Optimal rad setup: pressure

[bigben2k]

Following this thread:
http://forums.procooling.com/vbb/sho...&threadid=3587

Let's talk about pressure!

Why is a high pressure of coolant good for a waterblock?
Why is a low pressure of coolant good for the rad?

Here's hoping that the following individuals will join: pHaestus, gmat, gone_fishin, JimS, Sirpent, redleader, schoolie, jtroutma, and MeltMan.

This starts with gmat's statement:
Now, it goes without question that higher flow rate means better heat transfer. Why ?
* Increased turbulence
* Higher contact rate between water molecules and rad walls, less "dead spots"
* Better temp gradient through the *whole* water circuit because it's "averaged"
* My fluid dynamic courses are so many years behind i can remember everything
Since putting rads in parallel will divide flow... No good.

gone_fishin has also provided some good info.

Here's a recap:
Facts:
1-Ambiant air is at (around) 20C.
2-Coolant temp is increased by the power emitted by the CPU.
3-Coolant is cooled by transmiting its heat to the metal that composes the radiator.
4-The rad cannot lower the temp of the coolant below the temp of the ambiant air.
5-The heat from the coolant is transmitted to the metal of the rad at the same rate, regardless of the rate of movement of the coolant (given the same temp).
6-The rad lowers its metal temp to the ambiant air.
7-The fan helps the above purpose.
8-The rad will dissipate more heat from the coolant, if the coolant is hotter.

Anyone?

[Cova]

The amount of heat a rad can dissipate is largely based on the temp of the coolant in the rad. Even my little BI Prime could dissipate a few hundred watts of heat easily with a single fan, if I could keep the coolant temp at around 90C without frying the CPU. The bigger the temp delta between the coolant and ambient, the more efficient the rad will work.

However - as rads get less efficient (lower coolant temp), the WB becomes more efficient (bigger temp delta between CPU and coolant). Because the rad's we use (if they were running efficiently with hot coolant) are way overkill for the number of watts of energy we need to dissipate, I've always thought it best to make the block run as efficiently as possible - this means the most flow-rate (pressure, whatever) that you can get through it - and the coolant will heat up until it finds the balance where the rad becomes efficient enough to dissipate the heat produced by the CPU.


Many people argue about the water moving slowly through the rad to give it time to give off more heat. I'm not worried about the amount of heat that the rad can absorb from that particular molecule of water (or drop, or whatever unit of volume you want to measure) passes through the rad. I worry about the average temp of all coolant in the rad - the lower it is the less heat the rad can give off to the air - but it means cooler water is going to the CPU, and cooler water is coming from the CPU. Again - I don't care how much heat 1 unit of water passing through the WB can pick up on it's way through - I just to keep as cold of water as possible inside the block at all times.

[bigben2k]

and of course Cova! (Welcome)

I see what you mean. The flow rate to the CPU is the most critical issue. It also happens that the flow rate of a rig will depend a lot on the CPU wb, because it is usually the most restrictive.

The rad should perform almost the same, regardless of the flow rate.

However, gmat seems to thing that rads in parallel is bad... It seems to me that it doesn't matter what the flow rate is through each rad (for heat dissipation purposes), but it does matter for flow rate purposes. So parallel good? gmat?

[gmat]

note: flow rate is *not* pressure.
Actually (setting aside parallel branches) the flow is constant and the same in just any point of the circuit.

The pressure is a factor of several things:
* pipe diameter
* position in the circuit. It's another flame-war topic but on a pure fluid dynamics point of view, the WB should be just after the pump (high pressure) and the rad just before the inlet (low pressure).
Theres a point in the circuit where pressure switches from positive to negative. (not: thats *always* the case even with our poor centrifugal pumps) you'll want that point to be *before* your rad and *after* your waterblock(s).
For more info on that there's a very good article at overclockers.com.

[gmat]

Originally posted by bigben2k
flow rate of a rig will depend a lot on the CPU wb, because it is usually the most restrictive.

Beware. Flow rate depends on the *sum* of every flow resistance. Not just waterblock.

However, gmat seems to thing that rads in parallel is bad...[/b]

No really my pinkie finger tells me thats a good thing (see other thread on this forum). Basically it's because the 2 rads are less restrictive in parallel thus improving overall flow rate.
So the fact flow gets divided between rads is compensated by better overall flow.

(edit) ah and as i said before, one needs to test it out before building / modding some metal...

