Air Cooling -- Myths and setup tips for the novice performance / gaming builder

[An0maly_76]

I've noticed there seem to be some rather ridiculous and utter myths and misinformation regarding cooling, particularly air cooling. First and foremost, Linus himself did a comparison video testing various air and liquid cooler performance, concluding that while liquid cooling did deliver more consistent temperatures, they were only SLIGHTLY lower – 5C or less, and that air coolers performed quite comparably when properly sized and set up. Linus made a statement in that video – that liquid cooling was cool if you like to spend more time tinkering with your PC as opposed to using it. To me, that pretty much says it all, as a difference of less than 5 degrees in temp is more or less identical.

NOTE: If you’re planning a simple budget build with less than eight cores and entry-level graphics and other hardware, you won’t need a ton of cooling – in fact, stock air coolers should be just fine in most cases. However, if your build is a stepping stone to bigger and better later on, it makes sense to pave the way for this in choosing components with better cooling in mind for any future upgrade. Otherwise, you’re spending more money in the long run.

My 5900X build's cooling system uses an iCUE Commander with six 120mm fans, and the Scythe Mugen 5 which uses a 120mm fan, seven in total, but they don’t even run that hard below 70C. As illustrated in the screenshots below, CPU Package temperature idles in the 38C-42C range. Under light to moderate use, it hovers around 48-58C, sometimes as high as 63C. Cinebench R23 is the stress test to end all stress tests as far as I’m concerned. And even with a 30-minute CB R23 loop, temp peaks at 75-76C. Post-test recovery to idle temp range takes around 45-90 seconds. So it cools quite well, and at just 34 square inches of intake area for cooling, my Corsair 4000X case is hardly the best-ventilated case on the market -- the 4000D Airflow would do even better, although the 4000X has air filters where the D does not.

 

 

My build was hardly a typical process or timeline, but I admit that in the planning stage for my build, I was skeptical of air-cooling and thought the 5900X needed liquid cooling, but others here convinced me that a properly set up air cooler is plenty for even the most robust builds, and more reliable than liquid cooling. However, component selection and initial set-up are a bit more critical. As no one else seems to have done so, I thought I'd explain a few myths and misconceptions, as well as share some insight on component selection and tips for proper setup of an air-cooling system.

MYTH #1 – “My processor requires liquid cooling”

 

This simply is not true in 99.99% of cases. While it is true that more cores and more threads = more heat, my recent R9-5900X build is proof that air-cooling is more than sufficient for consumer market applications. This is not meant to bash those who like liquid cooling, but VERY little, if anything, on the consumer market actually requires liquid cooling. Also, I feel it often is a band-aid for poor component selection and / or an off-the-shelf answer for those who don’t understand how to properly set up air cooling, or the dynamics of cooling.

 

MYTH #2 – “Air coolers are junk. They don’t work”

 

This is not true, either. Air coolers do take a bit more care in component selection and time in initial setup. But properly set up, they are actually more reliable than liquid cooling in most, if not all cases. However, certain factors determine whether a particular cooler is suited to your needs. A poorly-ventilated case can actually insulate the components, making your system a furnace even the best cooling system can’t tame. Conservatively speaking, poorly-ventilated cases hamper cooling by as much as 30-50%, and that’s not the cooler’s fault. For example, you simply cannot stuff a 360mm AIO into an NZXT H510 and expect it to cool. It simply will not be able to get air to do so, because the front is solid steel and your radiator is blocking your airflow. Common sense.

 

Cooling requires three things – heat transfer, ventilation, and airflow. Hotter processors need bigger coolers, because a cooler’s surface area determines heat transfer ability. Too little surface area substantially hinders cooling. Think of it this way – if you were to cut a Mugen 5’s heat sink into three equally sized sections top to bottom and lay them end-to-end, its surface area is quite comparable to a liquid-cooled radiator. Air coolers simply stack their surface area, where liquid radiators spread it out. Some also use two fans, as opposed to one, which can make a difference.

