Understanding the Effects of Thermal Limits
From the outset, Arctic Flow was destined to be liquid cooled. But despite what you may think, I have absolutely nothing against air-cooling a PC. For decades I explicitly used beefy CPU coolers and plenty of fans within my builds, as they offered excellent performance without usually breaking the bank. I have briefly discussed air-cooling previously here, as well as a brief comparison of the advantages and disadvantages of ‘Air vs Liquid’ here.
I am also a big fan of the pre-built AIO (All-in-One) coolers that all of the usual suspects offer, having used 240mm and 280mm variants myself in previous builds. They offer outstanding cooling performance and are relatively straightforward to install.
So with that said, why would I be so adamant about building a fully liquid-cooled PC?
Whilst my reasoning at the onset of Project Arctic Flow was mostly theoretical (and based off the results of others), it was my own experiences with my hardware that reinforced the need for it.
On paper, the RX Vega 56 from AMD does not look that impressive compared to other Graphics Cards that are currently on the market; especially when you consider the newly released Navi and Super cards from AMD and Nvidia respectively. The Vega cards were ridiculously overpriced initially due to the ‘mining craze’, which caused hardware prices to become heavily inflated due to demand. Depending on the game, performance was usually on par with an Nvidia GTX 1070; but acutal performance could vary depending on the circumstances.
In reality, the Vega 56 has the potential to outperform an Nvidia GTX 1080 as well as the RTX 2070; but not in it’s ‘out of the box’ configuration. It requires some tweaking, cooling, and potentially even some modifications to do so; as I found out firsthand.
As I mentioned last time, I was able to purchase the Asus ROG Strix Vega 56 for an absolute bargain. The card comes with a beefy triple-fan setup as pictured above, as well as being aesthetically pleasing with a nice backplate and RGB accents. But the performance of the card is severely limited due to some poor choices in regards to the cooling of components; specifically the VRMs (Voltage Regulator Module).
VRMs are essential on motherboards as well as on graphics cards, as they convert the +5V or +12V supply voltage to the much lower (0.5V – 3.5V ) levels that the CPU and GPU require. These poor little fellas are kept considerably busy, as regulating the voltage to these components is one of the most vital functions within the PC environment; if voltage fluctuates too much, parts can quite literally fry.
On a motherboard the VRMs are quite easy to identify, there is usually a large heatsink covering them in the general area around the CPU socket on the motherboard.
On a graphics card however, some dissasembly is required in order to see where they are.
It is fair to say that the ‘typical PC user’ does not regularly monitor the temperature of their components, but ‘PC enthusiasts’ certainly do. If you find yourself using your PC for intensive tasks such as streaming/video editing/gaming, then monitoring your temperatures is a must. There are a plethora of monitoring options out there, which can provide you with the important items to keep an eye on: CPU Temp, GPU Temp, Fan Speeds, etc.
Monitoring these values can alert you to any potential problems with your components as well informing you that you have airflow problems within your case.
For Arctic Flow, the CPU and GPU temperatures looked great on air-cooling. My CPU would run in the 40s – 50s whilst the GPU usually hovered in the 60s under full load with a relatively slow fan speed (as pictured above).
But that single value does not tell the full story, as other temperatures were quite alarming.
The VRMs on the GPU were getting hot, REALLY HOT, as circled above on the left. In fact I have no idea how hot they could have potentially reached after prolonged use, and certainly did not want to risk finding out. As it turns out, the VRMs (circled on the top right) had horrible thermal pads which were too thin to effectively transfer heat to the heatsink (circled bottom right). So whilst the GPU temp was fine, the card was actually throttling due to the VRMs getting too hot.
Even so the card was destined to be liquid-cooled, something had to be done in the interim; as I had games that needed to be played!
It was a fairly easy fix to be perfectly honest, but not necessarily something a complete novice would want to attempt. I purchased some thicker thermal pads and also took the opportunity to apply some aftermarket thermal paste to the die. As demonstrated in the images above, the VRM temperature lowered drastically and as a result the card performed much better afterwards.
