Sidenote: Observations After One Week of a Custom Loop Watercooled PC

Last week, I wrote in detail about the build and setup of my entire new rig, from the desk to the PC.

With a week of use under my belt now, I can discuss a bit more of what using a watercooled PC is actually like.

Even for someone who reads too much tech news and watches far too many videos of people building custom PCs with watercooling and discussions about it, I had expectations that were slightly out of line. Not badly so, but definitely a bit. Watercooling, for as extreme as it once was, feels like it should be this mythical panacea that solves everything about heat in your system and makes it a non-factor. The truth is that it kind of does, but also kind of doesn’t.

What watercooling offers, and the real TL;DR of this post, is simple – lower temperature levels at typically-lower noise levels. Aircooling in 2021 is getting better, and a massive Noctua air cooler for your CPU coupled with a 3-slot GPU air cooler can get you thermal performance that is pretty good, with the CPU being right near where an average CPU+GPU waterloop will be and the GPU being worse off, but still working well all the same. My goals in undertaking watercooling were simple – give me better temperatures for my CPU and GPU at lower noise levels since I can have fewer fans, run them at 100% as I tend to do anyways, and get a system that is pretty damn quiet to my ears while still offering exceptional performance.

Did I get that? Yes, but it wasn’t without some challenges.

Don’t Expect A Full Loop With CPU and GPU to Make Both Parts Happier

My expectations going in, and the one place I kind of messed up, were that a loop with the amount of radiator space I had would keep both my CPU and GPU happy, cool, and better off than my old system, where the CPU was cooled by a 360mm all-in-one liquid cooler and the RTX 3080 I acquired in December was cooled via its stock, 3-fan, 2.2ish slot cooler. The expectation led to a bit of a disappointment, at least at first – my CPU was running pretty similarly to my old system in nearly all ways – about as hot and about as performant, while my GPU was staying chill with lower temps in all ways – slightly lower idle temps (29 degrees Celsius average versus 34C with the air cooler) and significantly lower load temperatures (hanging out in Elysian Hold in WoW on water was pushing 40C vs. 60C on air). When I thought about it, though, it made sense. CPUs in 2021 (well, since boost became a thing, really) push themselves to an upper bound when there is moderate exertion on them, and Ryzen especially loves doing this. My 3900X will, quite often, boost to mid-4GHz speeds while I’m reading news or writing a blog post because some storage task in the background throttled it with IO requests.

It also made sense to me because the CPU normally doesn’t have to share cooling with anything else, and has a great mechanism for mounting a ton of cooling directly to it. A CPU has a large surface area around the socket designated as keep-out, space that an air cooling heatsink is designed to fill. A CPU can have a massive cooler that weighs as much as a modern graphics card bolted right on top of it, serving solely to cool it. A GPU, on the other hand, has to handle a severe lack of mounting strength. Every gram of weight added to a GPU cooler can cause sag that could, in theory, destroy a motherboard PCIE slot, and there is a pretty sharp limit on how long a card can be before it becomes unwieldy in many systems, or how tall it can be. Most manufacturers have only just now started to breach full 3-slot designs, which require extra bracing metal on the card that serves just to stiffen the whole card for a heavier air cooler! Gigabyte has their Aorus Extreme RTX 3080 that takes up 4 slots, but that isn’t even a great cooler design! Given all of that, I can see how a CPU might not be a fan of going into a full-system loop where a GPU would love it.

