this post was submitted on 18 Jun 2025
16 points (100.0% liked)

Hardware

549 readers
4 users here now

A community for news and discussion about the hardware side of technology. Questions and support posts are also welcome, so long as they are relevant to hardware and interesting technologies therein.


Rules

1. English onlyTitle and associated content has to be in English.
2. Use original linkPost URL should be the original link to the article (even if paywalled) and archived copies left in the body. It allows avoiding duplicate posts when cross-posting.
3. Respectful communicationAll communication has to be respectful of differing opinions, viewpoints, and experiences.
4. InclusivityEveryone is welcome here regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity and expression, education, socio-economic status, nationality, personal appearance, race, caste, color, religion, or sexual identity and orientation.
5. Ad hominem attacksAny kind of personal attacks are expressly forbidden. If you can't argue your position without attacking a person's character, you already lost the argument.
6. Off-topic tangentsStay on topic. Keep it relevant.

founded 11 months ago
MODERATORS
 

What is the rationale behind GPU heatsink fin design?

So I have seen a few GPU heatsinks and I wonder why some of them are how they are.

GPU placement in Cabinet

The originally intended and most widely used placement for ATX cases at least, is installing right on the PCIe slot, which goes horizontally and with the air incoming from the front of the case.

gpu_traditional_fit

Then we have the "vertical" placement using the riser cable, which changes the direction from which the GPU fans take in air, but keep the GPU front, in the same direction, again lining up with the incoming air from the front.

gpu_trendy_fit

Finally the very rare vertical placement, which has the front of the GPU rotated towards the top of the case, making it not line up with an airflow incoming from the front.
This also makes the GPU's ports inaccessible the standard way, giving a reason to the rarity.

gpu_exotic_fit

Some examples I found for this

Here we see that the first 2 placements would make up most of the ATX PC builds with the third one being either for different case styles or for extensively customised builds.

Fins, fans and airflow

I would assume that having the air flow along the fins would be better than it crossing the fins at ⦜90°. And even if the air flow due to the case inlet is being ⟂ to the airflow from the GPU fans (in case of front incoming air flowing along the fins), it should still lead to overall increase in air pressure (hence, air density) between the fins.^[Source: Mental simulation] Considering that GPUs tend to have gaps in the IO shield to let the air go out the back, I would assume they (designers of the thermal dissipation solution) want air from the fins to go out the back, which would be better with fins parallel to front incoming air.

gpu_airflow_illust

Inference and Doubt

From the above 2, it would make sense for most GPUs to have their heatsink fins going along their length instead of their width, right?

Then what's up with the ASRock lineup, with all cards other than the Passive model, the Creator cards (which have the front covered by the shroud anyway, so no incoming airflow) and the watercooled cards (which is not applicable) having the fins ⟂ to front air inlet?
And of course they are not the only one doing it that way.


Follow Up

While discussing in this thread, I realised another point (which I didn't state anywhere in the comments):

  • While it is desirable to have higher air density, it is not desirable to have air increasing in density while in the inter-fin space. Because that would cause the heat released when changing density to be transferred to the fins.
  • Another way to say that is, the air will get hotter, while increasing its density. So, how much this factor matters, will depend upon the initial temperature difference between the air and the fins and how long the dense air stays in the inter-fin space. Oh and also the composition of air.
top 20 comments
sorted by: hot top controversial new old
[–] [email protected] 6 points 1 week ago (2 children)

Speculation: the airflow on the GPU fans completely dominates the airflow on the GPU's heatsink, making "ambient" airflow direction essentially irrelevant for airflow that goes directly across that heatsink except as it pertains to reaching those GPU fans.

Having the short-fin direction also allows significantly more "exhaust area" for the GPU's hot air than the long-fin direction (effectively both long sides are "GPU exhaust" instead of both short sides being "GPU exhaust") and additionally the exhaust that would face into the front of the case -- which I assume would be less efficient due it fighting incoming airflow -- doesn't have to be used.

[–] [email protected] 3 points 1 week ago

This is what one of our engineers found out when trying to pick an orientation for an FPGA heatsink on a board, fins inline with the chassis fans or perpendicular and he found that the perpendicular placement stayed cooler since the heatsink had a fan blowing the hot aire to both sides and the chassis fans blowing it away than having the fan on the heatsink fight the incoming air from one side and exhaust from one side

[–] [email protected] 1 points 1 week ago (1 children)

I see, so the ⟂ orientation might actually be better and if the other variables, like fan quality and fin material were the same, I could have expected an increase in cooling performance.

[–] [email protected] 3 points 1 week ago (1 children)

Some other factors to consider:

  • Parallel-to-slot (long) fins can make it harder to effectively use multiple fans because the airflow directions (once the air is redirected upon hitting the GPU heatsink) will clash between each fan (see image).
  • Heatpipe design might be simpler or better when the pipes are mostly-straight across the perpendicular-to-slot (short) fins compared to having to have bent pipes of much more variable length.
  • Short fins themselves might be easier or cheaper in manufacturing somehow?
  • Long fins can mean that air spends more time being heated before being ejected (making its average temperature higher before ejection, lowering heat transfer).