[jtroutma]

This could be off the topic however I just had a small brainfart! (yes you read that correctly )

In essence what we are talking about here is contact time...... or surface areato be more elact. In order to get rid of the heat our CPUs are putting off, we need to have:
A) Large surface area in order to move the heat from a small object to a large object so that the surrounding air can remove the heat.
B) Air Flow to accelerate the heat disapation from the large surface area to surrounding air.

In short, we are trying to MOVE the heat away from our CPU dies (which is tiny!) to a place where we can have a very LARGE surface area as well as a large airflow.

With that said, where do we want all that heat in our system? On the CPU die? HELL NO! In the Radiator? OF CORUSE!! We want as little heat around our CPU as possible and we want to move all that heat (or as much of it as we can) to the radiator to get rid of it. Now wouldnt it make sence to have the heat removed from the CPU as QUICKLY as possible and placed in a location where it can be "dumped" for as LONG as possible?
BTW Radiators tend to ack as large heat storage area due to the volume of water being inside them.

If we have a very high flowrate through the entire system, the CPU will get rid of its heat fast but the radaitor will not have very long to get rid of it (so it goes right back to the CPU and gets warmer IF we have a low flowrate, the radiator has time to remove the heat but the CPU cant get rid of the massive amounts of heat it is making fast enough (hence high temps).


OK I think I will just lay that out on the table and let people chew on it (or me ) for a while.

BTW BigBen you are almost inviting a riot with this thread!!
This WILL be FUN!!!!!!

[gmat]

Originally posted by bigben2k

5-The heat from the coolant is transmitted to the metal of the rad at the same rate, regardless of the rate of movement of the coolant (given the same temp).

Beeep !
Heat transfer is a factor of dS/dt (S= surface) or if you prefer it's a direct factor of Surface Area... When flow rate increases, surface area does as well...
So given the same temp, heat is transmitted from coolant to metal at a faster rate when flow is higher.

[Cova]

hehe, I seem to be building a reputation in a hurry here But whatever keeps me busy at work is a good thing, and I haven't gotten flamed too bad yet.

As I seem to keep doing, I'm gonna compare the water-cooling "circuit" to an electrical circuit again. And I'm gonna re-state what gmat just did - parallel rads is a good thing - in an electrical circuit it would be like parallel resistors (rads resist current flow). This would mean that the overall resistance of the water-cooling circuit would be less, which would mean more flow-rate. My post above just explained why IMHO you can never have too much flow-rate - and I'd think that potentially the extra flow-rate from parallel rads could make more difference in CPU core temp (which is what we really are concerned about in the end) than the extra cooling surface of the rads.

Parallel rads = more water at nearer to ambient temp through the CPU block. Except that it takes up a lot of space in the case, I can't see anything bad about it.

As for the whole pressure thing - I've got proof in my own cooling system that our little pumps do make enough pressure to have very different pressure-points in various places in the circuit. In fact - my cooling setup may contain the proof that the BI Prime is actually more restrictive to water flow than the Maze-2. In my system, going from pump->cpu->rad->pump (no res - sealed air-tight) I actually had to re-inforce my tubing from the rad to the pump. I'm using thin-wall Tygon which is VERY easy to stretch/crush, and the suction of the pump actually sucks the tygon closed from the inside on that short stretch of tubing. The lines from the pump->block and block->rad don't exhibit any flattening of the tubing at all. And this is with only a ViaAqua 1300 (400gph) pump, though the thin-wall tubing is what makes this pressure easily visible.

[chazz469]

Originally posted by bigben2k
... The rad should perform almost the same, regardless of the flow rate.

However, gmat seems to think that rads in parallel is bad... It seems to me that it doesn't matter what the flow rate is through each rad (for heat dissipation purposes), but it does matter for flow rate purposes. So parallel good? gmat?

I think that duals rads would be a redundancy, and not necessarily the most efficient. As Cova pointed out, the rads most of us use are able to dissipate more heat than we put out. So splitting the flow to go to two rads would decrease flow rate, which according to the above statements, flow rate is most important.

I think that the idea(s) that low pressure is good for rads and high for the wb came from a number of things. But basically, the idea that pressure is directly related to flow rate. It's not. Pressure is related to mass or volume, and flow rate to velocity (if I understand it right). So in this case, pressure is a misnomer for flow rate. Now to rationalize it, slower flow rate in the rad gives more time for the water to cool, and faster flow rate in the WB removes the heated coolant faster from the WB replacing it with cooler water. Once the system has reached equilibrium, the rad changes roles. It goes from cooling the water to maintaining the temperature level of the entire system (yes this is still cooling, but not in the same sense). So, now the important thing is to get the cooler coolant to the WB and move the hot coolant out of there, needing flow rate.

