Some fans are less effective in pull configuration versus push. The difference in temps would likely be only a few degrees either way though.

 

Push is better. But fans pulling or pushing cool air into the rad is ideal. Not having the hot air from the case going through the rad and out.

 

Quote:

Originally Posted by stillsurfn

I have a general question.....

Is it more efficient an effective for cooling performance to have a fan push air into a radiator or pull air out of a radiator ?

This is assuming the fan being used is a static pressure fan, since those are supposed to have a more focused flow for restricted air flow situations, like a radiator.

But, the question is independent of where it the radiator is located or any other heat related issues.

Anybody know ?

While what MNMadman said might be true in rare cases, fan airflow is basically the result of the pressure differential between intake and exhaust sides of impeller, .. and as such impeller has no idea if radiator is in front or behind it. But .. and it's a big 'BUT' .. the exhaust side of a fan has impellers 4-5mm farther away from fan mounting surface (fan housing) then on intake side. This is because exhaust side of fan has the framework from surrounding housing to center where motor is. This 4-5mm difference can mean fan will be much noisier and sometimes does is not able to pull air into itself as well and it can push it out. This is true regardless of where fans being mounted and used .. be it on case vents, cooler or radiator.

As for the difference in temperature of air coming out of radiator and into a fan versus being a few degrees cooler going through fan before going into radiator, it's not enough of a temperature difference to make much if any difference to fan or fan bearing life.
Quote:

Originally Posted by MrTOOSHORT

Push is better. But fans pulling or pushing cool air into the rad is ideal. Not having the hot air from the case going through the rad and out.

So you suggest using pre-heated air coming out of radiator to cool components inside of case instead of cool air coming in? Radiators are effected less by the temperature of air going into them than air cooled components are, therefore it's better to have air a few degrees warmer feed a radiator instead of air cooled components.

 

If you're doing low speed fans pull is usually better, high speed favors pushing. But saying that it's a tiny amount of difference.
You shouldn't have any issues and the best setup is whatever works in the build normally.

Some fans like eloops have been bad for pull setups as they catch on the radiator in some cases though.

 

The best thing about a pull configuration is that the dust tends to accumulate on the accessible side of the radiator. On a push configuration dust builds up between the radiator and fan, making it a little more difficult to clean.

 

Quote:

Originally Posted by doyll

...fan airflow is basically the result of the pressure differential between intake and exhaust sides of impeller, .. and as such impeller has no idea if radiator is in front or behind it. But .. and it's a big 'BUT' .. the exhaust side of a fan has impellers 4-5mm farther away from fan mounting surface (fan housing) then on intake side. This is because exhaust side of fan has the framework from surrounding housing to center where motor is. This 4-5mm difference can mean fan will be much noiser and sometimes does is not able to pull air into itself as well and it can push it out.

Good explanation. So, like I said, some fans are less effective at pull configs. The CPU/GPU temperature difference as a result would only be a small one -- a few degrees maybe.

 

Thanks guys,

The reason I'm asking is that I have a somewhat unique situation.

I don't use a computer case, so I have no "heat buildup" or GPU heat transfer issues to the CPU area.

I don't worry about case intake or exhaust because there is no case.

I do use a noctua d15 and one thing that i've always wondered about it the "stock" configuration of the d15 with the center fan.

In stock configuration, one of the radiator fin stacks is always going to be using hot exhaust air from the other fin stack. One stack gets fresh air .... the other gets heated air.

So, I was wondering if it would be better to mount the 2 fans on the outside of the fin stacks, with no fan in the center.

Then the choice is wether to push air into each the fin stacks or pull air out of the fin stacks......

Any thoughts and opinions ??

 

Quote:

Originally Posted by scientist

The best thing about a pull configuration is that the dust tends to accumulate on the accessible side of the radiator. On a push configuration dust builds up between the radiator and fan, making it a little more difficult to clean.

This is the reason I run my fans in pull config. On my Thermaltake Core P5, the fans pull the air through a dust filter first so the rad doesn't even get dusty.

 

Quote:

Originally Posted by doyll

So you suggest using pre-heated air coming out of radiator to cool components inside of case instead of cool air coming in? Radiators are effected less by the temperature of air going into them than air cooled components are, therefore it's better to have air a few degrees warmer feed a radiator instead of air cooled components.