And in my particular use case, there are factors I had to keep in mind too. My Ryzen 9 3900X is a first-run CPU, and the first run of the 3000 series ran hotter than later silicon, with idle at 50C not unexpected under fantastic cooling. On the other hand, my RTX 3080 is an MSRP model from notoriously cheap MSI. The heatsink is thinner than most 3080s, split into segments rather than a contiguous block of metal, and has 5 decent-sized heatpipes with 3 smallish fans. It gets the job done, but rather unspectacularly. It also has a “graphene” backplate, which obviously MSI knows is not as thermally conductive as they’ve made it seem. I know that for a few reasons – first, it feels like plastic and is obviously cast like a plastic part, second, it had a metal shim where the VRAM is on the card that it used for heatsinking, and third, the 3090 version of the Ventus lineup has a metal backplate because the 3090 has memory chips on the back of the PCB. So obviously, MSI’s “graphene” is bullshit and with a proper aluminum GPU backplate and the thermal pads I added to sink memory, GPU core, VRM, and even some extra to pull heat out of the PCB and give it an exposed surface to dissipate into air, the graphics card is going to thrive in the new setup.

Plus, after some BIOS tweaks and the newest BIOS for my B550 Taichi from ASRock, the CPU performance came up enough that the new system beats my old 3DMark scores on both CPU and GPU and the average idle temperature is lower. The CPU can still max out at around 75C under load, but the GPU has only ever gotten sort of close to 50C, and that is in Control with all settings maxed and RTX on without DLSS. Stress test, indeed.

Watercooling Makes Tons of New Noises That Will Scare The Shit Out of You

When I first got my system running, there was a trickling sound pretty steadily coming from the front of the case. Oh no, I thought – is there a leak I cannot see?

Nope! Turns out that a vertically-mounted large radiator with fittings at the top can take a few pump cycles to fully fill, and it becomes necessary sometimes to tilt and jostle the system to encourage air to exit the loop and collect in the top of the reservoir, where you can loosen up a stop plug to let it out. If you run your pump fast, you might sometimes here a little gurgle as small air bubbles pull through it, and you might have similar weirdness as those air bubbles enter parts like waterblocks and cause turbulence or weird bits of disturbance in the loop. Then, your platter hard drives turn on and the sound of their motors, which you haven’t heard much yet in the new system, makes you think the pump is doing something dumb, and oh no!

Which leads nicely to the next point, actually…

First Time Watercooling Makes You A Paranoid Lunatic

An O-ring only needs to make compression against the terminal on a component to seal. It doesn’t have to be flattened into oblivion, or so compressed that you cannot see it – just has to mate with the fitting surface and the terminal surface, and that’s all. I start with this, because the other day, I legitimately opened my case and tried really hard to tighten a fitting further because I could see the O-ring, despite no leakage or even any signs of potential leakage, like moisture outside the waterblock or any sort of pooling around the fitting, pressure drop, or air bubbles entering the loop in the tubing near the offending O-ring. After common logic dictated I shouldn’t worry about it, I dropped it, but then I took a shower and the whole time I thought about how I might return to my PC with it awash in purple drink. That (obviously) didn’t happen!

This weekend, when updating the BIOS on my motherboard, I noticed after the update that my liquid temperatures were climbing even at idle, and while the CPU and GPU temps were about where they had been pre-update, it scared me. Not logically, though! I started jostling the case for air bubbles, tightening fittings in reach, looking frantically around the case to make sure nothing had gone wrong. And nothing had! You know what did happen, though? Updating the BIOS cleared my settings and I had not gone back in to set my fan speeds (including the PWM signal for my pump) so the liquid was slower-flowing, and the fans were not pushing heat out at the same pace as before the update. A quick reboot and fix there and my liquid temperatures scaled like normal once more – all the tilting, shaking, and idiotic grabbing and twisting of fittings had been for nothing.

Also, I have the return run from the loop back into the pump set to the anti-turbulence tube at the top of the reservoir. Now, this is fine and works, but it has a quirk that drives me crazy even still – when you look at it in operation, nothing seems to be flowing through the exposed length of the tube! Now, obviously, something is – I can see fluid in the tube routed into that fitting, and the loop isn’t draining or losing fluid anywhere, so it is obviously being replenished as it should be – but I have Googled if that is how it should look about 5 times now.

Now, this is at least partially on me, because I love tinkering with that kind of stuff and I was paying too much attention, such that I was starting to find problems that weren’t there (even the faster onset of liquid temperatures wasn’t an issue) – but I think that anyone installing their first watercooling loop where it has the potential to wipe out a graphics card that took 4 months to get might become similarly obsessed!

