfortunately, this applies to radiator lengths only, which makes sense because the thickness is limited by your specific hardware (for example, you may be unable to install thick rads on top because of your motherboard’s tall heatsinks, or in front because of a long graphics card).
Another problem with case specs is a lack of information about the possibility to utilize two radiator spots simultaneously. For example, the Phanteks Enthoo Evolv case supports 360- mm rads in front and on top, but is it possible to mount two medium 360-mm rads at both spots? The quick answer is “no”, you will have to go with either a slimmer SE rad in one place (which is not good given the results in the aforementioned review), or mod the case top so it would be possible to squeeze fans above the top rad, or (preferably) mount a 360- mm rad on top and a 280-mm rad in front. A double 140-mm radiator will perform close to the triple 120-mm one, and the PSU shroud will not allow anyone to see that a front rad is shorter than the top one and does not stretch all the way down.
To sum up, be sure to investigate what rads your case can accommodate thoroughly. Read radiator specs on a manufacturer’s website, measure your chassis, Google for other folks’ build logs and so on. Poor planning can be quite costly.
Another radiator spec is fin density: the number of fins that dissipate heat measured in FPI (fins-per-inch). The more fins you have, the larger the dissipation area is, hence the better. On the downside, you may need to run your fans at higher speeds to force the air through rads with a high fin density, which increases the overall system noise. Again, FPI is a less critical spec than the radiator length. If you have a sufficient surface area, it would be better to go for slightly more “transparent” rads with a bit lower FPI. Try guessing which of the EKWB radiators shown below has a lower fin density.
Let’s say you’ve done your homework and picked rads that fit in your case. Is there a way to tell whether or not they will deal with a rig’s heat efficiently? Well, typical advice would be something like “go for one 120mm rad per each cooled component included in the loop, a 240mm rad if you are going to overclock that hardware”. It is an okay rule, but personally, I prefer getting more precise and detailed answers. And here’s one: find out how many Watts of heat those rads can dissipate. If a manufacturer’s web page has no such specs, contact their support team with this question, they should be able to give you at least an approximate number. Then, seek for TDP numbers for each component cooled (e.g., an Intel i7–7700K processor plus an NVidia 1080Ti card equals 341 Watts). I guess you’ve already figured what one should do with these numbers: the total rad dissipation rate must be at least slightly higher than the overall hardware TDP.
This topic applies to the entire loop, but since radiators are the most massive metallic components of a circuit, I’ll leave it here.
Do not mix different metals in one loop to avoid corrosion
If you’ve chosen a copper rad, make sure to avoid aluminum parts anywhere inside other loop components. Vice versa, do not buy waterblocks and fittings made of copper\nickel\nickel-plated brass if you have aluminum rads. Copper is known to be one of the best heat conductors among metals, so just stick to copper parts, and you’ll be fine.
A water block is a heat spreader hidden in a (typically) acrylic body. One side of a heat spreader touches a die with its polished cold plate, while the opposite side provides a stack of microchannels that contact a coolant. These microchannels save the same purpose as radiator fins — they enlarge the contact surface area to accelerate the heat transfer.
The overall construction is pretty straightforward with no moving mechanical parts or high- tech production requirements, which makes waterblocks probably the best candidates to save a few bucks on.
Speaking of CPU waterblock, buy whichever fits your motherboard socket. Modern blocks have adaptable mounting brackets for almost all existing Intel and AMD sockets. Some blocks, like Supremacy EVO from EK Waterblocks, may require quick customization since they ship with several inserts and jet plates enclosed and depending on your processor you may need to replace the ones which are factory-installed. Just read the installation manual that comes with your block and this procedure should not pose any difficulties.
With GPU blocks things get more complicated because of different card manufacturers have different board designs. The key word here is “reference.” Be careful — it does not mean the same thing it used to a few years ago. I’m going to use NVidia cards as an example, but same applies to AMD products as well.
A “founders edition” design is a PCB of NVidia’s Founders Edition (FE) cards. These are cards manufactured directly by NVidia (or copied entirely by their partners) and have NVidia-designed blower coolers.
A reference design is a PCB design developed by NVidia for their partners (Asus, Gigabyte, MSI, Zotac, etc.). These 3rd-party manufacturers clone the NVidia’s reference PCBs and attach their cooling solutions, colored shrouds, RGB LEDs, etc.
FE and reference designs are not the same and may feature different board elements’ layouts and\or back I\O panels.
