• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Sticky Situation - Cleaning Up Water Cooling

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Nope. The odd thing is water in it's pure form is an excellent heat carrier. You can look it up on Wiki. Auto antifreeze is used to increase the freezing point and increase the boiling point of water in a car cooling system, and as anti-corrosive. It actually isn't as efficient as pure water, but the anti freeze is needed.

So distilled water and a few drops of a biocide is better than all these other fun things out there. We don't boil, we don't freeze, and we don't have any worry of corrosion in a modern loop.
 
I just want to say that I'm a complete noob to water cooling, yet I have taken a hydraulics class and I wanted to make sure I'm getting what is being said here..

Something for the Flowrate sticky. The old one has tons of great info, some applies, some is way out and does not apply to our needs. This will kinda sum it up.

Flow rate is important to the effective operation of a CPU block. Modern blocks need 1 to 1.5 GPM of flow rate to effectly remove heat. below 1 GMP you can impact the CPU. Above 1.5 the benefits drop off and flatten out on curves. Look inside any modern block. And if you can find some of the simulations posted how the water acts in the block, I have seen them in the past.

Fluid needs to be of high enough flow to get very turbulent inside the block. This causes more molecules to contact the warmer copper surfaces, thus collecting more heat to be dissipated in the rad.

I don't want to get too technical here, but the amount of head these pumps produce isn't enough to create a high pressure scenario in regards to having more molecules make contact. You're talking about compression and water, in liquid form, is as dense as it will get. I think what you're saying is you want the flow rate to be just right so that the water has time to absorb the heat. Not too fast to not absorb the heat, yet not too slow for the water to stay and act as an insulator. What I think you mean is having a good velocity profile through the whole system.

We want linear laminar? flow in the hoses, but not in the blocks. You reduce flow and you can severly impact the cooling capability of the CPU block.

Technically, the water will be turbulent through the whole loop. Pumps, blocks, and radiators will cause a turbulent flow because everything except the pump will cause a drop in energy head. So, the type of flow is not really an issue because it is not a high water pressure scenario. It boils back down to good volumetric flow rate, which is dependent on the pump. Keep in mind that flow rate never changes in a Pipes in Series, which is what water cooling is.

Another thing. Slowing the flow in the rad also reduces flow in the blocks. So as it enters the block it's cool. When it exits it's warmer than it could be, because it's flowing slower. So besides the impingment problems metioned above, you can look at it this way.

Unless you mean volumetric flow, I think you're talking about velocity. The pump controls the volumetric flow rate, not the radiator and block.

Lastly, we are dealing with a closed loop. Lets set the flow rate at 2 GPM. For easy demonstration purposes this will cause a water molecule to make a loop every 5 seconds, and stay in the rad for 1 second. So, every minute the molecule will be in the rad for 12 seconds.

Lets reduce the flow to 1 GPM. Every 10 seconds. 2 seconds. The molecule will still be in the rad for 12 seconds. Reducing flow rate does not keep the water in the rad for any longer...........................

I don't know where you're getting that data as it's nearly impossible to 'track' a molecule in that closed loop.

In summary, the purchase of a good pump and not getting crazy with two many blocks and 90 degree fittings your flow rate will be just fine. All modern rads and recommended pumps are just fine even with two GPU blocks and a CPU block. Add in a full set of Mosfet blocks etc and you might be hurting for flow rate. How do you know? Your temps are poor. It can be tested with a 10 gallon bucket. Fill use the 10 gallons as your res. Bleed the loop. Then put the pump out hose into the sink and turn it on. How long does it take the 10 gallons to get to 2 gallons? There is your systems flow rate.

All of this can be supported by tests at Martins and Skinnees and other info on the web. Someone is welcome to add all the goodies and links if they wish to make it into a true sticky.

I really want to see a few graphs from Martin & Skinnee. I really would like to see a test done on one PC with a water cooling setup of just having the CPU cooled, and see how the temperature of the CPU changes when you change the volumetric flow rate of the pump. Then another graph of another water cooling setup but this time, the GPU is also water cooled and the same comparison is done...

Otherwise, all I have to say is that by reading this and applying my knowledge I've learned in hydraulics, the key factor that determines good cooling is volumetric flow rate from a pump with a good pump pressure head.

You want it so that when the water reaches the pump on the intake side, the drop in pressure head due to ALL of the blocks, radiator, and elbows is still above atmospheric. It also prevents water from 'lingering' inside a block and start acting as an insulator rather than dissipate it.

high pump head, and the right volumetric flow will create a very efficient cooling set up.
 
I'll ignore all the parts of what you said that were correct - they are all very good and I have nothing to add. :thup:

I don't want to get too technical here, but the amount of head these pumps produce isn't enough to create a high pressure scenario in regards to having more molecules make contact. You're talking about compression and water, in liquid form, is as dense as it will get. I think what you're saying is you want the flow rate to be just right so that the water has time to absorb the heat. Not too fast to not absorb the heat, yet not too slow for the water to stay and act as an insulator. What I think you mean is having a good velocity profile through the whole system.