I suggest taking cool air from outside of the case into the rad then going into the case then out the exhaust.

 

Quote:

Originally Posted by stillsurfn

I have a general question.....

Is it more efficient an effective for cooling performance to have a fan push air into a radiator or pull air out of a radiator ?

This is assuming the fan being used is a static pressure fan, since those are supposed to have a more focused flow for restricted air flow situations, like a radiator.

But, the question is independent of where it the radiator is located or any other heat related issues.

Anybody know ?

Hi,
If a radiator is on the front of a rig it's easy to do a push pull on the rad being as an intake because one would already have places for case fans there.

On the top push pull is sometimes tough without covering the top of the board but again always as a intake would be best.
On top if I couldn't do a push pull deal I'd put the fans to the case and the rad under so it's a push.

 

Quote:

Originally Posted by MNMadman

Good explanation. So, like I said, some fans are less effective at pull configs. The CPU/GPU temperature difference as a result would only be a small one -- a few degrees maybe.

Not even that much. The spacing difference results in noise differences, but like I said, the pressure differential between intake and exhaust sides of fan is what makes fans work, and fans have no way of knowing which of their sides is mounted to a surface.
Quote:

Originally Posted by MrTOOSHORT

I suggest taking cool air from outside of the case into the rad then going into the case then out the exhaust.

And how are you suggesting to keep the heated air coming out of radiator from mixing into other airflow inside of case warming it up before it cooler other components? Because that is what happens, .. the heated air coming out of radiator into case mixes with cool in the case and warms it up.

 

Quote:

Originally Posted by doyll

Radiators are effected less by the temperature of air going into them than air cooled components are, therefore it's better to have air a few degrees warmer feed a radiator instead of air cooled components.

Even if this is the case, most custom WC setups are going to be cooling the only components significantly temperature constrained limited (CPU and GPU), and the inside of the case being several degrees warmer is of little concern.

Knocking ~5C off peak CPU and GPU temps is easily worth everything else being ~5C warmer.

 

Quote:

Originally Posted by Blameless

Even if this is the case, most custom WC setups are going to be cooling the only components significantly temperature constrained limited (CPU and GPU), and the inside of the case being several degrees warmer is of little concern.

Knocking ~5C off peak CPU and GPU temps is easily worth everything else being ~5C warmer.

Did you notice you are posting all of these excuses in the air cooled forum?

Besides, CPU and GPU are only that, CPU and GPU, and being this is 'Air Cooled' forum we are not using custom loops like in the 'Water Cooled' forum, but even there the HDDs, SSDs, other motherboard components, RAM, etc. are all air cooled. Experience with air cooling has proven a 1-2c increase in air temp inside the case usually results in a 1-1+c increase in component temps being monitored. When I was water cooling a 4-6c increase in air temp resulted in a 1-2c increase in component temps. To me that is a big difference

You are not going to knowk 5c off of CPU and GPU temps by using radiators as intake vs exhaust. If you can I would like to see relevant data showing what the system was as well as what all other relevant component temps were with radiators as exhaust versus intake.

 

delete

 

Quote:

Originally Posted by scientist

The best thing about a pull configuration is that the dust tends to accumulate on the accessible side of the radiator. On a push configuration dust builds up between the radiator and fan, making it a little more difficult to clean.

As was already noted, best thing is to have filtered intake so so no dust accumulates on radiator or fans, but only on a filter that is easily removed and cleaned.

 

Quote:

Originally Posted by doyll

Did you notice you are posting all of these excuses in the air cooled forum?

No, but that's ok as it applies to air cooled setups too.
Quote:

Originally Posted by doyll

there the HDDs, SSDs, other motherboard components, RAM, etc. are all air cooled.

All of which tend to have enormous temperature margins, in practice, relative to the CPU and GPU.

Another 10C on my HDDs or SSDs wouldn't hurt them. My memory would still be in spec if it were 30C warmer than it typically reaches under heavy load. Chipsets and VRMs, with a few exceptions, generally have considerable margin to work with.