 

Another pair of factors apply to the larger, higher-performance air coolers as well. The impact of heat sink size has already been explained. However, what’s underneath impacts cooling capacity as well, and I don’t mean the processor. First, some have more heat transfer tubes than others. For example, the Scythe Mugen 5 has six heat tubes transferring CPU plate heat to the heat sink. Most, if not all Noctua offerings, such as the NH-U12A, have seven, others can have as few as three or four. And this directly impacts cooling capacity, as to a point, more tubes = better cooling.

 

That being said, it MUST be made for your processor and suited to your setup. My 12-core 5900X cools quite well with the six-tube Mugen 5, so while it may not be an official formula or manufacturer’s recommendation, it makes sense to theorize that air coolers need one tube for each pair of processor cores. And I think this is the biggest factor in misconceptions about air-cooling, that those having issues with air-cooling are simply not selecting the right cooler for their needs.

 

But there’s another factor to look at here, as shown below. This is one area where airflow comes into play. The right-side cooler has six heat transfer tubes, compared to the left-side cooler’s four. In addition, the right-side cooler’s tubes are also perfectly round throughout the cooling circuit. The left-side cooler’s tubes flatten / taper sharply at the CPU plate, reducing their diameter. This is a bottleneck and also impacts cooling capacity. No pipe (or tube in this case) can flow more than its smallest diameter.

 

 

So, as you can see, a PC system’s cooling and performance both come down to component selection. Case selection has a bigger impact on this than you might think. There’s nothing wrong with a case that uses tempered glass / acrylic panels, but be sure it either provides some sort of air gap around the nose or side for air intake, as well as ventilation to the top.

 

System board choice can impact cooling as well. Many don’t know that boards have a voltage regulator module (VRM) that controls voltage to the processor. This effectively controls clock speed, but also has an impact on temperature as well. And cooling the VRM is critical with processors with higher clock speeds and core/thread counts. This is easy to determine. Some boards have no heat sink near the processor socket. Others may have one or two, some have larger ones than others. Bigger is better in this situation, especially with processors known for running hotter.

 

Something to note about fan choice. Quieter fans may not perform as well as others, but you’re not stuck with airplane noise, if you’re willing to put a little work into a custom fan control profile, which I’ll get to shortly. And with a well-designed and well-ventilated case (open grill front such as the Corsair 4000D Airflow), three 120s can flow more than two 140s. My setup has three 120s pulling in at the nose, two exhausting topside, one exhausting to the rear, and one blowing across the cooler heat sink.

 

MYTH #3 – “Air coolers aren’t cost effective”

 

MAJORLY not true. Most budget / entry-level builders can find air-coolers to serve their initial needs for under $30-$50, depending on their CPU choice. Even the best ones are cheaper than a liquid cooling setup. Noctua’s NH-D15 averages $100-$110 USD. The Scythe Mugen 5 cools quite well and at an average of $50-$60 USD, I think it is hands-down the best bang for the buck for CPUs over eight-cores.

 

EASY SETUP OF AN AIR COOLING SYSTEM

 

First, use the proper thermal paste, and properly apply it. Too little won’t cool well and too much will spill onto the board and other components, potentially damaging them. You don’t need a lot, just a thin coating on the processor lid (two or three dots on the lid center are usually sufficient), and take care to wipe any excess squeezing out over the lid edges as the cooler is tightened down.

 

Tuning your case fan control curve is key. Most don’t like the airplane noise of constantly wide-open fans. Good news – it’s not necessary anyway. In a well configured cooling system, the fans generally need not exceed 70-85% capacity, as the upcoming screenshot of my fan curve shows. Also, the more options your fan controller software has, the better this will work. I use iCUE, which has adjustable settings to lock the fans at fixed percentages and RPM, as well as a custom fan curve, such as the one I use, shown below. Note that one size does not necessarily fit all when it comes to fan control curves.

 

 

Step 1: Most fan controllers can be set to a certain RPM and / or percentage of total output. Most fans max out around 900-1400 rpm, some at 1600. With the system idling (only your fan control software running), set all fans wide open for five minutes to get your lowest temperature. I know this will be annoying to listen to initially, but it is necessary to find the coolest temp your system can run. If you’re not seeing low 40s Celsius or better here, something is hindering cooling performance – my 5900X idles at 38C-42C.