This was also a good test-run for me; as whilst I am confident in my abilities to dismantle and reassemble tech components, it is still nerve-wracking to disassemble relatively expensive graphics cards. But it was something I would have to do once again in order to apply the water block.
Running Cool, But Loud
The GAMMAXX GT CPU cooler from DeepCool is an excellent air cooling solution; it runs cools and quiet, looks the part, and also is relatively inexpensive compared to other brands. My GPU on the other hand, even so it was now running cool and peforming well, was considerably loud. Probably even more noticeable due to the open design of the Thermaltake Core P5 that Arctic Flow uses. At idle the PC was fairly quiet, even during video rendering under full CPU load it stayed quiet as well; but when gaming the GPU fans ramped up (out of necessity) a considerable amount.
Since I use a headset when at my PC, it did not bother me too much; but needless to say, the rest of my household could easily tell when I was playing a game on my PC.
For many, this is the primary motivation to construct a custom loop for a PC; as it is possible to have a PC that is practically silent under full load. Liquid cooling allows the user to fine-tune their PC by manipulating the pump and fans speeds to almost inaudible levels. Likewise if noise is not a factor, then the user can aim for maximum cooling performance; depending on the radiator fans used, a setting of 100% could potentially still be quieter than those on an air-cooled CPU or GPU.
That flexibility, as well as the aesthetic benefit of a custom loop were my primary motivations for Arctic Flow.
It was also one of my ‘bucket list’ items to be perfectly honest, a challenging yet rewarding endeavour that I would be hopefully be proud of afterwards.
This is where the fun begins.
Planning and Parts Used
The most important aspect of undertaking a custom liquid cooling loop is the planning stage. It has always been a personal mantra of mine in my professional life:
80% Preparation, 20% Execution: 100% Success
Which definitely holds true when it comes to putting together a PC, even more so when the intricacies of liquid cooling are added to the mix. I had originally planned on using some of the more well-known brands for some of the liquid cooling components, but ultimately the prices were cost-prohibitive; so instead opted for fittings, blocks, fans, and pump from prominent brands which are commonly used in Asia. These are of course names such as: Barrow, Bykski, and Barrowch.
These brands are slowly gaining recognition in the West, as their quality is surprisingly high considering the extremely low price. For comparison, I was able to buy everything I needed for about the same price as what a single block and a pump would have cost from the Western brands. The downside was of course having to wait several weeks for the components to be posted from China.
The only exception was the radiator I opted to use. Ironically enough, raditor reviews/benchmarks are some of the most dry things to peruse; but not all radiators are created equal. You would actually be amazed at just how poorly some of the radiators from ‘big name’ brands perform. But at the same time, I was not willing to ‘roll the dice’ and experiment with a radiator from Barrow; so opted to go with the infamous Black Ice® Nemesis® 360GTS® from Hardware Labs.
Not an inexpensive piece of kit when compared to the prices from other brands, but frankly it is one of the most important aspects of any liquid cooled system; so it was money well spent.
The first order of business was to basically dissassemble Arctic Flow and prepare the CPU and GPU for the water blocks. Whilst I know a lot of people record time-lapses of the build process, since this was my first time attempting a custom loop I wanted to stay focussed on the task at hand whilst also taking my time; so only pictures I am afraid.
I was pleasantly surprised with the build quality of the water blocks; now this is not to say I was expecting them to be horrible, it is just that given their low cost I had fairly realistic expectations. Even so they felt and looked the part, I ensured to check that all of the screws holding them together were sufficiently torqued for my own peace of mind.
The next order of business was to mount the pump, which was the only component I was slightly trepidatious about.
The pump is incredibly important because if it were to fail, then the entire PC would be in jeopardy; if the fluid is not flowing, then the system will eventually overheat. I spent months researching all of the various brands in regards to reliabilty and longevity, ultimately coming to the conclusion that a pump from Barrowch would be my best option for a couple of reasons.
