How to choose between GPU cooler designs [Exhaust VS. Open-air] - Overclock.net - An Overclocking Community

Forum Jump: 

How to choose between GPU cooler designs [Exhaust VS. Open-air]

 
Thread Tools
post #1 of 9 (permalink) Old 09-15-2016, 11:29 AM - Thread Starter
Old dog, old tricks
 
ronnin426850's Avatar
 
Join Date: Sep 2009
Location: Europe
Posts: 9,738
Rep: 574 (Unique: 474)


Introduction:

Hello!
Here I will try to explain some commonly ignored aspects of GPU cooling that should be taken under account when deciding on what GPU to buy to best fit your situation.

Who is this guide for:
This guide is for people on a budget, who want to get the best overclocking and gaming experience out of their mid- or mid-low-end PC (I'm looking at you, AMD crowd).
It could also be useful to people who want a 2 or more GPU setup.
This guide does not apply to people who do liquid cooling, submerged builds, or caseless builds, or will spend top dollar for whatever toys the other cool kids have smile.gif

Why is this guide needed:
People are being massively mislead when asking for advice on GPU selection. There are these two special groups of people - the hardware snobs, who believe they are correct because they own a 5000$ build that plays Crysis 5 and it's awesome; and the fanboys, who will dismiss anything that is not their favorite thing. These people have been giving wrong advices to inexperienced PC enthusiasts and overclockers lately. This guide is needed, it order to teach people how to decide for themselves what is better, and not to rely on the opinions of people with questionable knowledge and ethics.

Why should you trust me:
You shouldn't. That's the point. You should learn how to estimate your requirements, and trust yourself, instead of some guy on the internet.

The problem:

There are generally two types of GPU cooling: Exhaust cooling, and Open air cooling.

The common conclusion on the Internet is that exhaust cooling (often incorrectly called "reference cooling") is terrible, and should be avoided at all cost.
Proof of this point is sometimes provided in the review stats of popular hardware sites or forums: Open air cooled GPUs stay cooler, are less noisy, and sometimes even cheaper.

So why would you want to buy a reference cooler?

Let's examine the differences and intricacies of each:
Exhaust cooling:
- Has 1 side-blowing fan (incorrectly called "a turbine")
- Has a heatsink that runs for some of the length of the GPU
- Has an opening at the rear of the card
- The air pushed from the fan passes sideways through the heatsink and exits the case via the rear opening
- Has lower static pressure on the surface of the heatsink

Open air cooling:
- Has between 1 and 3 conventional fans
- Has one or two heatsinks, often coupled with heatpipes
- The air pushed from the fans runs through the heatsink fins vertically
- May or may not have higher static pressure on the sink.

Exhaust coolers are shrouded, which means that the air pushed by the fan is forced to go in one direction - through the sink and out the back end. This means that the entire airflow is optimally utilized, and the whole surface of the heatsink gets airflow. The generally lower surface pressure on the sink means that the heat transfer fin to air is generally worse for the same area and airflow. That is why the heatsinks for exhaust coolers and their fins are designed for maximum usage of the air flow - they are often thicker and spaced further apart. The heatsinks may also utilize heatpipe technology, although the effectiveness of the heatpipe in this case is limited by the size of the sink and the fact that is makes direct contact with the chip. Leaving all other aspects aside, this is a very effective cooling design, that does not generally depend on immediate environment, and provides a great solution against heat buildup.

Open air coolers are, well, open, which means that air pushed from their fans is free to go wherever. A portion of the air goes sideways, and depending on the design of the shroud, may never pass through the heatsink. A portion of the heatsink that is directly under the fan motor is usually left without airflow. If the card has more than one fan, in most designs the airflow of two neighboring fans collides in the area between them, making it harder for heat to leave the area of the heatsink. The heatsinks themselves typically have greater area, compared to exhaust coolers, because they do not need to leave space out for the fan. Often times the sinks do not have direct contact with the core, but are instead based on many heatpipes with aluminum fins running through them. This type of cooling depends on the surface pressure of the air blown by the fans directly onto the heatsink. The fins are in close proximity to the fan, they are thinner in order to reduce the resistance and improve heat transfer, and more closely spaced in order to maximize the area and compensate for being shorter. The airflow of open air coolers is often greater in the case where the card has 2 or 3 fans, and lower when it has only 1 fan. Although the heatsinks of open air coolers are generally better designed than those of exhaust coolers, the cooling effectiveness is limited by the chaotic airflow and heat buildup around the cooler.
This means that open air cooler's effectiveness depends greatly on the environment - ambient temperature, stable and sufficient airflow in the case, space to breathe, proximity to other heat sources.