For example, this is the compatibility list for the EK-FC1080 GTX Ti water block. Note that many 1080 Ti cards with a reference board design (like Inno3D 1080Ti X3 or EVGA 1080Ti SC2 Gaming) are missing from this list. It means the “FC1080 Ti” block is compatible with Founders Edition cards only, whereas for reference design card you need to buy the EK-FC GeForce GTX FE block instead (which, ironically, has the “FE” in its name while the actual FE-compatible block does not). I personally contacted EKWB with this question, and the answer was that the FC1080 block MAY fit a reference card not listed among supported ones, but you have to try it at your own risk.
Custom (non-reference) cards manufactured by NVidia’s partners feature both custom coolers and reworked PCBs. Major waterblock manufacturers offer blocks that exclusively support specific card brands. For example, the FC1080 G1 block is designed for top-tier Gigabyte G1 cards, the FC1080 Aorus block also supports Gigabyte cards (but those of the Aorus model line only), and the FC1080 TF6 block fits MSI graphics cards.
Manufacturers need to receive a sample card from GPU makers and research it before they can start developing a compatible plate, so blocks for non-reference GPUs are typically unavailable for some time past a card release. What’s worse is that some non-reference cards have no full-cover blocks that fit them, at all. For instance, the FC1080 Strix block supports cards from the Asus Strix line, including 1070 cards. However, an Asus 1070 Dual card is not compatible with this block. And since it has a non-reference PCB, reference blocks won’t do either. Plan a loop long before you start picking its actual parts: be sure you will be able to get a block for a GPU you’re buying.
Reference VS aftermarket GPUs
Many first-time builders are confused whether they should get a top-tier after-market card “A” or a cheaper reference card “B.” Given that you found full-cover blocks for both cards, the question is not that simple. Buying card “B” seems more reasonable at first sight, since both cards have identical GPU dies and memory chips and the majority of an aftermarket card “A”s price is its custom cooler (which you are going to replace with a waterblock anyway). However, aftermarket cards can have separate minor advantages that make them a viable choice.
- Reworked power phases may boast better performance and push the overclocking potential a bit higher.
- More expensive VRMs and capacitors are more likely to last longer if you are continually subjecting your card to extreme load levels.
- Onboard fan connectors in Asus Strix cards allow you to monitor fan speeds based on GPU temps without employing third-party software like Speedfan.
- Finally, a good-looking and well-built stock backplate can spare you from buying a separate one (given that it is compatible with the installed waterblock).
The choice is yours to make; purchasing an expensive aftermarket card for water-cooling is not necessarily a waste of money as many do believe.
Universal GPU blocks
A few paragraphs ago I mentioned that you might be unable to find a “full-cover” block for a non-reference card. Some people believe that a “full-cover” block is a block that stretches along the entire board. This is not quite right, the FC1080 TF6 block leaves both sides of MSI cards’ boards open, but it’s still a full-cover waterblock. A full-cover block touches a GPU die, memory chips, and VRMs. As opposed to that, a universal block cools the GPU die only. It is quite evident that universal blocks are less efficient and require additional air cooling for card VRM zones. You should buy these blocks only if you’ve purchased a GPU before planning to build a water cooling loop and found out there’re no full-cover blocks for your card.
Most modern cards have stock backplates that increase the card durability and somewhat help with cooling board VRM sections. If you can mount a stock backplate back onto the card after installing a waterblock (e.g., the FC1080 TF6 block mentioned above is compatible with stock MSI 10x0 Gaming card backplates) — excellent. If not, you may buy a replacement from the same water cooling manufacturer. A waterblock filled with a coolant adds a lot of weight to your card, so having a backplate as an additional reinforcement is nice. So is a minor VRM cooling. But for me, both of these are not principal backplate advantages. What’s most important, you get decent leak protection: the GPU stretches across the entire case, and if a leak forms somewhere in the upper part of a loop, an unprotected GPU has very high chances to get soaked and fried. And secondly, a proper backplate adds a lot of sexiness to your build. And don’t even bother saying that it’s not a relevant factor to you! :)
RAM Blocks, Motherboard Blocks, and Others
Nowadays you can water-cool almost every part of your PC: RAM sticks, motherboard VRMs, disk drives, M.2 sticks — you name it. Doing that is optional, most of these parts are very durable and designed with operating at high temps in mind. For instance, Corsair gives ten years warranty for their overclocked memory sticks — they wouldn’t do that with fragile hardware, right?
