CPU Cooling: Air vs Liquid

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Posts: 1
Joined: Sat Apr 25, 2020 10:31 am

CPU Cooling: Air vs Liquid

Post by marvelsferb7 » Sat Apr 25, 2020 7:08 pm

I'm looking to build a top-of-the-line PC soon and plan to OC as well (debating between a 9700k and 9900k depending on benchmarks). Regardless of a specific build though, what's everyone's thoughts on liquid vs air for an enthusiast?

Historically, I've always used one of those bulky, aftermarket fan+heatsinks. I've looked at quite a few liquid cooling solutions and they appear much more aesthetic, but I'm looking for pure performance and durability (ie temps during full load, longevity, failure rates, etc).

Anyone made the leap recently from air --> liquid or vice-versa? And if so, what made you finally pick one or the other?

Posts: 3613
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: CPU Cooling: Air vs Liquid

Post by Abula » Sun Apr 26, 2020 1:22 pm

Most of the time Air is quieter than Water, but if take a 9900k and OC heavy you will create so much heat that water will be likely quieter than air under those conditions. I no longer peruse OCing as i used to, i still do it for benching a new pc, but i leave it at stock for 24/7 use, as i prefer quiet operation than slightly better performance, this is choice you have to take.

As there is a real rivarly atm on the CPU market, we will see both companies try to release better CPUs each gen, and a lot of times they will be already very close at their limit, the turboboost and pbo are more aggressive than past gens, making ocing something that most (not all) are not really that interested. That said because of the CPUs are close to their limit, they also run hot, so it could go either way, i still think Air is quieter for the average user.
GameMi - Intel i9 9900K + Gigabyte Z390 AM + Asus RTX2080Ti Strix | MiniMi - Intel i7 8700K + ASRock Z370ITX + MSI RTX2080Ti
StreaMi - Intel i9 7940X + AsRock X299 Taichi + Asus GTX1080Ti Strix | WorkMi - AMD 1950X + AsRock X399 Taichi + MSI GTX1080
ServeMi - Intel Xeon E3-1230 + Supermicro X9SCM-F + 144TB | StoreMi - Intel Xeon E3 1240v5 + Supermicro X11SSM-F + 42TB | CamMi - Intel i7 4770K + MSI H81i + Blue Iris

Olle P
Posts: 660
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Re: CPU Cooling: Air vs Liquid

Post by Olle P » Tue Apr 28, 2020 6:12 am

I've recently gone to water cooling.
Will write a long commentary on it.

Posts: 1
Joined: Fri May 01, 2020 11:30 am

Re: CPU Cooling: Air vs Liquid

Post by agv » Fri May 01, 2020 12:29 pm

marvelsferb7 wrote:
Sat Apr 25, 2020 7:08 pm
I've looked at quite a few liquid cooling solutions and they appear much more aesthetic, but I'm looking for pure performance and durability (ie temps during full load, longevity, failure rates, etc)
Gamers Nexus is one of the very best objective and honest review resources for PC components today. They are not silence-focused, as SPCR was/is, but they do value sonics as well as performance in their review-perspective. A few weeks ago they released a data-driven video addressing just this question.

The video was prompted by their testing of a new AIO CPU cooler, the Arctic Liquid Freezer III. It's the best available right now, from a performance and sonic perspective. In the video the point is made that sonics and thermals are fundamentally linked, and to discuss one without the other is meaningless. This is a key truth that many performance-oriented review and news sites ignore.

A key point made in the video, stemming from extensive testing, is that when comparing air cooling to closed-loop liquid cooling, liquid has a performance and sonic edge. But it's limited to about four or five degrees Celsius at equivalent sonic levels. This difference results from using the best-available equipment in terms of both the air and liquid options.

But performance and sonics, while the most important metrics, are not the only metrics, as you said above. On durability: if you maximize durability by using the best liquid and the most reputable equipment, you have about five years before you need to check on and maintain a closed loop in the best, problem-free setup. You will almost certainly need to add liquid. But even with no corrosion stemming from bad choices of liquid, erosion still occurs. After that five-year period you must check for eroded metal. You will absolutely find some, but it may not be a concern at that point. Nickel may erode, exposing the underlying copper, for example, but not compromising the structural integrity of the piece. Or structural integrity could be threatened. You need to check.

Even if structural integrity is sound after five years, it will not last forever. Liquid-cooling parts all wear out comparatively rapidly, and we're talking about a handful of years before the danger requires attention, not a decade.

If bad choices are made with regard to liquid, corrosion can be a problem much earlier, and biologics even earlier.

Durability aside, catastrophic failure is a remote possibility. With good equipment and sound practices, this need not be a pressing concern. But there is a small chance of losing expensive components to a liquid failure. If the equipment is not top quality, seals can fail earlier, pumps can die, or other failures can result in flow stoppages or leaks. The risk just a fact of life that you accept or don't. How small is the risk? With top equipment and/or safety protocols, very small, but it exists. Most people handle probabilities pretty poorly. This risk is something that is not high enough to create anxieties. Just don't risk your equipment with liquid cooling if you cannot possibly replace it, and take every reasonable precaution with your system.

That's the long and short of the issue for mass-market cooling components. Liquid cooling is much more complex and requires more design considerations. If you address those well in designing your system, then be aware of the systems that check for failures in pumps and flow. Be aware of the needed maintenance schedule. In return you will get 4 or 5 degrees of Celsius at high heat outputs, which can be spent on a minor increase in OC speeds. You will also get a different aesthetic, which you might prefer. If you don't engage in rather aggressive overclocking, you will get no thermal advantage. At stock or moderate-OC levels, air performs as well as liquid.

