As previously mentioned in one of last month’s posts, with this write-up I’m wrapping up the coverage of the various building blocks I’ve assembled in preparation for building my three planned new PCs. This last topic, on thermal considerations and solutions, is a critically important (albeit, like power supplies before it, a commonly under-prioritized) one.

After all, not only are the power supplies themselves less than 100% efficient in converting incoming AC to a variety of DC output voltage-and-current combinations, the various ICs and other circuits powered by those DC voltages are equally if not more inefficient in transforming that DC power into meaningful calculations and other work. Inefficiency equals heat, and this heat needs to be rapidly and robustly removed from the system in order to ensure long-term functionality.

Let’s begin with some basic statements, which will likely seem elementary when you read them but which I suspect are commonly overlooked at system implementation time:

• Heat rises (duh, right?)


• Ventilation intake and outflow vents must be free of close-proximity potential airflow-impeding objects around them, and the intake vents should have access to as-cool-as-possible ambient air


• Airflow through the system must also be carefully considered. Not all possible vents (specifically the ones with active fans associated with them, i.e., “forced,” versus just passive) can be air intake (“push” positive pressure) portals; some of them must also be outflow (“pull” negative pressure) vents, otherwise there’ll be no path for the generated heat to escape!

On that last bullet point in particular, I’m going suggest you now press “pause” on this piece and go read my colleague (and former boss) Bill Schweber’s excellent write-up on the topic (see “”) published in EDN about a year ago…then come back here afterwards, of course. Bill’s write-up includes a diagram of probably the most common airflow configuration for PCs: inflow through the front, with (passive) airflow out the top and (active) airflow out the back:

Benefits of this approach include that the ambient environment in front of a computer tends to be cooler (not to mention less obstruction-filled) than that behind the PC, and that it also tends to be less dusty. The dust-accumulation impacts on system reliability introduced in Schweber’s piece cannot be understated; while it’s sometimes possible to add dust filters to a PC case’s air intake vents, doing so tends to constrict airflow through those same vents…notably (but not just) if you forget to clean the filters periodically!

That said, there are potential downsides to this approach. Aesthetic (cosmetic) considerations can constrain the number and size of intake vents at the front of a system. And noise coming from fans mounted at the front is also generally more audible than that from fans mounted at the rear.

The subject of fans begs for a blog post all its own: generally stated, they move more air at high vs low RPM and when larger versus smaller, although both faster speed and larger sizes also commonly correlate to more audible noise. That all said, advanced fan blade (and cluster), bearing and motor designs can both mitigate noise and boost airflow, albeit at higher cost. For these reasons, fans are most common at the back panel of a PC, regardless of which direction they’re transferring air, with passive (and sometimes fan-assisted active) vents at the top of a system to aid in heat removal, exploiting the earlier-mentioned “duh” that hot air rises.

All modern GPUs that I’m aware of leverage at minimum a heat sink over the graphics IC to aid in heat transfer away from the package, and most graphics boards also include one or multiple built-in fans. Look, for example, at the push-pull two-fan arrangement used in the Founder Edition branded cards that NVIDIA has developed for its :

The fan on one side pulls air into the board, where it flows over the GPU and memory heat sink(s) and is subsequently exhausted by the fan on the other side of the board. However, of course, that GPU-generate heat is at that point still contained within the PC enclosure; system venting is also required to subsequently exhaust it to the outside world.

PC power supplies, as previously mentioned, also generate their fair share of heat. But depending on how you orient them within the system, they may also act as a system heat outflow device…although again, as always, there are trade-offs. An example is the large fan in a PSU, such as in one of the units discussed in one of last month’s write-ups:

The fan is associated with an air intake vent; the other vent, shown at the back panel where the AC input plug is also located, is a passive outflow vent. In a traditional top-of-PC PSU configuration such as the one shown below, from another prior write-up of mine (see “”) the fan is at the bottom of the PSU:

It pulls in air from the interior of the PC and exhausts it out the back. Sometimes PSUs are mounted at the bottom of the PC versus the top; the same concept applies, with the PSU installed “upside down” so that the intake fan is at the top.

In both cases, however, the PSU’s internal circuitry is exposed to warmer-than-ambient air that’s already been heated by other PC subsystems such as the CPU and GPU, with resultant potential PSU reliability impacts. Alternatively, therefore, some PC enclosures include passive vents at the top or bottom that match up with the PSU intake fan location. If you correspondingly mount the PSU in the “reverse” orientation from “normal,” you end up with a “closed airflow” configuration that draws in cooler air from the outside instead, at the trade-off of the PSU then not assisting in overall system airflow.

DRAM has historically not been known as a particularly “hot” PC subsystem, but in the modern era of high clock speeds and multi-clock-edge data transfer rates, heat spreaders are increasingly common. A set of four Corsair Vengeance LPX DDR4 DIMMs I recently bought (for another planned system I haven’t yet told you about) even came as an “Airflow” kit also including an mounted on a bracket and intended for installation above them:

On that note, here’s another general comment on fans. Keep in mind that each one you add to a system will need a corresponding power “feed,” either coming direct from the PSU (therefore usually running at full speed all the time) or via a motherboard-located connector (where BIOS-controlled, temperature-defined speed control is often available). It’s also possible to daisy-chain multiple fans to a single motherboard connector via a multi-tap cable, although all fans connected to that cable will then run at the same speed (also, don’t exceed the connector’s maximum current-source capability!). That all said, more motherboard fan connectors are better, especially for high-end systems that beg for more involved multi-fan setups.