In a small case, even one equipped with good intake and exhaust fans in a standard configuration, an open-air cooling may create swirls which interrupt the airflow, result in buildup of heat around the card, and cause the exhaust fan to create negative static pressure around the CPU.
In large, spacious cases this effect is avoided.
Exhaust coolers do not have such problems, although in a situation where the card manages to pump out more air than the intake fan pumps in, you may still have negative pressure around the CPU. However, in this case you do not have air swirls in the lower half of the case that would prevent even air distribution towards the CPU.

In cases with enough space and stable airflow, the better design and more airflow of an open air cooler results in noticeably lower temperatures on the GPU, sometimes in the matter of tens of degrees Celsius.
Exhaust coolers do not depend that much on environment, which also means they do not benefit from a good environment. A good airflow in the case will not improve the work of an exhaust cooler as much as it would improve the work of an open-air cooler.
This is somewhat remedied with the introduction of backplates to exhaust-cooled cards. The backplate makes it so exhaust cards benefit more from good environment. The negative aspect is that the backplate causes heat buildup on the side of the CPU instead of underneath the GPU.

Due to the side-blowing aspect of exhaust airflow, the fan of an exhaust card needs to work harder in order to produce the sufficient airflow in order to cool the heatsinks the same way a conventional downward blowing fan would, in the ideal case. Aslo, due to mechanical constraints in directing the airflow, the fan can not be of the same diameter as a conventional fan. This means exhaust cards have smaller fans that are forced to work harder. This results in the noise that a lot of people list among the reasons to avoid exhaust cards.
GPU manufacturers of course know of this, and remedy it somewhat by using less agressive fan curves. This results in higher temperatures under load, that a lot of people list as another reason to avoid exhaust cards.
However, often times it is the case that an exhaust card boosted to 100% fan speed will keep cooler than an open-air card boosted to 100% fan speed, albeit being much more noisy. So the cooling performance itself is not worse on these cards, and the two reasons to avoid them are in fact only one reason - the temperature to noise ratio is worse.
This is proven with many examples across many reviews.

However, there are aspects of the standard testing methodology that are not discussed, but are of great importance for cooling performance.

Some of the most important:

1) A lot of these reviews are done with the best parts possible, in order to show the true performance of the GPU being tested, without other limiting factors. This often means that the GPU is placed in a large, expensive case, equipped with many performance-grade fans, or on a test bench with open air flow, on a high-end motherboard with many PCI-E slots, and on the one that is not closest to the CPU. These are ideal conditions for open-air coolers.

2) A lot of these tests are only ran for short periods of time, due to the need to test a lot of hardware fast, write and edit the review, and publish it in time to beat the competition. This means no prolonged gaming sessions, and the effect of heat buildup is never seen.

3) Almost none of the tests measure the temperatures of other components, or test for OC stability of other components. If an open-air GPU would cause your CPU to overheat, you'll never know that from a review.

So you are to take these proofs of exhaust cooler's inadequacy with a grain of salt.

In a much more realistic scenario, you have a ~150W TDP GPU that you'd like to overclock, paired with ~80W TDP CPU that you'd like to overclock, cooled by a single tower CPU cooler, on a mid-range mobo with 2 PCI-E slots, placed in a mid-range case with average or below average cable management capabilities and 3 case fans - 2 front intake and 1 exhaust, because you read that positive pressure is good.
Let's see how this scenario plays out.