Prioritizing the CPU and GPU, even at the expense of other component temperatures, is going to be a sound strategy far more often than not.
Quote:

Originally Posted by doyll

You are not going to knowk 5c off of CPU and GPU temps by using radiators as intake vs exhaust. If you can I would like to see relevant data showing what the system was as well as what all other relevant component temps were with radiators as exhaust versus intake.

Even reasonably well ventilated cases can have large temperature deltas under load (the air coming out of my primary system right now is at least 15C warmer than the air going in) and I've certainly had cases where switching a radiator from exhaust to intake, or switching the fans nearest to an air cooler, saved more than 5C on the components being cooled.

Couple of examples of older setups of mine where I improved temps significantly, by doing my best to put as much cool external air through and over critical components as possible:

https://d1rktuf34l9h2g.cloudfront.net/6/6a/6a103131_IMG_1533.jpeg

https://d1rktuf34l9h2g.cloudfront.net/7/78/78d91697_DSCN0754.jpeg

In the first image, which is a 4.3GHz i7 970 and a 918MHz GTX 480, I have both radiators as intakes. Air flow from the CPU radiator also blows through the VRM heatsink, while flow from the fan on the GPU radiator blows directly on the VRM/memory cooling plate of the GPU. I tested all sorts of different orientations, and both the CPU and GPU core, were a few C cooler this way than any exhaust setup, while the board VRM and GPU VRMs were both closer to 10-15C cooler...the air blowing across their heatsinks was warmer, but there was vastly more air flow than otherwise, which more than made up for the difference. The drives and southbridge were somewhat warmer, but the mechanical HDDs were still colder than ideal, and the SB never ran hot. The IOH temps were about the same either way.

The second setup (3930K @ 4.3GHz, and a pair of 7950s @ 1080/1500) is all air cooled, obviously. Again, I tried may cooling layouts, eventually settling on no exhaust fans at all, and six intakes. Cool air was coming in from all sides except the rear and being forced through and around all critical components. Both the CPU and GPUs were more than 5C cooler this way when the system was under heavy load than with a more conventional fan setup featuring three exhausts (top and rear), and three intakes (two front, one side panel). With everything intake, no air flow from the GPUs was being ingested by the CPU cooler, and as much air was going over the backs of the GPUs (and out the open I/O panels) as through their heatsinks.

 

Quote:

Originally Posted by Blameless

No, but that's ok as it applies to air cooled setups too.
All of which tend to have enormous temperature margins, in practice, relative to the CPU and GPU.

Another 10C on my HDDs or SSDs wouldn't hurt them. My memory would still be in spec if it were 30C warmer than it typically reaches under heavy load. Chipsets and VRMs, with a few exceptions, generally have considerable margin to work with.

Prioritizing the CPU and GPU, even at the expense of other component temperatures, is going to be a sound strategy far more often than not.
Even reasonably well ventilated cases can have large temperature deltas under load (the air coming out of my primary system right now is at least 15C warmer than the air going in) and I've certainly had cases where switching a radiator from exhaust to intake, or switching the fans nearest to an air cooler, saved more than 5C on the components being cooled.

Couple of examples of older setups of mine where I improved temps significantly, by doing my best to put as much cool external air through and over critical components as possible:

https://d1rktuf34l9h2g.cloudfront.net/6/6a/6a103131_IMG_1533.jpeg

https://d1rktuf34l9h2g.cloudfront.net/7/78/78d91697_DSCN0754.jpeg

In the first image, which is a 4.3GHz i7 970 and a 918MHz GTX 480, I have both radiators as intakes. Air flow from the CPU radiator also blows through the VRM heatsink, while flow from the fan on the GPU radiator blows directly on the VRM/memory cooling plate of the GPU. I tested all sorts of different orientations, and both the CPU and GPU core, were a few C cooler this way than any exhaust setup, while the board VRM and GPU VRMs were both closer to 10-15C cooler...the air blowing across their heatsinks was warmer, but there was vastly more air flow than otherwise, which more than made up for the difference. The drives and southbridge were somewhat warmer, but the mechanical HDDs were still colder than ideal, and the SB never ran hot. The IOH temps were about the same either way.