In the “old days” (extending to today, of course, for those of you still using rotating magnetic storage media) it was common to ensure that a system fan blew cool air around the hard drive(s), whose platter(s) (and associated motor(s)) tended to generate plenty of heat. This dedicated-fan assistance was particularly critical with dense (multi-platter) and/or high-RPM (high performance) HDDs, and with multi-drive “stacks”. And as with DRAM, the HDD’s successor, the SSD, is getting denser and faster, meaning that heat sinks are increasingly common (see “”):

Last (but not least) let’s look at the CPU. While low-power processors for embedded systems sometimes use only “bare” packaging, CPUs absent heat sinks are now commonly seen only in the lowest-end PC configurations. At minimum, as I showed in a beginning-of-year post (see “”) there’s a heat sink (usually aluminum, ) in combination with ducting that routes the heated air to a system fan for exhaust from the system enclosure:

More commonly, there’s a CPU-dedicated fan mounted right at the top of the heat sink (see “”):

And high-end systems, and/or those built using diminutive enclosures with limited airflow and capacity for fans, leverage lessons learned from automobiles. A personal perspective: my first Volkswagen camper van, a , had a four-cylinder air-cooled engine. It frequently overheated, especially when going uphill on hot summer days, and more generally it was quite underpowered out of thermal-constrained necessity.

Its successor, a , employs a modern six-cylinder water-cooled (and antifreeze-augmented) engine that thankfully shares none of its predecessor’s shortcomings. Similarly, with high-end PC CPUs it’s becoming increasingly common to replace the heat sink with a “water block” mounted to the top of the CPU and connected to tubing which, via a pump intermediary, routes the () coolant liquid inside to a fan-augmented radiator mounted at a system panel (frequently but not always the front), then back to the CPU to transfer away more heat.

While it’s possible to build up such a liquid cooling setup yourself, you need to worry about getting air bubbles out of the entire system (which, as with an automobile, also benefits from periodic flushing and replacement), as well as the possibility of leaks at component junctions due to tubing failures, etc. Or you could take the “easy route,” as I’ve done, and just purchase an AIO (all-in-one) system that includes everything pre-assembled. In either case, you also need to concern yourself with the potential for pump failures, of course, but the outcome is conceptually no different than the failure of a heat sink-mounted fan in a traditional setup.

Speaking of which, what thermal solutions have I come up with for each of the three systems I’ll be building? The mini-ITX system based on the AMD Ryzan 7 3700 X uses :

It includes a front-mounted 120 mm fan and has mounts for two 80 mm fans on one side…which I won’t be able to use, because I’m installing a dual-slot graphics card (you can have one or the other, but not both). Air vents elsewhere in the case are passive-only in nature. For CPU cooling purposes, I originally planned to use a I’d bought (along with ):

But upon further reflection, considering both the Corsair AIO’s age and its previously-used status when I acquired it, I’ve decided to hold onto it as a spare and instead go with a brand new , which set me back (I also bought the supplemental extended 3-year warranty for an additional $10.99).

One other nuance of this particular setup bears brief mention. Normally, on a motherboard intended for use with AMD CPUs, you’d find an AMD-compatible heat sink (or, in this case, water block) mounting bracket. However, as , the I’m using with has Intel LGA115x cooling mounts instead of the usual AMD AM4 cooling mounts. The reason why, AnandTech points out, makes sense explained: “There aren’t as many low-profile coolers available on the market for AM4 as there are for LGA115x. Including an Intel mount on this smaller form factor AMD board actually improves cooler compatibility.”

The Intel system, based on a Core i9-10850K processor, is housed in a somewhat roomier (albeit still diminutive) mini-ITX enclosure, :

In this case (bad pun intended), there’s a 120 mm intake fan upfront, along with an 80 mm intake fan on one side, the latter intended to blow cool ambient air directly onto the CPU. As with the Cooler Master Elite 110 enclosure discussed earlier, air vents elsewhere in the Elite 120 Advanced are passive-only in nature. And as with the earlier AMD CPU-based system, I plan to again harness a Cooler Master MasterLiquid Lite 120 AIO, mounted to the front panel and replacing the existing 120 mm fan, to “chill” the Intel processor.

And speaking of “AMD CPU-based systems”, what about the third PC-to-be, which ? The enclosure in this case is the comparatively spacious :

And it’s got plenty of active air movement potential: a preinstalled 120 mm fan in the rear, along with room for up to a 360 mm liquid cooling radiator (comprising up to three 120 mm fans or two 140 mm fans) upfront and plenty of passive venting elsewhere. Speaking of which, the power supply mounts at the bottom, and there’s an associated fan vent for it, so I plan to do the “reverse” installation mentioned earlier.

For now, I plan to go with a two-fan radiator configuration, holding off on the three-fan AIO also in my possession for the previously hinted-at “another system to be named later.” I’ve got three to choose from (I cannot resist a good sale): a brand new dual-120 mm fan-based NZXT Kraken X53 (I’m not fond of the RBG lighting, but I don’t have to enable it), which I’d bought for :

And an also-brand-new dual-120 mm fan-based (ditto on the RGB lighting) for , and also purchased with a four-year $13.99 optional extended warranty (fans fail, don’cha know):

And an open-box dual-140 mm fan-based (have I mentioned my distaste of RBG lighting?) for :

Right now, I’m hoping that one of the dual-120 mm configurations will suffice, thereby enabling to use the other as a spare for this PC and reserving the dual-140 mm AIO as an option for the “system to be named later.” And speaking of which, with this set-up series of posts complete, I really need to get going on the actual builds now, don’t I? Until then, I as always welcome your thoughts in the comments on thermal considerations or anything else I’ve so-far discussed.

— is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.

Related Content

The post appeared first on .