You buy an open-air GPU. You place it on the topmost PCI-E slot because the bottom one faces it directly to the PSU and it will get no air. You OC it mildly till it dumps about 170W of heat in your case (optimistic combined from GPU chip, VRM and VRAM). Your front intakes airflow gets swirled by the GPU fans and mixed with the hot air from the sinks. You GPU is being cooler by warm-ish air. The hottest air gets sucked upwards by the exhaust fan, passes through the CPU heatsink, heats it up, and is then propelled out the back end if the case.

Your GPU is being cooled by warm air.
Your CPU is being cooled by warm air and may be unable to OC.
Heat is ejected from the case by one fan.

You buy an exhaust GPU. You place it on the lower PCI-E, because its fan is at the very end of the GPU and beyond the PSU, and will not be choked by it. You OC it mildly till it blows 170W of heat out the back. Half of your front intakes airflow gets sucked by the "turbine" (actual turbines are generators, this is not a turbine) of the GPU, absorbs the heat from its sink, and is ejected out the back. The other half gets sucked by your CPU and top exhaust fans. It absorbs the heat from the CPU and is ejected out the back.

Your GPU is being cooled by cool air.
Your CPU is being cooler by cool air, sometimes making a difference of more than -10C under load.
Your GPU is further away from your CPU, heat from one will not interfere with the other.
Heat is ejected from the case by two fans.

This is roughly the configuration of all mid-low and mid-range gaming PCs.
These are ideal conditions for exhaust coolers.

Of course, they still will be more noisy. However due to fan curves, the difference in noise will not be as shocking as the review measures make it out to be, because in these conditions an open-air cooler's fans will be forced to work harder in order to try and compensate for the inadequacy of the entire setup, and will be in some cases just as noisy as a "turbine" working on half-time, depending on the design of the "turbine".

Finally, I'd like to address a discrepancy in two of the arguments against exhaust coolers:
1) They are designed by the GPU manufacturer to be the absolute cheapest cooler that will just barely do the job, board partners then invest into making better, open-air coolers.
2) They are more expensive than open-air coolers, while being worse, why would you buy one?

So, are they the absolute cheapest, or are they more expensive, make up your mind!

Conclusion:

As a conclusion I'd like to say this:
- Do not believe what people on the internet tell you. The fact that a person has a GTX 1095 and a Core i8 worth 1500$ each, or that they have used BrandX for the past 15 years and "know everything about them" does not make them an expert.
- Do not outright believe what review sites tell you. They do not review YOUR setup. They review 1 board, under conditions that are often not realistic.
- Your PC is not a combination of independent components, that if all of them separately are good, the PC will be good. Your PC is an interconnected system, especially when we talk about cooling and airflow. Everything can, and will affect everything else, unless you take steps to prevent it. Exhaust coolers are the most effective way to almost entirely exclude the greatest producer of heat from the equation.
- Use your own brain, do your own research. "Research" does not mean "read on 10 different places a subjective opinion". "Research" means "find out how and why things happen. Find out how different designs affect those things. Find out what your specific situation is, and apply this knowledge to take the maximum out of it".

PS: For another example of "don't buy the best", check out this other thread:
https://www.overclock.net/t/1597824/how-to-not-fall-for-gpu-marketing-tricks

Sorry for the wall of text, thank your for your attention, and good luck!

thumb.gif

If you have any notes and remarks, I'd be glad to take them smile.gif
ronnin426850 is offline  
Sponsored Links
Advertisement
 
post #2 of 9 (permalink) Old 09-15-2016, 01:38 PM
pass the butter
 
lovetobuild's Avatar
 
Join Date: Sep 2012
Location: Leeds
Posts: 537
Rep: 26 (Unique: 19)
+rep

I aim to misbehave.
Desktop
(15 items)
Server
(5 items)
CPU
FX 4170
Motherboard
ASUS Sabertooth 990FX
GPU
RX 580
RAM
Corsair XMS3
Hard Drive
Samsung SSD 840 Evo
Power Supply
Seasonic M12II Evo
Cooling
Corsair H110
Cooling
Many fans
Case
Aerocool X-Predator
Operating System
Windows 10
Monitor
Dell P2715Q
Keyboard
Microsoft Sidewinder X4
Mouse
Logitech M500
Mousepad
Aukey 900x400mm
Audio
SB X-Fi Xtreme, Logitech Z506, Betron HD800X
CPU
E8500
Motherboard
AsRock ConRoe
Power Supply
XFX TS 430
Case
Custom
Operating System
UnRAID
▲ hide details ▲
lovetobuild is offline  
post #3 of 9 (permalink) Old 09-20-2016, 03:42 AM
New to Overclock.net
 