 

Step 2: Every 1-2 minutes, dial your fans down 5%, or 100-250 RPM. Keep doing this until the temp starts rising, then dial the fans back up another 5% or 100-250 rpm to regain your lowest temperature. At this point, note your lowest temperature and fan RPM or percentage setting necessary to maintain it.

 

Step 3: If you don’t have it, download and install Cinebench R23.

 

Step 4: Set your fan control to wide open again.

 

Step 5: Set Cinebench R23 options for a 30-minute multi-core test, then start, minimizing the Cinebench window and switching to your fan control.

 

Step 6: Watch your CPU Package temp until it stops rising. If all is well, it should top out around 85% of max safe temp. NOTE: If it rises beyond 90-95% of max safe temperature for longer than 30 seconds, abort Cinebench and check to see that all fans are running and blowing the correct direction. Fans can create a deadlock of airflow when positioned and running against each other.

 

Step 7: Noting your highest stable temperature reading, repeat Step 2, noting the percentage / RPM setting necessary to maintain your recorded stable peak temp. When this has been determined, you can terminate the test.

 

Step 8: Open your fan controller software’s custom fan curve settings. Most will show dots on a graph, marked as fan percentage / RPM vertically, temperature setpoints horizontally. Each dot represents a setpoint.

 

Step 9: If silence is preferred, set your first setpoint at zero percentage or RPM just below or at your lowest temperature. Set your second setpoint by your lowest temperature and corresponding fan percentage / RPM setting. (Example: 0% below 30C, 20% at 30C)

 

Step 10: Find your fourth, fifth or sixth setpoint. Set it to your highest temp and fan percentage / RPM setting.

 

Step 11: The hard part is now over. The other setpoints between your high and low can now be tweaked to find a balance between stable temperatures and noise level.

Edited October 14, 2022 by An0maly_76
Revised, more info

[jaslion]

20 minutes ago, An0maly_76 said:

t makes sense to theorize that air coolers need one tube for each pair of processor cores

My e5 2697V2 on a hyper 212 plus would like a word with you there :p.

 

But nah good writeup and it basically comes down to having enough heatpipes to rapidly take the heat away and then having enough surface area and mass to absorb that heat into before it gets taken away by fans.

[PDifolco]

Pretty good full sized article

 

Yet I'm a watercooling guy, and for example I get 22.5K CBR23 over your 20.3K, with temps around 82C... That's why you go watercooling, so that you can tweak and get big OC while still not throttle your CPU.

 

But I agree, it's for people that like to tinker with their PC over just using it, it's more expensive, more fault prone, needs maintenance... There's a case for GPU watercooling as well, it's super efficient and silent, sure expensive but not that much with $1,500$+ high end GPUs! And while I could think getting back to CPU aircooling just over "practicallity", I'd keep my GPU watercooling anyway.

[An0maly_76]

1 minute ago, jaslion said:

My e5 2697V2 on a hyper 212 plus would like a word with you there :p.

 

But nah good writeup and it basically comes down to having enough heatpipes to rapidly take the heat away and then having enough surface area and mass to absorb that heat into before it gets taken away by fans.

Yeah, I think clock speed and case airflow / ventilation equally impact this as well.

[jaslion]

2 minutes ago, PDifolco said:

Pretty good full sized article

 

Yet I'm a watercooling guy, and for example I get 22.5K CBR23 over your 20.3K, with temps around 82C... That's why you go watercooling, so that you can tweak and get big OC while still not throttle your CPU.

 

But I agree, it's for people that like to tinker with their PC over just using it, it's more expensive, more fault prone, needs maintenance... There's a case for GPU watercooling as well, it's super efficient and silent, sure expensive but not that much with $1,500$+ high end GPUs! And while I could think getting back to CPU aircooling just over "practicallity", I'd keep my GPU watercooling anyway.