Now, all of the above refers to mass-market, AIO, closed-loop solutions. If you move into custom-designed liquid cooling designs, with hard tubes and open loops and bespoke parts, well, you could likely get more thermal performance. Another handful of Celsius degrees is plausible, at the very, very high costs and inconveniences involved. The costs escalate very quickly, into many hundreds or thousands of dollars. This is not about performance in a practical sense - it's a hobby in and of itself.

So there are two possible reasons to get into liquid instead of air. Well, three. One can be dispensed with quickly. The main forces driving the wider adoption of liquid are commercial forces and aesthetics / ownership pride. The component manufacturers have a direct economic interest in creating an additional mass market for components, and aggressively creating demand. This has dovetailed with consumers' desires to admire their own purchases and designs to create a vastly expanded liquid-cooling market.

There's nothing wrong with preferring the look of liquid parts. That's one perfectly good reason. Being suckered by the marketing forces that want you to want additional components, that's a bad reason. The last reason is the only performance-based reason. If you want a very minor to barely moderate speed increase by aggressively overclocking your gear, you can get it by moving to liquid, at the cost of higher sound levels over stock. You won't get higher performance "for free" on liquid, that is, without higher sound volumes. And you won't get huge performance gains.

If you want to roughly double those gains at extreme costs, that's not a practical interest. That's a hobby. You can extend that hobby with all kinds of extreme pursuits, like liquid-nitrogen cooling and crazy hardware setups that are loud as hell and wildly impractical but reach new thermal lows. That can be a lot of fun if it interests you. There are communities and competitions to join.

So get into liquid if you want to, just be sure that you will get what you want. Most people who get into liquid expect more practical results than are available.

Posts: 2
Joined: Fri May 08, 2020 3:51 pm

Re: CPU Cooling: Air vs Liquid

Post by aleemiqbal » Fri May 08, 2020 7:14 pm

I personally prefer Air

Olle P
Posts: 660
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Re: CPU Cooling: Air vs Liquid

Post by Olle P » Sun Jun 07, 2020 11:52 pm

marvelsferb7 wrote:
Sat Apr 25, 2020 7:08 pm
Anyone made the leap recently from air --> liquid or vice-versa? And if so, what made you finally pick one or the other?
Here's my story:

There were several reasons:
1. I've been wanting to do it for about 15 years, but not thought I could motivate it financially. (Still can't, but at least I found it affordable!)
2. I've got a master’s degree in engineering combined with some autistic characteristic that makes me dig deep into any interest of mine. Having done a lot of theoretical work on water cooling I really wanted to test (some of) my theories.
3. One "theory" is that water cooling is good at preventing rapid temperature changes in the parts, thus preventing thermal fatigue.
4. "Just because!"

The final decision to go to water cooling was taken by me a few years ago, and then I purchased the parts required. Due to various family reasons I didn't get to install it in a timely manner though.
There were also some problems with the hardware chosen, like a refurbished CPU cooler that didn't come with the required screws (had a double amount of Intel screws that can't be used for my Ryzen CPU) and a pump that's noisy and probably not powerful enough. So late last year I opted to buy a new CPU block, new reservoir and pump.

In the meantime I had also bought (and installed) an All-in-One CPU cooler for my son's computer.

Anyway, installation of my custom loop with soft hose was fairly straightforward. Filling it up with water didn’t go quite as smooth as expected (because of filling port a bit narrower than optimal) but was eventually done okay.
When testing there were no leaks or so. My GPU is running cooler than expected. The big let-down though is the CPU block. It really doesn’t perform well. Starting a stress test it takes just about three seconds for the CPU to get a 30 degree temperature increase before getting somewhat stable! (I’ve tried several different approaches to fix this problem but not yet found out what’s the cause of the poor thermal coupling between CPU and water.)
My third point above is thus proven both right and wrong: The GPU does indeed vary in temperature much slower now than before, as expected. The CPU, on the contrary, does not.
My first pump and reservoir, a BT-LT in a 5.25” bay, is a fairly noisy combo. The replacement AlphaCool Eisball with a D5 755 pump is as good as silent. I need to hold my ear within 15cm to hear it running at all. (And it’s mounted on top of the computer case!)
Cooling of the water is passively done by a twelve module AlphaCool Cape Cora HF convector mounted on the wall in my study.

I have yet to do some more checks on my son’s computer to see how well the AIO performs and how it influence the graphics temperature. I did a somewhat unorthodox install in the Fractal Define C case:
The radiator is mounted in the top, with the fans pulling air into the case. All openings surrounding the radiator are closed off to prevent air from leaking out that way.
Two 14cm fans are mounted in the front, also pulling air in. Surrounding openings are closed off.
The fan mount at the rear behind the CPU is completely shut, using adhesive tape. Don’t want any air to move out there.
Only the openings at the rear below the graphics card are left open, so that all air pulled into the case must pass the graphics card and is thus accessible for cooling the GPU.
Fan control is the issue here. The radiator fans rely on the CPU temperature. That’s very straightforward. The front fans should ideally be dependent on the GPU temperature, but that can only be done if both the motherboard and graphics card are of Asus high end models, which neither are. As is the computer is loud while gaming, with some fans running at full tilt, and I have to find out what fans and why.

Post Reply