8051's Avatar
 
Join Date: Apr 2014
Posts: 2,809
Rep: 22 (Unique: 16)
The cooling on my 6970 w/its blower style cooler was bad. The VRM's overheated as well as the GPU when I tried to overclock. Throttling was also a problem.

My Arctic Cooling Accelero II+ was far superior to the blower cooler and the only way I got it above 1000MHz without throttling, I was also able to substitute more powerful fans (as I've done w/my 980Ti).
8051 is offline  
Sponsored Links
Advertisement
 
post #4 of 9 (permalink) Old 09-20-2016, 03:57 AM - Thread Starter
Old dog, old tricks
 
ronnin426850's Avatar
 
Join Date: Sep 2009
Location: Europe
Posts: 9,738
Rep: 574 (Unique: 474)
Quote:
Originally Posted by 8051 View Post

The cooling on my 6970 w/its blower style cooler was bad. The VRM's overheated as well as the GPU when I tried to overclock. Throttling was also a problem.

My Arctic Cooling Accelero II+ was far superior to the blower cooler and the only way I got it above 1000MHz without throttling, I was also able to substitute more powerful fans (as I've done w/my 980Ti).

I remember high-end HD6xxx cards had serious problems with heat, they were too powerful for their own good smile.gif 6970 is a 250W chip, Jesus! When you add VRAM and VRM heat to the equation, you need some serious cooling, both on the card itself, and in the case, in order to keep it operational.

Also, the design of this particular card was terrible. Let's examine it:





First of all, 8-phase VRM for a 250W chip is INSUFFICIENT. Some high-end 200W GPUs have 12 or 16-phase VRMs in order to keep them stable.
Next, notice how the MOSFETS only make contact with the base plate of the cooler - no independent heatsinks, no fins, no nothing.
And most importantly - MOSFETS and *underneath* the fan. They get no airflow whatsoever. No wonder you had issues. Whoever made this failed terribly at his job.
ronnin426850 is offline  
post #5 of 9 (permalink) Old 09-20-2016, 01:36 PM
New to Overclock.net
 
8051's Avatar
 
Join Date: Apr 2014
Posts: 2,809
Rep: 22 (Unique: 16)
Quote:
Originally Posted by ronnin426850 View Post

I remember high-end HD6xxx cards had serious problems with heat, they were too powerful for their own good smile.gif 6970 is a 250W chip, Jesus! When you add VRAM and VRM heat to the equation, you need some serious cooling, both on the card itself, and in the case, in order to keep it operational.

First of all, 8-phase VRM for a 250W chip is INSUFFICIENT. Some high-end 200W GPUs have 12 or 16-phase VRMs in order to keep them stable.
Next, notice how the MOSFETS only make contact with the base plate of the cooler - no independent heatsinks, no fins, no nothing.
And most importantly - MOSFETS and *underneath* the fan. They get no airflow whatsoever. No wonder you had issues. Whoever made this failed terribly at his job.

Funny that you would mention the MOSFET cooling on the 6970, because the only part I retained from the stock cooling solution when I mounted the Arctic Cooling Accelero II+ was the portion attached to the VRM's -- and they worked fine cooling the power MOSFET's once they had axial fans directly blowing down on the heat plate.
8051 is offline  
post #6 of 9 (permalink) Old 09-20-2016, 02:26 PM - Thread Starter
Old dog, old tricks
 
ronnin426850's Avatar
 
Join Date: Sep 2009
Location: Europe
Posts: 9,738
Rep: 574 (Unique: 474)
Quote:
Originally Posted by 8051 View Post

Funny that you would mention the MOSFET cooling on the 6970, because the only part I retained from the stock cooling solution when I mounted the Arctic Cooling Accelero II+ was the portion attached to the VRM's -- and they worked fine cooling the power MOSFET's once they had axial fans directly blowing down on the heat plate.