Gpu watercooling is quite ok actually. About 200$ tops for a all in one kit from alphacool. Often on sale too. If you get last gen stuff it's often on there for about 100$. Bonus is that you can hook that stuff up to a loop later if you so wish. It's all components that work in a normal loop too. They also offer cpu blocks that just click right into the quick disconnects and boom custom loop.

[Somerandomtechyboi]

55 minutes ago, An0maly_76 said:

“My processor requires liquid cooling”

 

This simply is not true in 99.99% of cases

My heavily abused and overclocked wolfdale and westmere cpus would like to have a word with you

 

Tbf im aiming for closer to 5ghz on platforms that shouldnt even be going over 4.5g on ambient cooling but oh well this 92mm has served me well for my ocing, time to upgrade it and turn it into a watercooler

 

And since im using aquarium and car parts my ghetto loop is atleast on par with air price/performance wise, though im aiming for below 70c at >1.5v which is something no aircooler will ever hope to acheive on a near extreme overclocked westmere

[An0maly_76]

18 minutes ago, Somerandomtechyboi said:

My heavily abused and overclocked wolfdale and westmere cpus would like to have a word with you

 

Tbf im aiming for closer to 5ghz on platforms that shouldnt even be going over 4.5g on ambient cooling but oh well this 92mm has served me well for my ocing, time to upgrade it and turn it into a watercooler

 

And since im using aquarium and car parts my ghetto loop is atleast on par with air price/performance wise, though im aiming for below 70c at >1.5v which is something no aircooler will ever hope to acheive on a near extreme overclocked westmere

It actually seems you prove my point here. Let's review -- "heavily abused and overclocked" -- "aquarium" -- "car parts" (?) -- "1.5v" -- "92mm" which I'd say puts you pretty solidly in the .01% to which that does apply.

[Somerandomtechyboi]

3 minutes ago, An0maly_76 said:

It actually seems you prove my point here. Let's review -- "heavily abused and overclocked" -- "aquarium" -- "car parts" (?) -- "1.5v" -- "92mm" which I'd say puts you pretty solidly in the .01% to which that does apply.

And then theres also the morons that think they can cpu overclock on a new platform but cant cause no aio or aircooler will handle a 126/7/900k at 5.4+ and keep temps low enough to not hit a deadzone where temps are the main limit even if its only 80c, in which case theres not many but alot more than idiots like me that aim for 5ghz on over decade old platforms

[An0maly_76]

7 minutes ago, Somerandomtechyboi said:

And then theres also the morons that think they can cpu overclock on a new platform but cant cause no aio or aircooler will handle a 126/7/900k at 5.4+ and keep temps low enough to not hit a deadzone where temps are the main limit even if its only 80c, in which case theres not many but alot more than idiots like me that aim for 5ghz on over decade old platforms

Remember, YOU said that. Not me.

 

To me, more than 2-5% overclocking is the equivalent of throwing a 300-500 hp shot of nitrous oxide into Grandma's pristine, bone-stock Town Car. It will put up with it for a little while, but eventually disaster will strike. ADDED: However, I'd be lying if I said I hadn't wondered if my 5900X could hit 6 with a liquid nitrogen setup.

Edited April 14, 2022 by An0maly_76
Revised

[Somerandomtechyboi]

2 minutes ago, An0maly_76 said:

Remember, YOU said that. Not me.

 

To me, more than 2-5% overclocking is the equivalent of throwing a 300-500 hp shot of nitrous oxide into Grandma's pristine, bone-stock Town Car. It will put up with it for a little while, but eventually disaster will strike.

Then my overclocks are the equivalent of shoving a massive turbo on a civic and being able to run it daily for lower boost and being able to run for a little while under ludicrous boost

 

And that analogy you made is more geared towards ryzen cause over 1.5v and the cpu just degrades to sht while prob only barely being able to break 5g even at lower temps while intel can run 1.5v without degrading to sht and actually benifit from it by running 5.4+ even though its pointless cause of efficiency reasons

[An0maly_76]

1 minute ago, Somerandomtechyboi said:

Then my overclocks are the equivalent of shoving a massive turbo on a civic and being able to run it daily for lower boost and being able to run for a little while under ludicrous boost

 

And that analogy you made is more geared towards ryzen cause over 1.5v and the cpu just degrades to sht while prob only barely being able to break 5g even at lower temps while intel can run 1.5v without degrading to sht and actually benifit from it by running 5.4+ even though its pointless cause of efficiency reasons

I am, after all, running a 5900X. LOL. Pretty sure when this one was new, stock against stock, the best i9 at the time couldn't hang with it though.