Watercooling Teaches You A Lot More About Science

Measuring a PC’s thermal performance by component operating temperatures is fine, but also flawed. The real measure is delta T over ambient temperature, defining how much hotter than the environment the PC is in that the components get. In watercooling, there’s a further twist – the temperature you most want to watch is the temperature of the liquid in the loop, provided that your components are not overheating and melting down, that is. In my new case, I have a thermal probe that was provided with the Corsair XD5 pump/reservoir combo, and I’ve wired that to an XSPC 7-segment display screen that shows me the liquid temp at all times the system is on. What you really want is for that liquid temperature to be lower – specifically, watercooling threads and posts I read suggest a holy grail of 5C over your room ambient, with most people realistically being able to get somewhere between 10-20C delta. Of course, I live in the US where room temps are reported in Farenheit, so I’ve spent far too much of the last week converting the thermostat readout and checking it against the liquid temps to determine how well my loop is holding up (pretty great, average gaming delta T is between 8-12C), which was mostly because of the paranoia I mentioned above!

In the end, what I learned is that a lot of nerds obsess over the liquid temperature too, but that ultimately, as long as your liquid isn’t meeting the temperature at which the tubing starts to degrade (usually around 60C) then it is okay, provided your parts aren’t throttling or getting excessively hot themselves, and even most poorly designed watercooling loops will still keep things well below that tube-melting temperature. Given that, my fuss over the liquid temps reaching 35C after hours of gameplay was not warranted, but again – the paranoia!

Watercooling Gives You A Quiet System, Provided You Tune It

Air cooling being quiet can be done, but it takes massive air cooling towers, large, ugly Noctua fans, and being willing to sometimes still reach temperatures where the boost algorithm of a Ryzen CPU will step back clockspeed (or an Intel CPU will happily brr right along with their dumb boost), not to mention getting stuck with a GPU that has a massive footprint and requires a lot of massaging to get to run well.

Watercooling, on the other hand, starts you off on a much better footing. While what I mentioned above with the CPU is still true, you can get a watercooled system to run quieter than most aircooled systems with relative ease. There’s an easy mode version (buying an Aquacomputer fan/pump controller that is rigged to liquid temperatures and curves your fans for you based on that reading) or you can do what I did, which is still easy mode – crank the fans and pump through your BIOS to the point of your noise ceiling and let it be from there. You can set curves based on liquid temperature with some motherboards too (if they have the two-pin header for a thermal probe, which my Taichi sadly does not) or you can tune based on a component temp for your application – in all likelihood the GPU for gaming and the CPU for productivity-focused machines. For me, the choice was easy – even with 6 fans and a PWM pump at 100%, this system is still far quieter than my previous rig, which had 12 case fans, 3 GPU fans, an AIO pump on the CPU, and in order to tame the heat, required most of those things be running at 100% (my GPU fans maxed at 92% of their speed because higher pushed a high-pitched whine from them which even made me cringe).

Generally, because of how liquid absorbs heat and how the whole of a watercooling loop functions, you can get lower fan speeds to perform better because you have a much longer window for heat soaking before the system hits an equilibrium, while in air cooling, a heatsink gets to its maximum heat load pretty damn fast and you need fans to whisk that heat out quickly. Because you have fewer fans (generally) in a watercooled system, you can run those fans faster to improve performance without as severe of a noise impact. A few minutes spent adjusting speeds and tweaking things to your tolerance can get you very far.

For me personally? I’ve always been a 100%er on fan speeds because I have a lot of noise tolerance, and I find the adjustment of speed on fans to invite unwelcome pitchy sounds as I can hear the adjustment and the changes in noise more than a static, steady 100% fan speed, so I tend to set a static speed and leave it. In this case, I have half the fans of my last system, and a far more efficient and quiet pump compared to an AIO, so 100% everything is quieter just by nature of being less stuff moving at full speed!