Like I said, getting airflow at them is very important. I bet you'll see even better results if you remove the plate and attach some finned sinks. I myself had an Accelero 2+, great cooler, but I had to buy a few small aluminum sinks to attach to VRM and VRAM. Temps went down massively.
ronnin426850 is offline  
post #7 of 9 (permalink) Old 04-06-2017, 08:47 AM
New to Overclock.net
 
kevinf's Avatar
 
Join Date: Apr 2011
Location: Ontario
Posts: 1,804
Rep: 66 (Unique: 59)
I really don't understand people's distaste for blower coolers either. With the correct setup, your overall system will be cooler, and can be quieter as well.

my setup:
2 front intake, side intake (over GPUs), top + rear exhaust.
(2) blower (correct term) style reference cards.
PSU intake from the bottom directly out.

this effectively creates (3) "independent" airflow paths.


the bottom CPU cooler fan (huge Mugen tower cooler), might grab some excess heat off the back of the GPU, which in my case is fine. front intakes may help cool the GPUs as well, which is fine, but 80% of the GPU heat is straight out of my case.

Thanks for a great article, and the "how-to-not-fall-for-gpu-marketing-tricks" thread was really good as well.

typer.gifecho "The `uname` Club"typer.gif

Steam profile Hit me up for some fun gaming TF2, L4D2, HL2DM, Unreal 2k4
kevinf is offline  
post #8 of 9 (permalink) Old 04-06-2017, 10:01 AM - Thread Starter
Old dog, old tricks
 
ronnin426850's Avatar
 
Join Date: Sep 2009
Location: Europe
Posts: 9,738
Rep: 574 (Unique: 474)
Quote:
Originally Posted by kevinf View Post

I really don't understand people's distaste for blower coolers either. With the correct setup, your overall system will be cooler, and can be quieter as well.

my setup:
2 front intake, side intake (over GPUs), top + rear exhaust.
(2) blower (correct term) style reference cards.
PSU intake from the bottom directly out.

this effectively creates (3) "independent" airflow paths.


the bottom CPU cooler fan (huge Mugen tower cooler), might grab some excess heat off the back of the GPU, which in my case is fine. front intakes may help cool the GPUs as well, which is fine, but 80% of the GPU heat is straight out of my case.

Thanks for a great article, and the "how-to-not-fall-for-gpu-marketing-tricks" thread was really good as well.

Thank you thumb.gif
ronnin426850 is offline  
post #9 of 9 (permalink) Old 04-06-2017, 12:44 PM
New to Overclock.net
 
b0oMeR's Avatar
 
Join Date: Sep 2013
Posts: 559
Rep: 17 (Unique: 17)
Quote:
Originally Posted by ronnin426850 View Post

Thank you thumb.gif

Generally from my experiences a GPU with an open cooler is better for a case that is larger/has directional airflow.
The blower style is for when you need air to disperse from a certain direction ( through the back of the case), this is only beneficial in a smaller case like a HTPC/m-itx/atx style motherboard as there will be very limited air flow and lack of the ability to direct air out like with multiple case exhausts.

During multiGPU setups, an open air cooler will blow hot air directly onto the next card making one card a constant 10-15c hotter than the first. I have this problem with my MSI 970s, One card runs at 80 and one will run at 65. ASIC quality has part to do with this but mainly its because one card is throwing its old heated air onto the next.

If i had a blower style where both cards are blowing hot air directly out of the back, although this is a less efficient method of cooling as you lack the surface area, it is more efficient in this use case because you aren't dumping old air onto another piece of hardware that is trying to breathe in new air.

Also cases that are all open air like my ThermalTake Core P5 will be more suited for open air coolers mainly because the supply of fresh air is much better than a normal case where the fresh air supply is dependent on the intakes.
b0oMeR is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off