Most vulnerable parts among all of these are, probably, motherboard capacitors. Manufacturers only state how many hours will these capacitors endure at a particular temperature, e.g., 5 years at 130 degrees Celcius. Operating at higher temps burns through capacitors’ lifespan and water-cooling them might be reasonable. And it does not necessarily mean you need to buy an additional piece of equipment: many vendors offer MB blocks combined with regular CPU blocks into something called “a monoblock.” Like the EK block for the Asus Rampage V Extreme board.
This bad boy has multiple cold plates to touch a CPU, a PCH, VRMs, MOSFETs — anything that gets hot under heavy loads. Which means if you get a monoblock, you don’t need a regular CPU block anymore.
But, to quote John Constantine, “there’s always a catch” — same as graphics cards, motherboards have unique element layouts. And for that reason, you cannot get a “universal monoblock” that would fit any board, as it would work with a simple CPU block. Instead, you will need to (again!) do some research looking for those blocks that are compatible with your specific hardware.
If you decide to include additional water blocks in a loop, remember that every one of them increases the coolant temperature and the overall resistance that reduces the fluid head pressure. To neutralize that, you will need to enlarge the radiators’ surface area and ramp up the pump a bit higher.
Fans are probably the worst loop component to write (and read) about. All reviews you can find are more or less subjective, and no one guarantees you can get what you hope for before you purchase a specific fan and test it yourself. I’ve seen a JayzTwoCents’ review where Jay was literally touching a spinning EK Vardar fan with his ear, claiming he can barely hear it working. My experience with Vardars was different: sure these are still lovely fans in terms of performance, but not as silent as I thought they would be.
Commonly, you find two fan types on the market: airflow optimized fans that aim to create the highest airflow possible, and static pressure optimized fans designed to produce high air pressure to push the stream through any obstacles. Canonical examples for these fan types are Corsair AF (stands for Air Flow) and SP (static pressure) fans.
In this figure, the left fan with blue rubber ring is an SP model, and the one on the right with a red collar is AF. Just by looking at these you can suggest that the difference between an airflow fan and a static pressure one is their blade shape — and that would be correct. SP fans have wide and flat blades that cover the opening almost entirely, which leaves no room for the air to escape and forces it to move forward through any obstacles. In a water cooling loop, main barriers in a stream’s way are radiator fins. This leads many people to believe that static pressure fans are the only viable choice for a water cooling. For the most part, that is true: airflow-optimized fans can be unable to push the air through fins effectively. However, it is not a rule of a thumb. Radiator with lower FPI count is more “transparent” and is not a big issue even for airflow fans. Moreover, some all-rounder fans, not advertised as static-pressure ones, deliver a decent air pressure. For instance, you can easily mount stock Phanteks Enthoo Evolv fans on a rad, and they will outperform Corsair’s SP LED fans, which are much worse than regular SP models from Corsair. As always, “the truth is out there,” so read and watch as many reviews as possible to figure out which is a decent fan model, and which is a paid ad.
Tubing and Fittings
Since tubing cannot latch onto water blocks without fittings, let’s discuss both of these parts in the same section.
Rigid or flexible?
There’re two tubing types: rigid and flexible. A flexible tubing is basically a hose, similar to what AIOs utilize.
Rigid tubing is, well, rigid. Pretty self-explanatory, isn’t it? :)
And as you may have guessed, there’s no right answer to “which one is best?” again. I’ll try to provide some insight, but it all comes down to your own best judgment once again.
- Safety. I am convinced that safety-wise flexible tubing is best, hands down. Compression fittings provide a “nose” that you need to push into a hose, and the grip so tight that it takes a lot of force to take that tubing off even when you want to. Coupled with a compression fitting’s ring that seals the connection, it’s nearly impossible that the tubing will accidentally slip. There are compression fittings for rigid tubing as well and don’t get me wrong, they are not loose or dangerous either. But if you try pulling a rigid tubing by hand, it will detach much much easier.
- Complexity. Flexible tubing is easier to work with: cut the hose, connect both sides and you’re good to go. Just be careful not to bend a tube at a very sharp angle to avoid a kink that can limit or completely block the flow. In turn, rigid tubing requires more effort: you will either need to bend it using additional instruments (e.g., a heat gun), or buy 45- and 90-degrees adapters to route straight tubing pieces from one loop node to another.