- The colour from the addressable LEDs would match with the blocks as well as the distrubtion panel.
- Due to the extremely low price, a replacement could be purchased if something were to happen. Even buying a new pump every year would still work out cheaper than servicing a more expensive pump every other year.
This reasoning was also why I opted to go with the water blocks as well, as some of the more expensive options are plagued by issues; so rather than bother with the lengthy RMA process, inexpensive replacements seemed as a more reasonable alternative if anything were to malfunction.
As you can see above, I opted to utilise an elbow and extender fittings to provide the pump with a little extra rigidity; even so it is also mounted via screws into the bottom mounting plate. What I did not expect was the difficulties I would have with the small run from the pump to the distribution plate above it. The bend was too tight to use hard tubing, but also too short to use flexible tubing without kinking it. Rotating the pump was also not an option, as the inlet fittings would not have been tight enough; so a slow leak would have formed.
A simple 90 degree fitting solved the problem, but not until a week later; as I did not have one on hand at the time. I happened to tweet about this very issue earlier this week. The interim fix was an eyesore to say the least, but I desperately wanted to get the system assembled and tested.
I also took this opportunity to somewhat clean up the cabling within the case’s back panel, but did not take it to the extreme as I still wanted relatively easy access if I needed to change things out. The important thing was ensuring the radiator was unobstructed and that everything could connect correctly.
Speaking of connections, I have to rant and rave about this 16-way controller from Barrow. Fan controllers as well as RGB controllers are fairly expensive items, RGB controllers especially. One of the most infuriating things as well is compatibility, as many brands of fans or strips will only work with their own brand of controller. Even so my motherboard has the ability to control all of these items natively, this particular controller was required since I opted to go with fans from Barrow (not only for price/performance, but again the whole colour-matching aspect); which have a proprietary 6-pin interface (combined PWM and RGB cable).
Initially I thought this was a mild inconvenience, but since the controller was fairly inexpensive I just accepted that it would be needed.
Let me just say, this cheap little controller board is absolutely fantastic and an absolute must-buy for anybody using multiple fans or RGB components; regardless of brand.
This controller is multi-voltage, so you can plug in 12V RGB or 5V ARGB (Addressable) components by using different pins, 4-pin or 3-pin fans, flow sensors, etc. Furthermore, it comes with a connector cable so the controller hooks up to your motherboard so that everything is controlled natively from your own software (in my case, the Asus software). If however you do not have a motherboard with RGB support, then you just flick a switch (hidden by my cable runs above) and use the included wireless remote to control the lighting of your components that way.
Absoloutely brilliant, especially for $13 USD!
After building the foundation and ensuring all of the RGB components and fans worked (these can be tested without having to power on the PC), that was enough work for one day. Bending the PETG tubing for the first time would begin first thing the following morning.
They Make it Look So Easy
I have been watching videos on YouTube about liquid cooling for years, so in theory the process of heating up hardline tubing and bending it seemed fairly simple. If you have watched a video on anything and thought to yourself ‘yeah, I could do that’ then you know what comes next; it is very rarely just that simple. These people have years of experience to the point that these processes are second-nature to them, with that said, everything went quite well considering this was my first time giving it a try.
In all I only had 2 instances of a bend not quite going to plan with Arctic Flow, but I was only off by very slight margins. The best advice I can give is to take your time and first practice with the whole heating up process and how the properties of the tubing change. It takes a little while to get a feel for it, but once you do it becomes fairly straightforward.
Even so I had planned my loop extensively, I overlooked the fact that technically my distribution block was designed for Intel motherboards. What this meant was that the tubes running to the CPU would be perfectly parallel with an Intel system, but with AMD there was just a slight difference in vertical positioning of the CPU. So my initial plan to have all of the horizontal runs lining up parallel to each other was out of the window.
So I decided to get a little creative and instead ran the lines at slightly different depths to give it a little character.
