Air Flow vs. Water Flow (is more always better?)

V

virtualrain

Limp Gawd
Joined
Jul 31, 2005
Messages
182
I ran a simple experiment on my dual pump system, comparing delta-temperatures (air to CPU) at idle and load in four different configurations:

  1. Quiet (pumps and fans both undervolted to ~ 7V)
  2. More Air Flow (pumps undervolted but fans running at full speed)
  3. More Water Flow (fans undervolted but pumps running at full speed)
  4. Noisy / High Performance (both fans and pumps running at full speed)

The chart below reveals some interesting facts...

airflow_vs_waterflow.png


Conclusions:

As expected, running both the fans and the pumps at full speed provides lower delta-T's than running them undervolted.

Also, not surprising, is the fact that increasing fan speeds (while keeping the pumps undervolted) provides a noticable increase in cooling performance... more airflow is definitely better than less.

The interesting revelation here, at least with my configuration, is that running the pumps at full speed without additional air flow only serves to dump more heat into the loop and actually performs WORSE!

My conclusion is that just increasing water flow without a proportionate increase in air flow, may actually hurt temperatures.

It's also interesting to note that you can actually gain a lot of peace and quiet by undervolting your fans and pumps at idle while still providing more than adequate cooling. Even at load, the undervolted components may still provide enough cooling perforamnce to not affect your overclock.

You can read more about my loop and the testing process here.

I'm interested in your thoughts and whether you feel this is consistent with your own findings.
 
virtualrain said:
I ran a simple experiment on my dual pump system, The interesting revelation here, at least with my configuration, is that running the pumps at full speed without additional air flow only serves to dump more heat into the loop and actually performs WORSE!

My conclusion is that just increasing water flow without a proportionate increase in air flow, may actually hurt temperatures.

.
Click to expand...
I run a two pump/two rad system and can verify what you are saying.
The only point you should make is that you are changing flow rates by undervolting the pump. Substituting a less powerfull pump (and supposidly one with less heat dump) may not yield the same results. An Eheim 1048 will dump alot less heat than a 655 undervolted.
http://forums.procooling.com/vbb/showthread.php?t=10825&highlight=pump+heat+dump
 
Motocross said:
With less flow the water stays in the radiator longer so it has more time to cool off?
Click to expand...

It also stays in the waterblock longer and has more time to heat up. Slower flow means a larger boundary layer between the water and the copper tubing of the radiator, resulting in worse heat transfer from the water to the radiator. More flow is always better in getting heat from the water to the radiator. The issue in this testing is that when you run the pumps at full bore they themselves are putting more heat into the water. In the testers configuration the additional flow was not enough to make up for the additional heat dump of the pumps.
 
Faster flowrate is better in any system. End of story.

However, too many people help spread a myth than you NEED extreme flowrates in PC watercooling systems but in real world setups it hardly makes a difference. A Half decent pump is enough and spending more on an upgrade is a waste.

Money is better spent on buying a bigger radiator, and if you're gunna go there then do what I did and get a full sized car radiator from a local scrapyard. Dirt cheap and better than any commercial PC watercooling rad you can buy.
 
spine said:
Faster flowrate is better in any system. End of story.
Click to expand...

Not quite. In real life, faster flow rate pretty much always comes at the expense of more heat dump. Granted, most people running one "normal" pump don't really operate in a regime where heat dump is significant. In these cases, the general rule of thumb is "more flow is better".

But to answer the question as worded in the thread title (implicitly assuming all other variables constant), then more is always better (for both air and water).
 
The transfer of heat is more dependent on the flow of air. This is basic physics. Air has a much lower specific heat than water, but the volume of air is much much larger (per mol) than that of water, which is what is important.
Faster flowrate will help you, as long as there is sufficient airflow. As seen here, there was not sufficient airflow to account for the increased flowrate.
 
spine said:
Money is better spent on buying a bigger radiator, and if you're gunna go there then do what I did and get a full sized car radiator from a local scrapyard. Dirt cheap and better than any commercial PC watercooling rad you can buy.
Click to expand...
Car radiator? Quite interesting. For most people, a standard heatercore is more than enough though. With something as big as a full radiator, you want to be sure you have a powerful enough pump. Extremely low flowrates or extremely high pressure can be bad in a loop...
 