- Materials. Any flexible tubing is, give or take, the same. It may be clear or colored, covered with an external layer or not, but in a nutshell that’s just a flexible hose. With rigid tubing, you have much more options. As a beginner, start with PETG tubing: it’s cheap, easy to bend, and much more durable compared to acrylic tubing. More advanced builders may go for glass tubing because of its crystal-clear appearance and high resistance to even most aggressive chemicals (remember what all flasks and vials for chemical and biological labs are made from?). Finally, an experienced modder can use copper tubing for a steampunk-styled PC. Or carbon. Or metal. Or plastic. There’s a lot of options to choose from.
- Chemical resistance. I’ve already mentioned glass tubing being the most resilient tubing you can use. Cheap options(flexible tubing, acrylic, PETG) are typically less durable. Flexible tubing slowly leaches plasticizers into the liquid, which clogs the loop and turns the tubing itself opaque. PETG tubing is sensitive to propylene glycol, so if you go with aftermarket liquids, check their contents first.
- Price. Given that you don’t opt for something fancy like copper or carbon tubing, both flexible and PETG\acrylic tubing is equally cheap.
- Appearance (aka sexiness). Remember the beginning of the article when I said there’s barely a real-life scenario that fully justifies a custom loop, and so you should go for it only if you count yourself an enthusiast? Well, following the same logic, if you were all about efficient cooling and nothing else, you would’ve purchased a high-airflow case, a couple of high-performance fans, and built a well-thought air cooling layout. But you’ve chosen a custom loop, something entirely optional, so in my humble opinion, you should push it to the limit and make the build look as satisfying as possible. And that spells rigid tubing. Spending a lot of money and effort on a build to have lianas of flexible tubing dangling in your case is very debatable if you ask me.
Tubing and fitting sizes
Tubings are labeled with two numbers: inner and outer diameters, or ID and OD. For instance, the tube marked as 12–16mm (or 7/16'’ — 5/8'’) has the OD of 16 millimeters with the 12-mm opening, which means this tube’s thickness is 2mm.
Tube size (and thickness) somewhat specifies its durability, but primarily this is a question of personal preferences. Thicker tubes are recommended for larger cases since they visually “fill” the free chassis space.
Once you have chosen the tube size, be extra careful to pick fittings of the correct size. Web-stores explicitly label fittings with tube sizes they fit, so for as long as you pay attention to what you are buying, you should be fine.
Speaking of fittings, I would strongly recommend against buying cheap ones. Water blocks are two acrylic parts sealed together, pumps stop when they fail, reservoirs are just plastic cylinders — all loop nodes are almost perfectly leak-free. If a leak forms in your loop, it will almost certainly be a faulty fitting. Always remember, you get what you pay for, so consider buying high-quality products from well-known manufacturers like Bitspower or Alphacool. However, even the best fittings can cause a leak because of rubber o-rings getting damaged by sharp tubing edges or just wearing down due to tension and coolant chemicals (rotary adapters are especially vulnerable in this sense). For that reason, keep an eye on your reservoir — if you notice a liquid level has dropped, power the PC off and thoroughly inspect all connections. Cases with acrylic or tempered glass panels are of great help since they allow you to watch the loop and spot potential problems on time.
Now, apart from a price, fittings also have different available types: barbed, push-in, quick disconnect couplings… However, in my opinion, compression fittings should always be your default choice, as they typically are the safest due to their tight grip.
The market offers a massive variety of adapters — 45- and 90-degree angled, T- and Y- splitters, extenders, etc. Again, remember that fittings and adapters are the weakest loop parts, so try not to overuse them. However, it might be a good idea to keep a couple of spare angled adapters just in case you bump into a very tricky bend that already cost you quite a few spoiled PETG tubes.
Extenders fall into the same “just in case” category. For instance, I was not planning to use any of these for my very first build, but tightening fittings in a narrow space between top and rear fans turned out to be extremely painful.
Bending rigid tubes
This section is admittedly the most poorly covered in the entire article, and I have a damn fine reason to keep it like that — bending tubes is not a single-time action, but a continuous process. As with every other process, you better learn that by seeing it with your own eyes, rather than reading an article and counting on the power of imagination. I will provide a few essential points regarding this process and share my personal tips, but still, open YouTube and look for it; there are plenty of good tutorials on the web.
- To bend a tube, you need a heat gun; the desired air temperature should be somewhere around 300 degrees. Also, make sure that a heat gun has a massive flat base, so it could firmly sit on your table with a nozzle facing upwards.