The second setup is all air cooled, obviously. Again, I tried may cooling layouts, eventually settling on no exhaust fans at all, and six intakes. Cool air was coming in from all sides except the rear and being forced through and around all critical components. Both the CPU and GPUs were more than 5C cooler this way when the system was under heavy load than with a more conventional fan setup featuring three exhausts (top and rear), and three intakes (two front, one side panel). With everything intake, no air flow from the GPUs was being ingested by the CPU cooler, and as much air was going over the backs of the GPUs (and out the open I/O panels) as through their heatsinks.

True, while there temperature margins are relative, and to me raising the air temp inside my case 10c is unexceptionable.

To me prioritizing the CPU and GPU at even at the expense of other component temperatures is definitely not a sound strategy.

If case airflow is setup properly component intake air temperature are only marginally warmer going into components at full load than at idle .. like no difference at idle and 3-5c when all components are being stress testing .. but I will agree if not setup properly they can be 15c or even 20c or even 25c warmer than room ambient.

While changing the fans helped you it may not help others. Obviously you are free to do as you please, but I'm definitely not going to be following your lead.

 

Here's a more recent, mixed cooling setup: https://d1rktuf34l9h2g.cloudfront.net/6/6a/6a5307ff_20161123_075814.jpeg

Also, the tower's power consumption, at the wall, during testing: https://d1rktuf34l9h2g.cloudfront.net/0/0f/0fa2244c_20161123_075625.jpeg

Using the AIO radiator as an intake there shaved double digit temps off the CPU at full load because the GPUs (twin OCed 290Xes) were dumping more than 700w of heat into the case right above the only place I could place the AIO as an exhaust.

GPU and system temps were only mildly impacted.
Quote:

Originally Posted by doyll

To me prioritizing the CPU and GPU at even at the expense of other component temperatures is definitely not a sound strategy.

If case airflow is setup properly component intake air temperature are only marginally warmer going into components at full load than at idle .. like no difference at idle and 3-5c when all components are being stress testing .. but I will agree if not setup properly they can be 15c or even 20c or even 25c warmer than room ambient.

Properly is relative to the amount of heat the components are generating.

At maximum load all of the system I posted were actually overloading the PSUs they were equipped with. The air flow needed to remove 700-1100w of heat from a box of these sizes with a sub 5C temperature delta is impractical if you don't want the system to sound like a tornado. I'd rather put the cold air right where it's needed than have to double or triple case air flow to get the same temperatures for the only components that are coming anywhere near their limits.

 

Quote:

Originally Posted by Blameless

Here's a more recent, mixed cooling setup: https://d1rktuf34l9h2g.cloudfront.net/6/6a/6a5307ff_20161123_075814.jpeg

Also, the tower's power consumption, at the wall, during testing: https://d1rktuf34l9h2g.cloudfront.net/0/0f/0fa2244c_20161123_075625.jpeg

Using the AIO radiator as an intake there shaved double digit temps off the CPU at full load because the GPUs (twin OCed 290Xes) were dumping more than 700w of heat into the case right above the only place I could place the AIO as an exhaust.

GPU and system temps were only mildly impacted.
Properly is relative to the amount of heat the components are generating.

At maximum load all of the system I posted were actually overloading the PSUs they were equipped with. The air flow needed to remove 700-1100w of heat from a box of these sizes with a sub 5C temperature delta is impractical if you don't want the system to sound like a tornado. I'd rather put the cold air right where it's needed than have to double or triple case air flow to get the same temperatures for the only components that are coming anywhere near their limits.

I'm sorry, but your statements of 'at maximum load .... systems were ... overloading PSU', 700-1100w of heat' and 'sub 5c temp delta' seen the thing science fiction is made of .. and make it impossible for me to believe.

 

Depends on RPM & radiator restrictiveness but push is generally better

http://www.xtremesystems.org/forums/showthread.php?278506-Martinsliquidlab-org-Radiator-Fan-Shroud-Thermal-Testing-V2 (ror https://martinsliquidlab.wordpress.com/2012/01/15/radiator-shroud-testing-v2/7/)


Quote:

In general shrouding seems to do very little. I can measure very minor gains in pull, but push alone is better than pull plus shroud.