[Somerandomtechyboi]

Just now, An0maly_76 said:

I am, after all, running a 5900X. LOL. Pretty sure when this one was new, stock against stock, the best i9 at the time couldn't hang with it though.

Yep, and that is why cpu overclocking is literally the most pointless thing you can do to a new platform and you are better of gpu and (v)ram overclocking since those still yeild decent amounts of performance relative to power draw

 

Only time when cpu oc is worth it is like me on these older platforms where cpus have tons of headroom and are already able to do 1ghz oc on stock volt

 

If i manage to get cmos fixed ill buy usb headers and sell this p6t off for something better, i might go ryzen considering the prices but knowing me ill prob go and switch to a better x58 board like ud5 or r2x cause haha 12 year old platform at 5ghz go brrrrrrr

[Dogzilla07]

Good write-up just wanted to add that

 

1. Sintered and grooved high quality heatpipes make a nice (arguably small) difference

2. Properly soldered base-plate and heatsink to the heatpipes makes a nice (bigger) difference

 

And that, while having a good, quality fan is arguably the most important part of a cooler, heatsinks can be designed for low and high airflow (dense and sparse distance between the sheets of aluminum). GamersNexus reviews of Deepcool Assassin III and Zalman CNPS20x are example of this, so is Scythe Ninja vs Scythe Mugen 5 for instance. If you got a heatsink like the CNPS20x high speed/high-airflow/pressure fans won't make as much difference as they would on the Assassin III (But they're still close, cause the Zalman is huge).

 

Also dual tower coolers like the NH-D15/S, and similar where the middle fan is sitting quite low and is poking beneath the heatsink make decent airflow for VRM coolers, which is often overlooked. That's why Dual tower coolers are such option, on top of having slimmer finstacks makes it possible to use only 1 fan or at most 2 (the 2nd fan is usually only 2-3c difference.

4 hours ago, An0maly_76 said:

Most budget / entry-level builders can find air-coolers to serve their initial needs for under $30-$50, depending on their CPU choice. Even the best

Even though they're not the best fans (the heatsink itself is very good) Thermalright Peerless Assassin 120 is still unbeatable for it's price of $41. As long as Thermalright continues undercutting the competition (dumping). Unfortunately we in the Europe do not have such a good deal yet  ID-COOLING SE-226-XT Black is also a very competitive option for $45 in black (and beats the black version of the PA 120 in price)

 

 

[An0maly_76]

I actually considered the Assassin, but preferred the subtle look of the Mugen 5 in black... Actually, went black / grey on much of my build. I'm not that big on chrome / aluminum when it comes to stuff like this.

Edited April 14, 2022 by An0maly_76
Revised, more info

[Dogzilla07]

7 minutes ago, An0maly_76 said:

I actually considered the Assassin, but preferred the subtle look of the Mugen 5 in black... I'm not that big on chrome / aluminum when it comes to stuff like this.

Assassin doesn't have the greatest compatibility, its smaller brother the Deepcool AK620 is much better in that regard (and has incredible fans), and beats Mugen 5 and beats/matches Fuma 2 in GamersNexus testing for a similar, slightly higher price than Fuma 2. But yeah it's looks are different from usual coolers.

Though on the other hand, in Europe we have a very competitive price for Arctic Liquid Freezer II 280 at 85 eur ($102 in the US), which outperforms all currently tested Air coolers in GamersNexus testing:

The new Mugen 5S is really awesome, I've recommend a m8 keep his original Mugen 5 when he upgrades to 5900x later  this year, and just with a change to a good fan, similar to your setup 

[An0maly_76]

See, I'd like to see a hybrid cooler... take something like what you just posted....