- Buy silicone inserts and push them into a tube before bending it. Otherwise, a tubing will collapse when you start heating it up. Silicone inserts are available at most online stores that sell water cooling parts. Choose the insert size that matches your tubing’s inner diameter.
- If you do multiple bends with one tubing, the silicon insert may be hard to remove. To avoid that, lubricate it with a kitchen soap before pushing it into a tube.
- It is difficult to bend a tube at the exact point you initially intend. Try marking this spot with a pencil. Avoid using pens and markers for this task, as their ink can “bake” into the hot tubing material.
- Avoid holding the tubing too close to a heat gun nozzle to prevent bubbles forming on a tubing surface. The same effect occurs when you heat up the pipe for too long without bending it.
- On the other hand, do not hurry and do not try bending a tube until it is hot enough. Otherwise, the tubing may crack.
- When heating up a tubing, move it left and right while simultaneously twisting it. This technique allows you to heat up all tubing sides near the bending spot equally. An exterior angle is subjected to a more significant tension, make sure to heat it up thoroughly.
- Measure twice, cut once. It is pretty hard to eyeball the required tube length, especially when you have little experience, so do not count on your intuition: measure all the distances and test after each bend whether a tubing you are working on is going to fit.
- When all the bends are ready, cut an excessive tubing and grind\polish the edges: sharp tubing edges can damage fittings’ o-rings and cause a leak.
Assembling a Loop
Phew, the hardest part is over, and you finally got all the loop pieces ready, lying in front of you. Time to start building!
Apart from computers, I am also curious about dogs, more specifically, dog training. One of well-known dog trainers whose videos I have been watching back in a day shared a sweet idea: never work with a dog if you are in a hurry. If you go for a walk with a limited time frame in mind, “Okay, we have half an hour to work on my dog’s behavior in a park,” forget it. There could be all kinds of distraction that won’t allow your dog to concentrate. Your dog could be a bit more absent today and have difficulties even with familiar commands you have learned before. That could easily make you feel impatient, anxious; you can start yanking the leash or even shouting at your pet — such training session can eventually bring more harm than not exercising at all. The right way to do that is to clear your mind and simply take a dog for a walk knowing what exactly you want to achieve today. Be prepared to spend 20 minutes standing in one spot if a dog keeps pulling. Be ready to stay for half an hour near the bicycle track if a dog is still nervous when another bike moves close to you. Learning something is not always easy, so you need to have a right mindset and be willing to spend as much time as it requires.
This fantastic advice applies to PC building as well — do not rush things, do not plan to finish everything in one go, be ready to spend more, much more time than you have anticipated. When in a hurry, you are more prone to mistakes, and even if nothing terrible happens, you may still be unpleased with the result. Crooked bends made in a rush have been bugging me for quite a few times, so trust me — if you start feeling tired or angry, take a break.
As for the loop assembly, here are my tips:
- If you are using the same chassis that hosts your current system, remove all the hardware from it — the more working space you have, the better.
- If you are running two or more GPUs in SLI\Crosshair setup, do not secure their rear brackets with thumbscrews after installing them into a case — you will need to push the cards a little to push small tubes in between them. See this Koolance article about different ways to route a tubing through multiple GPUs and make a decision whether you want to connect GPU blocks in parallel, or in series. Both methods have their pros and cons (e.g., see this discussion on the overclock.net forum).
- Put fittings and (optionally) extenders onto radiators before installing them. If you are planning to mount a rad directly to the case with fans on the opposite side, then install the fans first. You can plug all fan headers into a Y-splitter and secure excessive cables by taping them to a rad.
- It is okay to use splitters to connect multiple fans to one motherboard header. Some enthusiast motherboards like Asus Maximus IX Formula feature “high amp” headers (typically, marked as H_AMP on the motherboard) that allow you to connect a whole bunch of fans with a total power draw of 3 Amps. However, regular fan headers have the 1 Amp power limit, so check the fan specs and make sure their combined draw does not top this number as an excessive load may damage the header.
- Install radiators, reservoir, and pump. With these main loop parts in place, you can have a glimpse of what the final build will look like, so take your time and re-plan everything once again. You may want to take a few pictures of your build and use Paint to draw the tubing layout before going for bends.