Push is better than Pull in both high speed and low speed conditions

Push + Pull is a good gain over a single fan.

Quote:

Originally Posted by http://martinsliquidlab.petrastech.com/Radiator-Fan-Orientation-And-Shroud-Testing-Review.html
Push Vs Pull - This depends on fan speed/power. The high speed fans at 2000RPM with a 38mm fan thickness provided the best performance in a push condition. The slow speed fans with 1350RPM with a 25mm fan thickness provided the best performance in a pull condition. I would estimate that performance line is likely to cross in the 1500-1700RPM range where they are equal. So.... slow speed = pull, high speed = push, medium speed = it doesn't really matter.

http://www.xtremerigs.net/2015/02/20/pushpull-radiator-setups-learned-stop-worry-love-fans-less-noise/

 

Quote:

Originally Posted by AlphaC

Depends on RPM & radiator restrictiveness but push is generally better

http://www.xtremesystems.org/forums/showthread.php?278506-Martinsliquidlab-org-Radiator-Fan-Shroud-Thermal-Testing-V2 (ror https://martinsliquidlab.wordpress.com/2012/01/15/radiator-shroud-testing-v2/7/)

Spoiler: Warning: Spoiler!

http://www.xtremerigs.net/2015/02/20/pushpull-radiator-setups-learned-stop-worry-love-fans-less-noise/

So according to above push/pull is best and shrouds are radiator specific when testing with Yate Loon D12SM12 fans.
While Martin's testing is interesting read, how is this relevant to thread topic of 'push or pull air across radiator fins?'

 

Quote:

Originally Posted by doyll

I'm sorry, but your statements of 'at maximum load .... systems were ... overloading PSU', 700-1100w of heat' and 'sub 5c temp delta' seen the thing science fiction is made of .. and make it impossible for me to believe.

I'm not sure what you find unbelievable.

That worst case scenario loads with OCed components produce a lot of heat, that ~1200w from the wall means a 1000w rated PSU that is ~88% efficient is overloaded, or that removing ~1200w of heat from a Fractal Design Define R4 with the sort of temperature delta you mentioned is difficult to do without being very loud.

Perhaps you don't actually use your systems, or don't care ensure they'll work in worst case scenarios, but I do, and I expect mine to.

The best way to cool a collection of parts as quietly as possible, with what's on hand, is to focus on supplying cool air to what actually needs to be cooled, and to not fuss over the temperatures of the rest.

 

Quote:

Originally Posted by Blameless

I'm not sure what you find unbelievable.

That worst case scenario loads with OCed components produce a lot of heat, that ~1200w from the wall means a 1000w rated PSU that is ~88% efficient is overloaded, or that removing ~1200w of heat from a Fractal Design Define R4 with the sort of temperature delta you mentioned is difficult to do without being very loud.

Perhaps you don't actually use your systems, or don't care ensure they'll work in worst case scenarios, but I do, and I expect mine to.

The best way to cool a collection of parts as quietly as possible, with what's on hand, is to focus on supplying cool air to what actually needs to be cooled, and to not fuss over the temperatures of the rest.

Now that you explain your previously abbreviated post I now understand what you were trying to say, so yes, you are making some real heat there. 1200w on meter inline from mains might be 1000w of heat .. and which is everything including PSU and other fans, and I'm not sure a 1000w rated PSU at near it's limit is actually supplying 1000watts unless new, and now much of that wattage is getting to CPU and GPU and then dissipated through their coolers is also open to discussion. I don't know of any air coolers rated for more then 360TDP, so I am skeptical of a computer having1000w power consumption in home use. While possible that is extreme overclocking of a CPU and at least 2x GPUs .. to me it is extreme overclocking to when load is 300-400w on each CPU or/or GPU.

Out of curiosity, how long has this system been running with this kind of OC?

I do use my system at full load several times a week, but not every day. At full load they make about 35-37dB @ 1 meter, so not exactly quiet, but not loud either. My old i7 920 is OC'ed to 4.0GHz under TRUE Spirit 140 Power is right now running at 74c @100% with TY-143 fan @ 1250rpm in 21c room.