 

With an inline pump to a top-mount radiator. I never understood why anyone would front-mount a radiator, as that is traditionally the intake and would actually hinder cooling with heated intake air.

Edited April 14, 2022 by An0maly_76
Revised, more info

[Dogzilla07]

6 minutes ago, An0maly_76 said:

I never understood why anyone would front-mount a radiator, as that is traditionally the intake and would actually hinder cooling with heated intake air.

Looks, ease of installation, but yeah from a pure physics, performance and airflow standpoint it doesn't make sense. You don't need cold air for the CPU, it will almost certainly produce less power than the GPU. You just end up putting unnecessary heat into the system for the GPU to deal with it. And there's a bigger area to dissipate so having the CPU radiator on top as exhaust is much better. (The CPU and it's block will care less about using the hot GPU air, they'll be happy with just the airflow, regardless of its heat).

 

Even with seperate GPU and CPU AIOs, for most situations again the CPU as exhaust is more effective.

[GuiltySpark_]

40 minutes ago, An0maly_76 said:

With an inline pump to a top-mount radiator. I never understood why anyone would front-mount a radiator, as that is traditionally the intake and would actually hinder cooling with heated intake air.

In theory, yes, In practice this makes very little difference. 

[Dogzilla07]

1 hour ago, GuiltySpark_ said:

In theory, yes, In practice this makes very little difference. 

Absolutely, there's situations/outliers where it's a problem (it differs widely setup to setup), but for a huge chunk of users it's not that big of a deal (both CPU and GPU will just throttle down a bit, adjust the temp on the fly). Similarly to how the CPU water cooling doesn't care that it's not getting fresh air, so do most GPUs and their coolers not care if their air is not cold (especially low TDP cards and/or cards with big and very good coolers).

 

Which setup did you have in mind @GuiltySpark_?

 

1. for Custom loop = irrelevant

2. 2 AIOs = rarely relevant

3. Liquid Cooled GPU + Air Cooled CPU = slightly relevant

4. Air cooled GPU + Liquid cooled CPU = a bit relevant

 

But with a very important caveat, as long as there's no really bad hot pockets, air vortices. They will screw up temps big time, if the airflow inside the case is not optimal. And well this is assuming there's GamersNexus-certified intake meshfilter, and/or in general not obstructive intake/exhaust, and decent fans and a decent fan setup. As soon as some of those factors become sub-optimal, it's open hunting season for cooling/temperature instability ,and it's usually those other factors which are causing problems in a case.

 

And when you have those kinds of problems, sometimes setting up your radiator in a way to favour the higher TDP part does make actual tangible difference in practice, but for all the wrong reasons xD

 

A great rule of thumb is if your temps drop when you remove the side panel, something is wrong with the airflow in the case (the on-die/in-die temp should be the same, or within margin of error with the case closed up as it is with the case side opened). Of course with a very obstructed case this isn't always possible, and it's not enough to measure/compare the ambient temperature from a sensor far away from the case. To measure the difference in intake temp, as close to the cooling element as possible is paramount for consistency.

 

Another great rule of thumb is comparing the ambient temp outside of the case, with the temp right in front of a cooling element (either CPU cooler or GPU cooler)/fan connected to one. If it's more than 4-7c or so degrees celsius, something's wrong in the case/with the case (Unless it's an aesthetic design choice of course,  that's a perfectly viable reason).

[An0maly_76]

1 hour ago, Dogzilla07 said:

And that, while having a good, quality fan is arguably the most important part of a cooler, heatsinks can be designed for low and high airflow (dense and sparse distance between the sheets of aluminum). GamersNexus reviews of Deepcool Assassin III and Zalman CNPS20x are example of this, so is Scythe Ninja vs Scythe Mugen 5 for instance. If you got a heatsink like the CNPS20x high speed/high-airflow/pressure fans won't make as much difference as they would on the Assassin III (But they're still close, cause the Zalman is huge).

Yeah, so is the Mugen 5. I remember thinking "WTF?!" when I unboxed it.