Watercooling Is Gorgeous, Usually

Watercooling is aesthetically pleasing in nearly all cases, and with a small amount of work, it can look even better. My soft-tubing runs aren’t the most amazing or aesthetically-minded, but the criss-cross of clear tubing full of clear purple fluid being lit by purple and blue LEDs looks great to me. Certainly, aircooling can also have a sort of fun industrial feeling, but I really like the subdued look of my watercooling parts, especially because the subtle accents they do have look better for it.

Watercooling Is Expensive, But You Control The Scope

I spent a lot of money on just the components of my watercooling loop. Not counting fans, fan cables, or the RGB hub I bought to make its LEDs sync up, here’s a basic run down:

CPU Block: $50
GPU Block: $180
1 Each 120mm, 240mm, 360mm Radiator: $185
Tubing: $25
Purple Coolant: $18
16 Compression Fittings: $112
Four 90 Degree Adapter Fittings: $54
Ball Valve for Drain: $20
Pump/Res Combo Unit: $155
Liquid Temp Display: $15
Squeeze Bottle for Filling: $3
Total: $817

Most people can buy a full system for that amount, assuming non-COVID, non-supply constrained times. Yikes! However, this is all down to choices you can make in the design. A drain valve is optional – it is supposed to make life easier, but you can drain a loop without one just as easily, and that would cut down on fittings and the ball valve. You can do just a CPU loop or just a GPU loop and remove the waterblock cost for whichever piece isn’t going on water, plus reducing fittings, plus reducing radiator expense. Most recommendations are a single 120mm fan of radiator capacity per component on the loop, plus an extra 120mm worth if overclocking (I’ve heard this advice given as one 120mm per overclocked part or even just 1 extra fan worth of capacity in general if overclocking), so while I have six 120mm fans worth of capacity in my loop, I could have gone down to between 2-4 and still reached pretty great performance. Stuff like the 90 degree fittings and colored coolant are all aesthetic/ease of use focused, because you can buy a gallon of distilled water at the grocery store here in the states for under a dollar and then just need to source an anti-microbial/anti-corrosion agent to add to keep your loop free of nasties, and likewise with the fittings, you can use standard fittings and just be more careful and smart about routing your tubing.

One thing I also want to highlight here is that while I painted myself as maybe a bit carefree with my spending, by picking the Corsair ecosystem, I actually saved a lot of money. The CPU block at $50 was a steal compared to an entry level EKWB block that starts around $90. Provided I could even get a GPU block for my card from EKWB, that block would have cost as much, but then the backplate is sold separately and would have cost another $50-$65 on top of the cost of the actual block. Then the pump/res combo would have been around $30-$50 more, and EKWB’s radiators are all more expensive by almost double, and they perform worse in benchmarks! When I thought I was going to buy EKWB stuff, the cost of the watercooling alone was easily over $1,000! Now, the thing with that is EKWB does, generally, have higher-quality stuff (radiators notwithstanding) and they are a good brand – but you are paying a decent chunk of change to just have their little circle logo on your parts, while Corsair makes simple stuff with a unified aesthetic that performs nearly as well, and where it is lower performance, it isn’t by enough to make it bad.

The one buying advice thing I will offer is this – where Corsair tries to get you is in the non-watercooling part of the purchase. If you use their configurator tool it will sell you their LL120 fans at $45 a pop (I used the 3 included front case fans on my 360mm radiator, and bought 3 additional Cooler Master SickleFlow fans at $13 a pop instead) and will tell you that you must use a Commander Pro to control fan speeds, connect the temperature probe from the pump/res, and control their RGB LEDs. Wrong, wrong, wrong. Buy the fans you want and use your motherboard connectors or a fan hub that isn’t $70 fucking dollars. Use a motherboard temp sensor header or do as I did and buy a display readout. Lastly, the new Hydro X stuff from Corsair all includes adapter cables from their proprietary RGB connector to a standard 3-pin 5V motherboard header, and you can daisy-chain them by using a single such connector on the pump/res, then connecting everything else to that. Even with the RGB hub I bought to add more headers to my system, I still spent less than Corsair wanted on just the Commander Pro, and between that and the fans, I saved around $250 just on these additional components!