A crudely-drawn diagram of GPU fans directing airflow towards each other when heatsink fins are arranged length-wise (parallel to the PCIe slot).

[–] [email protected] 2 points 1 week ago* (last edited 1 week ago)

Yeah, I considered a similar image, but also:

pressure gradient expected for air pushing fans
(But I also realise that one would have to be careful when designing this, to prevent turbulence in the center, causing heat generation.)

pressure gradient including cabinet airflow

Which I considered would be a big factor.


Short fins should be cheaper, yes, if they use the skiving process. ∵ shorter fins ⇒ less total force required ⇒ cheaper machinery usable/ longer machine life due to lower stresses.


making its average temperature higher before ejection, lowering heat transfer

Well yeah, but I feel like the air coming out of the fins was not hot enough to say that.
I would rather put the factor of higher air density (caused by a higher pressure) between the fins enabling higher heat transfer.
Of course, which factor affect more in that case (pun intended), would be the engineering part.

[–] [email protected] 3 points 1 week ago

Just wanted to say this seems like a good summation and reporting.

[–] [email protected] 2 points 1 week ago (1 children)

I have seen my old RX 570 from MSI, work much cooler even with an absurd amount of dust in it than the ASRock 7600 and also noticed after the recent upgrade to my front inlet fans that they hardly made a difference to the GPU temps, which lead me to put some brain-time into modelling the airflow due to this GPU's fins.

[–] [email protected] 5 points 1 week ago (2 children)

570 TDP is 150w 14nm 7600 165w 6nm

Sorry but you cant really compare the heat output of the 2 cards.

If you had a passive cooler airflow would make a difference but the fans on a GPU are what do the cooling. The only thing those front fans do is provide cold air. The gpu fans arr designed for high static pressure, something a fan not touching the fin stack can't do.

Also changing the orientation of the GPU can have negative effects due to the heat pipes being designed for a specific orientation. See AMD 7xxx reference card debacle.

[–] [email protected] 3 points 1 week ago (2 children)

If you want to get technical, the heatsink/fins helping to move heat away from the chip (regardless of any fans) can also be considered "cooling".

[–] [email protected] 1 points 1 week ago (1 children)

The PCB too if you want to be a jackass about it

[–] [email protected] 0 points 1 week ago (1 children)

I think I saw some company advertising having specially added heat dissipation channels inside the PCB, near some hot components.

[–] [email protected] 2 points 1 week ago (1 children)

Once the PCB is heat soaked its basically useless at thermal disippation because fibreglass/resin (PCB) is an insulator. There is research being done to make micro channels into the silicon to directly water cool the chips but i doubt it will eer get to market or even be made for industry though

[–] [email protected] 1 points 1 week ago (1 children)

ceramic heat plates (or ceramic PCBs) can be used for very hot applications

[–] [email protected] 1 points 1 week ago

Yeah, not GPUs

[–] [email protected] 0 points 1 week ago

Well of course and so is the lid of the processor die and the material used to paste the lid onto the die.

When I think of cooling, I first think of the processor die's surface area, then the lid, then the heatsink after which come the fans.
The heatsink being an important part here, since that determines how much of a difference a fan will make.

e.g. I recently changed the heatsink on my SSD from the one provided with the motherboard to another one with some fins and placed a NF-A4x10 (x20 was unavailable) to point in its direction to get ~4℃ (yeah, just that and it is still ~15℃ over ambient, when idle (state 4)) of better cooling performance. Though I am yet to experiment with other fan directions and the fins are not particularly made to be used with a fan either. But still, a design with fins would be better than just a block of metal.

[–] [email protected] -1 points 1 week ago (2 children)

you cant really compare the heat output of the 2 cards

Yeah, that's the main reason I didn't try to be comprehensive about it.

gpu fans arr designed for high static pressure

I hope so. The fans along with the rest of the cooling system on the ASRock card seems just low quality, as compared to the MSI one (and I would have bought an MSI card if it were available at the time) and that is reflected in the memory temps, despite me making sure that it had a thermal pad, connecting the heatsink to the memory modules, unlike the ASUS cards.

[–] [email protected] 2 points 1 week ago (1 children)

https://youtu.be/WRm6N1A74oE Oh and a few years ago buildzoid explains why having a heatsink that has the fins going length wise suck (aka blower heatsinks)

[–] [email protected] 0 points 1 week ago

Yeah that's just the other extreme.
Such a long pathway that you need way more air speed than you should be looking for.

[–] [email protected] 1 points 1 week ago (1 children)

GDDR6/6x runs hotter than GDDR5. You need to learn a lot more about PC components m8

Like I said, you cant compare the 2 cards at all other than they look similar. Pretty sure the thermal engineers at ASROCK are much smarter than you take them for

[–] [email protected] 0 points 1 week ago

Pretty sure the thermal engineers at ASROCK are much smarter than you take them for

Yeah, probably just my bias after having that bad experience with ASUS Gaming laptop, in which they stuffed components way above the case's TDP.