Whilst the shot above only slightly evidences what I mean, and when dry it admittedly does not do it full justice; the effect when the system is running under my desk definitely gives it a sense of depth, feeling more organic than if they had been all lined up. Before you start thinking that I am just using that as an excuse for any mistakes made during the bending process, let me point out that trimming the tubes to the same lengths after bending is actually extremely easy.
Trust me, in person it just works.
Once all of the bends had been finished and all of the fittings triple-checked for tightness, the system was ready for the initial flush and leak testing. But Arctic Flow would have to wait, as I was weary and needed a stiff drink as well as some ‘chill time’ before proceeding any further.
Arctic Flow Begins to Flow
It goes without saying that testing Arctic Flow for leaks was of the upmost importance. I had already cleaned and rinsed out the radiator, as that is best-practice (due to debris/flux from the manufacturing process) as well as made sure that the blocks/fittings/tubes were all clean as well. As you can see in the image above, paper towels are placed underneath various components as a precaution; but as long as you leak test correctly, you will not actually damage anything.
What I mean by this is that as long as there is no power going to any of your components, water will not damage them. This is where that faulty PSU from Aerocool came in handy (yeah, still never heard back from that RMA request), as it allowed me to power up only the pump. If you do not have a secondary PSU, then just make sure to not connect your PSU to any component other than your pump and use a ‘jumper cable/cap‘ to enable the PSU to turn on.
With 5 liters of Distilled Water on hand, the plan was to fill the system up and check for leaks over an extended period of time (24 – 48 hours). If there were no leaks, then I would drain and refill Arctic Flow a few times just in case there were any ‘floaty bits’.
I was extremely happy that everything checked out and there were not any leaks. But sure enough, I had a few small bits of debris accumulating within my GPU block. Since my loop order is Distribution Plate (DP) > Pump > DP > GPU > DP > Radiator > DP > CPU > DP, this likely meant there was little bit of crud within the pump; which was indeed the one part that I had not previously flushed and cleaned. Whilst undoubtedly an inconvenience, I fully expected that I would need to do this; at least the GPU block had conveniently captured all of the bits for me.
So after a quick drain and partial disassembly of Arctic Flow, the crud was cleaned out and then reassembled; since now I would have to leak test once again since components had been disconnected.
After another day of leak testing, everything checked out and I was ready to fill Arctic Flow with the actual coolant that I had chosen. To be clear, there is nothing wrong with using Distilled Water; in fact it is probably one of the best solutions to utilise for a ‘first loop.’ Just keep in mind that a mild biocide and corrosion inhibitor will need to be added, as Distilled Water alone is not enough. For Arctic Flow however, I opted for a solid pre-mixed solution; not only for aesthetic reasons, but also since it contained all of the necessary additives.
This would also be the first time that Arctic Flow would also be fully powered on since constructing the loop throughout the week; so whilst I did not think I had damaged anything during the process, there was no guarantee that everything was in working order.
Running Cooler AND Better
This stage of the process is the tedious part to be perfectly honest. In a simple loop with a typical reservoir, it is best practice to let the system run for a decent amount of time in order for any trapped air to escape. Since Arctic Flow had several bends and the Distribution Panel, this process would take much longer than a simple loop would. By attaching a piece of soft tubing as pictured above and running the pump at 100%, this was the best way for me to let the system reach equilibrium. I would occasionally power down the system and rotate the entire PC in various directions, so air bubbles could get unstuck, then boot Arctic Flow back up so the air could escape.
The large air bubbles did not take too long to make their way out, but the ‘micro bubbles’ took a fair bit of time. If you look closely at that bit of soft tubing above, that is actually foam and not fluid escaping from the tube.
After several days of getting the system bled and ensuring everything was working, it was time to see how everything was working.