- Despite a popular myth, the order in which you connect all loop nodes does not matter (except for the reservoir that needs to feed the pump directly). When a system starts a workload, some components heat up more rapidly than others, and that indeed causes a certain temperature diversity. However, a coolant has a very high heat conductivity and it cycles at high speeds across the loop. A D5 pump set to 50% speed circulates a whole volume of coolant in just a few seconds. As a result, the liquid temperature gradually equalizes, we’re talking a couple degrees Celcius difference at max. In the end, you should worry about the ease of access and the aesthetics of the build rather than its nodes’ order.
- Do not overtighten fittings when installing them onto water blocks as a block body, often made of acrylic, may crack. Same applies to compression rings as well — do not use any instruments apart from your fingers and do not employ excessive force.
- Double-check the loop before filling it. Make sure all compression fittings’ rings are tightened, and all spare reservoir and block holes are closed with plugs.
- Most BIOSes are configured to display a warning and\or shut down the system if the CPU cooler is not spinning. With a custom loop, you should plug a 3-pin or 4-pin pump connector to this CPU_FAN header.
- Connect all hardware cables before installing tubes. If you are unable to access the CPU_FAN header after assembling the rest of the loop, plug the 3-pin or 4-pin pump tachometer cable while you still can. Otherwise — that is a perfect scenario — connect it (and an additional Sata\Molex power cable a pump may have) to another PSU: get the cheapest power supply you can get your hands on and use it to power the pump alone. Doing so allows you to safely bleed the loop without powering up the system, so if a loop leaks, it will not damage anything. A PSU does not power up when its main 24-pin cable is disconnected, so you need to jump-start it by connecting its green wire with any of black (ground) ones. To do so, you can use something like a paper clip. If you feel nervous about sticking metal things into a power supply, purchase a bridging plug like this one.
- Slowly fill the reservoir and let the gravity push the liquid into the lowest loop parts. Add more liquid until the res is almost full, then turn on the pump. Keep an eye on the res and a finger on a PSU power switch! When there is almost no coolant left in a res, turn off the pump — remember you cannot run it dry! Refill the reservoir and repeat the entire process until the loop is filled.
- A freshly filled system has a lot of air trapped in water blocks and radiators. These bubbles can airlock the loop and significantly decrease its performance, so you need to get it all out. Close the fill port and start tilting the case sideways. You can even turn the chassis upside down and gently knock all tubes and blocks with a fingernail. As more and more air escapes and gathers in a reservoir, you need to return a case to its normal position, open a res and top it with more liquid. You cannot get all air out at once, give it some time to escape the loop naturally. It may take up to a couple of weeks depending on the loop complexity, but sooner or later all air will be out. You can speed up the process by leaving available reservoir ports open to equalize the pressure, and running a pump at high speeds. Small air bubbles can gather inside the res (see the photo below) — this is completely normal. Tiny bubbles gradually merge into bigger ones and roll upwards, escaping the system.
- After the loop is filled and all the air you could reach is out, leak-test the loop by leaving the pump running for 24–48 hours. Make sure there are no leaks, then unplug the pump and connect it to your primary system PSU. Congratulations, the build is ready, you are awesome!
Fans speed has a direct impact on cooling — the faster they spin, the faster radiator heat dissipates into the atmosphere. At the same time, fans are primary sources of noise. Controlling fan speed allows you to find a sweet spot of harmony, where efficient cooling meets an acoustic comfort.
There are tons of apps that allow you to manage fans. Starting from the one many of you already have — BIOS. Modern motherboard manufacturers include a BIOS section that allows you to build a temperature curve that sets fan speeds depending on CPU temps (or temps of any other on-board sensor).
Same goes for desktop apps — you can download an official software for your board (like AI Suite from Asus), or find a third-party solution. Most of them have one, but very significant flaw — they cannot bind a GPU temperature sensor to speeds of fans connected to motherboard headers. This sucks if your GPU(s) are in a separate loop — controlling its fans and pump speeds based on motherboard temps is ridiculous. However, even with one single loop I personally still prefer to ramp my fans up only when a GPU heats up — because my usual workload affects CPU much less and besides, the processor’s TDP is almost one-third of a GPU’s TDP.
There can be many apps I am not aware of that allow you to get the job done, but I’ll mention the one I use myself — Speedfan. It is not an easiest app to get familiar with, but there are many web articles and YouTube vids that will be able to help you. Also, yes, it seems the app did not change its design since early 2000-s, but believe me, software features easily beat its unpleasant appearance.
In case you give Speedfan a try, be aware that it has troubles auto-starting on newest Windows versions. The solution is a bit weird, but it works — you need to create a new Windows scheduled task that will run the program.