Motocross said:
With less flow the water stays in the radiator longer so it has more time to cool off?
Click to expand...
Okay, it's time for the racetrack analogy again.

The finish line is the radiator, and the car is a block of water molecules. If the car can run one lap in sixty seconds, it crosses over the finish line once per lap, and takes around one second to pass over the finish line (let's assume that it's actually a fairly big painted line). Now, double the car's speed. Does the car spend less time passing over the finish line in the same sixty seconds?

It spends the exact same amount of time passing over the finish line, because it now does so twice in the same sixty seconds, while passing over it twice as quickly.
 
In your case, your loop does not benefit from any more additional flow/pressure to make up for the (at most) additional 14 watts of heat from your pumps (assuming 2-18 watt DDCs) at idle. Under load, the additional heat from your other components makes the 14 watts more negligible, although your blocks do not benefit much from the additional flow (you would likely see similar temperatures with a single DDC @ 12v).

On the other hand, your radiator loves the extra airflow from 12v. If it's a BIP or PA/HE, I'd be somewhat surprised, but those numbers with a BIX are pretty easy to explain. If you could tell us what fans/radiator you have, that would be greatly appreciated to accompany your results.
 
phide said:
Okay, it's time for the racetrack analogy again.

The finish line is the radiator, and the car is a block of water molecules. If the car can run one lap in sixty seconds, it crosses over the finish line once per lap, and takes around one second to pass over the finish line (let's assume that it's actually a fairly big painted line). Now, double the car's speed. Does the car spend less time passing over the finish line in the same sixty seconds?

It spends the exact same amount of time passing over the finish line, because it now does so twice in the same sixty seconds, while passing over it twice as quickly.
Click to expand...

What if the car going twice as fast isn't even bothering to remove heat from the block (because it is going too fast)? There is a point in which flow doesn't matter as much as the dynamics of the block and radiator. Most of todays blocks are not as dependent on flow as before. Hence why my Fuzion block work perfectly fine in a 8mm loop with a small pump. Would it perform better with a larger pump/tubing? Yes but not enough to justify the solution I am looking for. My overclock hasn't changed since switching to smaller tubing and a smaller pump...
 
Motocross said:
That makes sense. I think mwarps hit the nail on the head though.
Click to expand...

In a way, but not how you're thinking. mwarps' post is correct when read as saying that the heat transfer step where the thermal resistance is most affected by a flowrate change is the transfer of heat from the radiator's surface to the air. Therefore increasing airflow will usually help temps more than increasing water flowrate.

Say at the pumps full speed the water stays in the radiator for 1 second. In that time it is only able to give up say 75% (not a real number just whatever) of its heat and even though its coming around again faster its also picking up heat from the block faster. So when it gets the to block is still has 25% of its heat left from the last loop.

But lets say the pumps now running at say medium speed and now stays in the radiator for 2 seconds instead of 1. This time it has enough time to give up 100% of its heat and has no heat left over when it reaches the water block. You feel me?
Click to expand...
No. (I'm really not trying to be condescending, but it's really that simple.)
Again I ain't no scientist so if theres a scientist on here feel free to prove me wrong.
Click to expand...

It's a really simple problem once you sort out the proper language to talk about the physics involved. Talking about what % of heat gets given up in an amount of time spent in [insert loop component] is just not a good way to think about things. The whole point of the racetrack analogy is to get people to stop thinking in terms of how long the water sits in any one component in a pass through the loop.

Now, there are cases where the flow velocity itself really does make a difference. In other words the racetrack analogy does break down, but only when the velocity change being talked about forces the fluid flow from one regime to another. i.e. at a very low flowrate the flow is laminar, while at a much higher flowrate it is fully turbulent. However, for most scenarios this is simply not the case.
 
synaps3 said:
Car radiator? Quite interesting. For most people, a standard heatercore is more than enough though. With something as big as a full radiator, you want to be sure you have a powerful enough pump. Extremely low flowrates or extremely high pressure can be bad in a loop...
Click to expand...

You'd be surprised how easily a weak pump can handle it. I have a 22" x 17" car radiator that splits the intake into 3 tight and very long tubes that wind back and fourth through the entire radiator (unlike a heatercore which lets the water flow easily through many channels on one pass).