[Cova]

Originally posted by jtroutma
If we have a very high flowrate through the entire system, the CPU will get rid of its heat fast but the radaitor will not have very long to get rid of it (so it goes right back to the CPU and gets warmer IF we have a low flowrate, the radiator has time to remove the heat but the CPU cant get rid of the massive amounts of heat it is making fast enough (hence high temps).

Ahh - I was just waiting for someone to post this old argument up.

I was going to say flowrate doesn't effect how fast the CPU cools off - but in light of gmats post of "... When flow rate increases, surface area does as well..." I'll instead say that flow-rate has a relatively low impact on cpu temp relative to the temp delta between the coolant and the WB. So your first sentence is invalid = the CPU will get rid of it's heat fast IF the coolant is at a low temp. If that coolant is also moving fast, then it will make the CPU even cooler. However - because it is moving fast, each unit of water (again, molecule, drop, gallon, whatever) that passes through the block will absorb less energy and cooler water will be entering the rad.

If cool water is entering the rad - the rad will be inefficient (see my first post for explanation) The water will be not much cooler by the time it exits the rad again. Though this is not a problem, so long as it is cool coolant returning to the CPU. As the system runs for a while the coolant temp will rise to a balance where the coolant entering the rad is hot enough for the rad to be efficient enough to dissipate as much heat as the CPU is putting off. This is much more dependant on the average temp of the coolant in the rad than the length of time the coolant spends in the rad.

So - the faster the water is moving, the less energy each unit of water will pick up on a lap around the circuit, and the less energy it has to dissipate as it passes through the rad. The rad is a LOT bigger, and has a LOT more surface area than the block, so that unit of coolant should have no problem releasing it's energy into the rad in it's time through there, as compared to the small amount of energy it absorbed in the very short time it was inside the block.

[gmat]

Ah Cova thanx for reminding me the electrical analogy (which is perfectly valid until you reach turbulence problems).
Power dissipated through a resistor is P=RI2 (2=square).
R = resistance (fluid analogy = flowresistance)
I = current (fluid analogy = flow rate)
If you put resistors in // you split the current in 2 (knots law)
Draw your own conclusions

Now i dont say its a bad thing, like stated before one must consider the whole circuit. And less flow restriction is a good thing.

Chazz: flow rate is *not* velocity. Its a product of velocity times cross section.
And pressure is *not* flow rate.
Pressure changes greatly through your circuit.
Flow rate (in a closed loop) is THE SAME at ANY point.

(edit) hmm my analogy wasnt very clear. P in fluids will be your heat transfer factor.

[Cova]

Actually - in all my analogies...

Resistance = resistance to flow, makes water want to stop moving.
Current = flow-rate as measured in volume / time, not speed.
Voltage = pressure - a pump would be a battery

And power is something I'm not going to touch...

rads/resistors in parallel = 1/2 the current/flow-rate (assuming the same rads) through those pieces - flow-rate is not affected in the other parts of the circuit. But 1/2 the overall resistance - this does effect (increase) the flow-rate in the rest of the circuit.

So - it gives the slow flow in the rads that so many people on here love (though I personally think it doesn't help much - our rads are more than capable of the heat of a CPU) and higher flow-rate through the block, which is where I think the biggest advantage of dual-rads is.

Chazz - you must have written that while I wrote my second post. as I just said dual rads increase flow to every part of the system except the rads themselves (which get 1/2 flow each).

Edit: Now that you edited your analogy, I guess I didn't have to re-post it again. oh well.

[gone_fishin]

Originally posted by gmat
note: flow rate is *not* pressure.
Actually (setting aside parallel branches) the flow is constant and the same in just any point of the circuit.

The pressure is a factor of several things:
* pipe diameter
* position in the circuit. It's another flame-war topic but on a pure fluid dynamics point of view, the WB should be just after the pump (high pressure) and the rad just before the inlet (low pressure).
Theres a point in the circuit where pressure switches from positive to negative. (not: thats *always* the case even with our poor centrifugal pumps) you'll want that point to be *before* your rad and *after* your waterblock(s).
For more info on that there's a very good article at overclockers.com.