At idle the system is near silent and of course the temperatures are nice and cool. For reference, the ambient temperature of my closet under the stairs is around 22 degrees Celsius; so a few degrees above ambient is excellent. The VRM of the motherboard is air-cooled as is the Chipset, so those temperatures are to be expected; but they are well within safe operating temperatures.
In reality, idle temperatures are relatively meaningless as nothing is actually doing any work, therefore not generating much heat. But idle temps do at least allow you to identify if there any issues before really putting your hardware to the test.
So how do the inexpensive liquid cooling components perform whilst gaming you ask?
One of the most demanding titles I currently own is The Division 2, plus I do play a fair bit of it; so it made perfect sense to give that a try and see.
Whilst I clearly expected temperatures to be lower, I honestly did not think that the drop in GPU temperature would be that dramatic; 37 degrees after a couple of hours of gaming is just amazing. Whilst not pictured here, the ‘Hot Spot’ and VRM (of the GPU) temperatures are only a few degrees higher than the die itself. What this static image also does not capture is the difference in GPU performance after switching to liquid cooling; with the thermal limitations removed, the core speed of the Vega 56 now regularly sits in the 1580 – 1630 range.
That is without altering anything or attempting to overlock it mind you, the GPU just boosts automatically to that higher speed. Whilst this has resulted in noticeable FPS gains, I was pleasantly surprised at how it stands up against other cards.
Whilst there are a variety of ways to test for performance, a really nice quick and easy way for anybody to give their system a check is to use UserBenchmark. Whilst applications such as 3DMark and Cinebench are of course what most of the tech enthusiasts use to benchmark their hardware, UserBenchmark is a great way for everyday users to see how their system fairs. It can also highlight if something is not running the way it should in a manner that is fairly easy to read.
For a while I was unsure whether my decision to purchase the Vega 56 was the right choice or not, as I knew all too well that new cards were on the horizon. But since I was able to snag it for a bargain, I figured it was worth the punt.
Needless to say, I am happy with the decision; at 125% it is outperforming a GTX 1080 and Vega 64 whilst being within range of the new RX 5700 and RTX 2070. Once overclocked a little bit, it will likely perform slightly better and for much cheaper (that is including the GPU block).
Whilst a newer and faster Graphics Card is a nice idea, Arctic Flow is perfectly balanced at the moment. The liquid cooled Vega 56 is a perfect match for my 3440 x 1440 panel at 100Hz right now in all the games I play, but could very well fall behind in another few years time. I will cross that bridge when it comes, but anything more powerful would just be a waste due to the display I am using.
The CPU on the other hand is a different story. I said from the get-go that the Ryzen 5 2600 was just an interim solution, not that there is anything wrong with it. I have not even got around to experimenting with overclocking that yet either, because frankly it peforms just fine for gaming (boosting to about 3.9Ghz) as it stands. But since Arctic Flow is also going to be used for my Honours Project at University, which will involve heavy use of Virtual Machines; I need a CPU with more cores. So the newly released Ryzen 9 3900X with 12 cores and 24 threads will be perfect for that (hence why I chose 32GB of RAM, as it is certainly not needed for gaming); plus it will also give me a little boost to my gaming performance as well.
So Is Arctic Flow Done or Not?
Kind of yes, maybe.
For the most part everything is done except for the shiny new CPU, but that will not be getting purchased/installed until September. Beyond that there is nothing planned aside from enjoying the performance and silence that Arctic Flow now provides. I am still experimenting with pump speeds and fan curves, linking fan speed to liquid temperature whilst linking pump speed to CPU temperature seems to be working well so far; but everything else is sorted.
If a fantastic GPU appears in the next year that vastly outperforms what I currently have AND is inexpensive (the current options would not be worth the cost) AND has a compatible water block available, then perhaps I will consider it.
In the meantime, I will game when I can before delving into the final year of my BSc (Hons) in Cyber Security and Networks.
Oh, and of course an obligatory video of Arctic Flow below:
This video has been identified to potentially trigger seizures for people with photosenstive epilepsy.
Viewer discretion is advised.