Anyway, my old eheim 1046 handled it with no probs. I then moved to an eheim 1250 thinking it'd makes things loads better but it didn't. I then recently upgraded to a Laing DDC ultra and it again made little to no difference.

So I've learnt from bitter experience that paying loads on a fancy powerful pump is a waste of money. Getting that large car radiator for £10 was the best thing I ever did for my watercooling setup.

my_rad_1.jpg

Fresh!

my_rad_2.jpg

Fresh again!
 
Hahaha, Spine.

Woah.

Are those 10mm LED's on the top for the extra bling factor :D

And about pump and flowrate... I get NO temperature difference between my pump at "2" or on "5". Laing D5 FTW.
 
That's a useful test so thanks for that. I upgraded my 10w mcp350 to an 18w mcp355 (laing ddc pro to ddc ultra). Other than being slightly noiser, it didn't seem to help a lot, if at all. (Possibly because my radiator/fans are slightly underpowered compared to my system's heatload)
 
nonlnear said:
In a way, but not how you're thinking. mwarps' post is correct when read as saying that the heat transfer step where the thermal resistance is most affected by a flowrate change is the transfer of heat from the radiator's surface to the air. Therefore increasing airflow will usually help temps more than increasing water flowrate.


No. (I'm really not trying to be condescending, but it's really that simple.)


It's a really simple problem once you sort out the proper language to talk about the physics involved. Talking about what % of heat gets given up in an amount of time spent in [insert loop component] is just not a good way to think about things. The whole point of the racetrack analogy is to get people to stop thinking in terms of how long the water sits in any one component in a pass through the loop.

Now, there are cases where the flow velocity itself really does make a difference. In other words the racetrack analogy does break down, but only when the velocity change being talked about forces the fluid flow from one regime to another. i.e. at a very low flowrate the flow is laminar, while at a much higher flowrate it is fully turbulent. However, for most scenarios this is simply not the case.
Click to expand...


I think the easiest way to think about it is this :

You CPU is dumping X Watts into the water, this remains constant.

Your radiator is dissipating Y Watts into the water, this changes depending upon the difference between water and air temperature. A larger difference means more heat dissipation. (Faster airflow here keeps the air around the fins from heating up and hurting the heat dissipation)

Now the heat that the CPU is putting in remains constant. The only way the radiator dissipates more heat is if the temperature difference increases. So eventually the water will reach a temperature where while it is passing through the radiator, 100W is dissipated.

Notice that there isn't even any mention of time or speed. It is really that simple. Watts already has a time element built in. A Watt is one Joule/Second. You dont need to worry about water speed because in the big picture you are just talking about heat in versus heat out. If you want to get into the nitty gritty you can look at the inefficiencies along the way. The waterblock performance as a function of flow, the radiator performance as a function of water and air flow, etc.
 
Erasmus354 said:
I think the easiest way to think about it is this :

You CPU is dumping X Watts into the water, this remains constant.

Your radiator is dissipating Y Watts into the water, this changes depending upon the difference between water and air temperature. A larger difference means more heat dissipation. (Faster airflow here keeps the air around the fins from heating up and hurting the heat dissipation)

Now the heat that the CPU is putting in remains constant. The only way the radiator dissipates more heat is if the temperature difference increases. So eventually the water will reach a temperature where while it is passing through the radiator, 100W is dissipated.

Notice that there isn't even any mention of time or speed. It is really that simple. Watts already has a time element built in. A Watt is one Joule/Second. You dont need to worry about water speed because in the big picture you are just talking about heat in versus heat out. If you want to get into the nitty gritty you can look at the inefficiencies along the way. The waterblock performance as a function of flow, the radiator performance as a function of water and air flow, etc.
Click to expand...
I know. The cases I mentioned where the racetrack breaks down (and thus speed becomes relevant) have to do with critical Reynolds numbers that don't really come into play for 99% of watercooling scenarios. Believe me, I know what happens when you dig into the nitty gritty... ;)

And I was trying to point out to Motocross that thinking in terms of "time in the block/radiator" was the wrong way to approach the problem. I certainly wasn't advocating that line of reasoning! :p
 
You must log in or register to reply here.
Back
Top