This is a good point on pressure in the system. In my setup I have the rad first then the cpu, I will be switching this as an experiment. But as it is now, I can observe a few things about pressure. The tube between the pump and rad is very hard to compress compared to the tube between the rad and cpu. Also there is a distinct humming resonance that the rad gives off with no fans running (could this be some good turbulance?) Also squeezing off the pump to rad tube results in the water level rising drastically in the fill tube on the res as compared to squeezing the line from the rad to the cpu which results in the fill tube water level lowering. It seems I am set up the opposite of what is recommended.

[gmat]

Here are the equations:

http://www.engr.iupui.edu/me/courses/me314lab/lab10.htm

http://www.mas.ncl.ac.uk/~sbrooks/bo...08/node11.html

Hold on i'm looking for more...

[jtroutma]

Lets look at pressure as this:

Pressure is the driving force behind moving water. Volume is the amount of water we have in a given area (tube, glass, etc.) There are two ways we can create flow:
1) Move the water physically with another object (ie. piston, raise glass up high and pour, etc.)

2) Try and push more water into a give area where there is already water present. (ie. through a tube)

Our pumps are great and moving large volumes of water but they have very little "pushing" power to push the water through the lines. SO the pumps must rely on moving more water to compensate for not being able to PUSH a lot on the water.
Example: take your pump running and pinch the hose shut. The pump will continue to run and be stressed but not explode. Now if our pumps had high amounts of pressure, we would probably not be able to pinch the tube shut, and if we could, the pump might explode due to soo much "pushing" power being forced into the tube. (most likely though the tube would simply pop off the pump)

This also brings up the reason why we dont have to worry too much about crimping the ends of our hoses on our pumps and lines. Because our pumps (in most cases) can not put out enough pressure to "pop" our lines off.

NEXT!!!!

[gmat]

Toot !
Our pumps dont "explode" due to their very design. An impeller on a "floating" magnet / rotor.
They *do* produce pressure. And if you pinch a tube they produce their head pressure ! No more, no less.

[Cova]

So then what you are trying to say jtroutma is that our pumps produce very little pressure.

I agree - most of the pumps we use were designed to just move water around inside an aquarium, where there are virtually no restrictions to flow (very little tubing, no blocks/rads, just a filter to resist flow)

However - given that we are all using similarily low-pressure pumps, why does this matter? All of our pumps are low-pressure, yet still we need to design a good system to have that pump push the water through so as to make it as efficient as possible. The fact that we all have low-pressure pumps is irrelivant, we're discussing (arguing, flaming, etc.) how to best use the pressure we have.

[chazz469]

Originally posted by gmat
Chazz: flow rate is *not* velocity. Its a product of velocity times cross section.
And pressure is *not* flow rate.

I think that's what I was saying. Especially that pressure is
not the same thing as flow rate.

Originally posted by Cova
Current = flow-rate as measured in volume / time, not speed.

So flowrate is volume/time (i.e. gph) as opposed to distance traveled/time (i.e. mph).

I'm gettin' it....slowly......

[gmat]

Cova you read my thoughts

Lemme explain the equations for the not math-savvy:

Q = U x A x delta T

Where

* Q = heat transfer rate, in BTU/hr or Watts or other power units
* When the system is at steady state, this equals the power output of the CPU. Neglecting ambient losses in the tubing and reservoir, Q is the same for the radiator as for the water block. It's just going in at one point and out at the other.
* U = heat transfer coefficient, in (BTU / hr ft² °F) or equivalent units. Barring fouling or a change in flow, this will remain constant. Note: it's a direct factor of flow...
* A = area, in square feet This is fixed by the heat exchanger design. To increase it, buy a bigger rad.
* delta T = difference between the temperature of the hot stuff and the temperature of the cool stuff.
* If the change in temperature of the cooling material is significant as it passes through the water block or rad, one should use a log-mean average for the temps in the subtraction.

I hope that settles it...

[jtroutma]

A SMALL amount of pressure, YES, but a LOT of pressure, NO.

Our pumps are designed so that if something restricts the line, the impeller will just sit there and thurn water. In other pumps which are designed for high pressure systems, if the exit is blocked, the pump will either: burn out due to extreme stress, or cause an impeller to fracture or pump casing to crack, OR be releaved by a releif valve.

ONLY systems designed for pressure have to worry about these problems. (such as in hydrolic systems)

Remember that you can create pressure by forcing more volume into a smaller space but in our setups, its only possible to go up the our pumps rated head pressure.

Head Pressure = NO resistance; how much water can you force up that outlet with nothing attached.