So in short for this point – watercooling is expensive, prohibitively so in some cases, but if you’re interested in it, you can make a few smart plays to bring down the total cost or can pick a limited scope to get the rush of building a watercooling loop without spending a fortune on it.

In Closing: Watercooling Is Cool, Worthwhile, But Not For Everyone

Overall, I share these notes of my experience because I think it is important to have a newbie to the form discuss the pros and cons they ran into. Most watercooling forums I visited, videos I watched, and threads I read approached the topic with a certain level of assumed knowledge. I’ve been building and tinkering with PCs for nearly two decades, and the experience has been full of ups and downs, lots of little scares, and some difficulty in neatly getting answers for my own concerns with my loop. However, I am glad I took the plunge on it – I feel like I’ve learned a lot of very valuable lessons for my PC building hobby, my new system looks great and runs fantastically well (I still cannot wait to put a Ryzen 9 5950x in there to see what it does!), and it runs quietly and reliably with far less concern over temperatures and the heat wave my old PC would create at my desk when gaming (paranoia mentioned above aside).

Here’s the caveat, though – is it something I would recommend? Well…yes, but also no.

My yes is for these reasons – it is a fun, worthwhile endeavor to undertake, that will teach you a lot of things about efficiency, how your computer runs, and how to maximize cooling in it. You can exert a lot more influence and be rewarded for doing things well with a cooler-running, amazing looking enthusiast PC. I feel like my goal of making this system last 5 years is quite doable in this state, as it has so much to offer and already feels so much more alive and spritely than my old system, despite the core components literally being pulled from it! For the cost, it was a fun enthusiast project, and for all concerns about leakage or other disasters, if you pay close attention, watch a few guides, and properly hand-tighten your fittings and leak-test the full system, there isn’t that much risk – although you should treat it carefully due to the risks it does have!

My no, on the other hand, is simple – for what it offers, watercooling remains a niche interest that has a lot of costs, a lot of hidden costs (fittings add up very fast!), and at the end of the day, offers a marginally better PC experience for all of the cost and stress of the risks involved. If you value silent PCs, watercooling isn’t the only way to get there – it’s just the easiest. If you value fast PCs, you can make a lot of upgrades in your part list before watercooling becomes anywhere near the best way to add performance. If you like good-looking PCs, you can pull that off a lot of different ways without spending almost a thousand dollars and some gray hairs on building a custom loop. RGB is easy to add to nearly any system, things like better cases and more premium components can add a style without pushing too much added cost, and most of these things don’t require bringing water to your famously water-opposed PC components!

What I do like and why I feel like this is worth discussing is this – watercooling has come a long way from something people did with old car radiator cores and clamped tubing, and brands like Bitspower and Corsair are bringing more low-cost options out, such that even industry giant EKWB is starting to work on building better low cost options and kits while avoiding some of the low-cost watercooling problems, like using aluminum components (mixing metals is a huge no-no for a custom loop, and most stand-alone watercooling parts are copper, which, in a loop with aluminum, can cause galvanic corrosion and gunk things up badly!). For my experience, I can only speak to Corsair, but their ecosystem approach made buying my parts and hooking everything up very simple, and while I love to tinker, I’m glad that their components are all made simply and designed to simply be inserted into a new loop, ready to use. In the future, I will probably want to build another loop and would go with something from EKWB for the sake of getting the experience (and those slight upgrades in quality and performance), or at least more of a mix-and-match from vendors. But in the end, I was happy that my experience was pretty easy, with all the parts I needed available at a single vendor, and that despite my nerves and paranoia, I was able to set everything up and get it working trouble-free so that I could finally say I have a watercooled system.


2 thoughts on “Sidenote: Observations After One Week of a Custom Loop Watercooled PC

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

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