 

The PSU in the last system is a Corsair RM1000x that is ~88% efficient at maximum load. Pulling ~1200w the wall, the PSU is delivering a bit over 1050w DC, essentially all of which is turned into heat in side the case by the components consuming it. The PSU's waste heat is partially isolated by having separate intake/exhaust, but the whole chassis of the PSU gets quite warm and adds at least some of the 150w or so it's dissipating to the 1050w that needs to be removed from the case.

The PSU will supply whatever current is demanded of it, or it will shut off; either because OCP was tripped, the PSU failed outright, or the power it was supplying became too dirty/droopy for the system to not crash. Most quality PSUs can easily supply 10-15% more than they are rated for, at least when new, but I have killed several in short order by underestimating what worst case scenario loads would actually do. That RM1000x still works fine, as it was barely overloaded, and wasn't used too long that way before being moved to a less powerhungry system, but the Antec 750 TP, and two of those Seasonic SS-860XP2s didn't survive.

The above system was running 24/7 and any time I wasn't actively using it it was a combination miner, folder, and seed box. Running dual miners (where two separate algorithims are excuted concurrently on the same hardware to maximize gross hashrate, at the cost of extra load/power consumption) on the GPUs, Folding on the CPU, while several HDDs were active, kept the system averaging about ~900w at the wall. The 1200w test was a worst case scenario stress test where I was running eleven threads of Prime95 and full screen Furmark with all power limiters disabled.

One of the 290Xes eventually died, so I moved the remaining 290X to a dedicated mining box and installed a 1080ti in this. That Nepton 280L, which performed really well for a while, started loosing coolant to the point air bubbles were circulating in the loop. The orientation I had to install it in left very little margin for coolant loss.

Currently, I've backed the CPU off to 4.2GHz, which is the best it will manage with the NH-D15S on it now. I have one of my top fans as intake to get cold air to the front of the heatsink, otherwise my AORUS 1080Ti, which exhausts entirely inside the case, adds dramatically to CPU temps when it's heavily loaded.

 

Quote:

Originally Posted by Blameless

The PSU in the last system is a Corsair RM1000x that is ~88% efficient at maximum load. Pulling ~1200w the wall, the PSU is delivering a bit over 1050w DC, essentially all of which is turned into heat in side the case by the components consuming it. The PSU's waste heat is partially isolated by having separate intake/exhaust, but the whole chassis of the PSU gets quite warm and adds at least some of the 150w or so it's dissipating to the 1050w that needs to be removed from the case.

Spoiler: Warning: Spoiler!

The PSU will supply whatever current is demanded of it, or it will shut off; either because OCP was tripped, the PSU failed outright, or the power it was supplying became too dirty/droopy for the system to not crash. Most quality PSUs can easily supply 10-15% more than they are rated for, at least when new, but I have killed several in short order by underestimating what worst case scenario loads would actually do. That RM1000x still works fine, as it was barely overloaded, and wasn't used too long that way before being moved to a less powerhungry system, but the Antec 750 TP, and two of those Seasonic SS-860XP2s didn't survive.

The above system was running 24/7 and any time I wasn't actively using it it was a combination miner, folder, and seed box. Running dual miners (where two separate algorithims are excuted concurrently on the same hardware to maximize gross hashrate, at the cost of extra load/power consumption) on the GPUs, Folding on the CPU, while several HDDs were active, kept the system averaging about ~900w at the wall. The 1200w test was a worst case scenario stress test where I was running eleven threads of Prime95 and full screen Furmark with all power limiters disabled.

One of the 290Xes eventually died, so I moved the remaining 290X to a dedicated mining box and installed a 1080ti in this. That Nepton 280L, which performed really well for a while, started loosing coolant to the point air bubbles were circulating in the loop. The orientation I had to install it in left very little margin for coolant loss.

Currently, I've backed the CPU off to 4.2GHz, which is the best it will manage with the NH-D15S on it now. I have one of my top fans as intake to get cold air to the front of the heatsink, otherwise my AORUS 1080Ti, which exhausts entirely inside the case, adds dramatically to CPU temps when it's heavily loaded.

Thanks for details. Impressive system, but I woudn't want to pay your utility bill. Here running 700 watts 24/7 would cost about £80 a month.