[gmat]

http://www.stewartcomponents.com/adv...tem_basics.htm

they say it all...

[jtroutma]

Sorry to get off topic Cova. Was just trying to refute a small point
Now that we have established that the pumps put out little pressure but large volume............................. how can we use this to our advantage?

BigBen, SEE WHAT YOU HAVE STARTED MAN!!!!!!

[Cova]

Originally posted by gmat
Cova you read my thoughts

Lemme explain the equations for the not math-savvy:

Q = U x A x delta T

Where

* Q = heat transfer rate, in BTU/hr or Watts or other power units
* When the system is at steady state, this equals the power output of the CPU. Neglecting ambient losses in the tubing and reservoir, Q is the same for the radiator as for the water block. It's just going in at one point and out at the other.
* U = heat transfer coefficient, in (BTU / hr ft² °F) or equivalent units. Barring fouling or a change in flow, this will remain constant. Note: it's a direct factor of flow...
* A = area, in square feet This is fixed by the heat exchanger design. To increase it, buy a bigger rad.
* delta T = difference between the temperature of the hot stuff and the temperature of the cool stuff.
* If the change in temperature of the cooling material is significant as it passes through the water block or rad, one should use a log-mean average for the temps in the subtraction.

I hope that settles it...

I may be reading your thoughts - but I still don't understand 3/4 of the equations you post up here.

Most everything that happens in our systems though can be figured out very simply. Our measurement tools are not accurate enough to determine a lot of these small variables in all the equasions, but if you keep basic physics in mind its very simple.

[gmat]

Originally posted by jtroutma
A SMALL amount of pressure, YES, but a LOT of pressure, NO.

Of course

Our pumps are designed so that if something restricts the line, the impeller will just sit there and thurn water. In other pumps which are designed for high pressure systems, if the exit is blocked, the pump will either: burn out due to extreme stress, or cause an impeller to fracture or pump casing to crack, OR be releaved by a releif valve.

Because theit "impeller" is mounted on the engine shaft.

ONLY systems designed for pressure have to worry about these problems. (such as in hydrolic systems)

Wait... arent our watercooled PCs hydraulic systems ?
Besides we *do* need to take pressure into account. Look at Reynolds equations about laminar flow.

Head Pressure = NO resistance; how much water can you force up that outlet with nothing attached.

What ?...

Head Pressure = seal the output and put a manometer. Bingo, positive (and not negligible) pressure...And guess what it''s the very pressure (in water column height - yes its a pressure unit) that written on your pump casing !

[Cova]

Originally posted by jtroutma
A SMALL amount of pressure, YES, but a LOT of pressure, NO.

Our pumps are designed so that if something restricts the line, the impeller will just sit there and thurn water. In other pumps which are designed for high pressure systems, if the exit is blocked, the pump will either: burn out due to extreme stress, or cause an impeller to fracture or pump casing to crack, OR be releaved by a releif valve.

ONLY systems designed for pressure have to worry about these problems. (such as in hydrolic systems)

Remember that you can create pressure by forcing more volume into a smaller space but in our setups, its only possible to go up the our pumps rated head pressure.

Head Pressure = NO resistance; how much water can you force up that outlet with nothing attached.

Head pressure will be rated in PSI (or some similar unit) - how much water you can force up that outlet would be measured in GPH (or some similar unit). Those two things are not the same.

To go back to my favorite electrical analogies - our pumps are small batteries (low voltage, low pressure). If you hook even a small battery up with no resistance (eg. short it out), it will put out a LOT of current (likely fry itself) - if you hook up even one of our small pumps with no resistance, it will move a lot of water.

But really - all pumps are very similar - they create pressure (PSI), not volume (GPH). And a pump doesn't create pressure by pushing water into a small space - a pump creates pressure by transfering electrical energy into mechanical energy and exerting that energy on a fluid (and pressure is stored energy). Forcing more volume through a smaller tube creates more speed (MPH) for the same volume (GPH), and soon gmat will likely post up the formula that shows how to calculate them all based on one-another.

The reason we don't worry about pressure with our pumps, is because they are all mag-drive. The impeller is physically disconnected from the motor, connected only through a magnetic field. If the impeller gets stuck the motor can continue to turn, and so it doesn't burn out. Pressure is irrelivant - though typically a mag-drive system will produce lower pressure because the motor can't exert it onto the impeller without a physical connection.

Edit: I'm out for the day - gotta go do a lot of cooling-unrelated stuff before everything closes.