The only nitpick here is talking about timing being right - it's not necessary to have "time to absorb the heat". There is always water in contact with the components, so volumetric flow rate is the only factor worth talking about. Every molecule that passes picks up a portion of the energy (heat) no matter how long it spends in contact with the surface, so it doesn't matter how long any given molecule spends in one place as long as the flow rate is where it should be.

This also ties into the part about a specific water molecule and talking about seconds, it's an attempted example to convey the importance of volumetric flow rate. I don't think it does a good job at that tho.
 
Great input madhatter256, thanks. I'll address each point.

In old block designs there was simply a back and forth channel where the water lazily flowed around and the molecules if lucky were close to the copper and able to gather heat from the heatsink. The newer blocks like the DD TDX, Apogee, Swiftech GTZ etc etc have a ton of pins where each teeny .25 mm or less in diameter pin needs water next to it. Having enough pressure/flow into the top of these pins will push the water in a crazy convoluted mess so more molecules can touch the pins (Velocity good input). Yes, there is a balance. Over xx GPM doesn't increase the dispersion of the molecules much. You 'can' get an Iwaki $200 pump, but the gains, heatload from the pump, and noise aren't worth the extra cost or effort even tho your GPM through the bock will increase.

In areas where heat absorbtion isn't needed, it's best to have smooth flow. One reason Bitspower fittings have been a boon for watercooling. Take a standard 90 deg copper fitting from Home Depot. Inspect the inside with ID being equal of a BP fitting. Which will cause turbulence? Turbulence reduces flow rates. We need to keep the flow rate above xx or the blocks will not be as effective. A smooth drainage channel can carry more water per hour than a channel with large blocks in odd patterns. There is a reason for both in flood control, why not the same in a WC loop?

I like the term velocity. Good point. Thats what we need in a CPU block, velocity of water to create tubulence, like the water out of the end of a water hose with your thumb on it. These terms are so so so interelated and actually important for an engineer. Here? You can guide us and make sure we don't cross the common sense boundries.

Don't have to track a molecule, LOL, goodness. But you do understand the concept.

With your schooling and brains (you seem to have a good head on your shoulders) your welcome to do the tests, just need $$ for the equipment or a school lab. You'd be amazed what others have done in tests that are lost in some forums over millions of posts. I have seen tests done with proper flow equipment and adjusting flow rates and how it can impact temps etc. I have just read and read and melded it into basic common sense concepts. Thats what I do. Common sense watercooling.

Watercooling isn't 'roket science', just some basic understandings of flow rates and what is overdoing it with a weak pump and too many blocks. Or overdoing it the wrong way and spending too much on fat hose, monster pumps, and other things.

Your right on so many levels but we don't have to go that far, it's the concepts and understanding at the basic levels for 99% of us is what matters.

I really appreciate your input, don't take it the wrong way.
 
Last edited:
Just put this at XS, someone had a question. How can I reduce the fluff but make it into a good deep Delta T explanation?
...............................................

A delta measurement is used in many engineering terms in many disciplines. I guess it means the difference between two of the same thing.

In water-cooling it's simply the difference between the ambient air temps and a stable load on a rad. Load meaning heat. Stable meaning the load has been running long enough so the loop is stabilized, heat is made; it is removed and run long enuff for the water temps to get to the max under a load.

If your ambient are 60C and your water exiting the rad is 65C, you got a 5C DT. And it’s important you understand this simple concept. You need xx cooling for xx heat load with a resultant xx DT. It’s how you decide what size radiators you need as a minimum for the loop to perform better than air, and what’s needed for really superior max overclocks.

So for example, in a water cooled loop you generate 200 watts of heat. Your block pulls heat into the water; the heat is dissipated into the air by the rad/fans. Skinnee and Martin came up with a great chart for rad test results. Make xx heat, run xx fans, your DT is xx.

The efficiency of the rad determines the residual heat in the water as it circulates. A rad cannot remove ALL the heat. Heck if that was true, running a rad with no heat load would cause the water to go all the way down to below freezing theoretically. A great Delta T is under 5C, meaning you got a big rad for your heat load. Medium DT is 10C, and 15C is getting bad. CPUs need lower DT than a GPU loop.

Ultimately lucky folks with cooler temps year round can go with lesser rads. People with high ambient might need bigger rads for the same final core temp on a CPU that someone in Norway vs. Samoa can get.

Hope that helps a simple explanation, I'm not a Thermodynamic engineer.
 
i think we should start up a thread on benching delta's :) it's something i can be near the top in for once... i can't believe i'm still in the top 100 (98th) for 3dmark06 in the benching forum.

good explanation btw, simple and short enough to not scare me away.
 
Back