*spin-off* Feasibility of Water Cooling Radiator Setups
- Thread starter almighty
- Start date
Prophecy2k
Veteran
That's the bit I don't quite understand. If the water only gets a few degrees above ambient, it's not going to be efficiently transfering heat to the air. Hence surely it'll need a very large radiator? Or high air volume. I'm sure someone can provide the maths to show relation to heat disippation, water temperature and surface area. I can imagine a system like PS3 with the radiator on the bottom of a big fan and a peristatic pump moving water around (I'm pretty sure when I used them in biochem that we worked with very watery fluids, but perhaps the amount of effort to squeeze the pipe increases) geared off the fan motor, but I don't know if that's actually a feasible design.
Sorry Shifty I missed this one. I believe the relationship you are looking for is version of:
Q = U A dT'
Where,
Q is the rate of heat transfer in J/s or W
A is the surface area available for heat transfer in say cm²
U is the overall heat transfer coefficient of the system in W/cm²K
dT' is the LMTD (log-mean temperature difference) in degrees K
U is a constant and a function of the heat transfer mechanism (i.e. radiative, conductive, convective or combination of them), the properties of the heating/cooling fluids (e.g. thermal conductivies, velocities etc) and the dimensional arrangment of the flow path of each heating/cooling fluid. U is given by the design of the heat exchanging unit, it's materials, and flowing fluids from which heat is transferred one to another.
The LMTD is a logarithmic mean temperature difference used because with this system (i.e. likely counter or cross-current flow of air/water) temperature change along the flow path of each fluid is not linear.
So if you have ambient air at temp. t1, and water into the radiator at temp. T1, say your outlet temperatures are t2 (air) and T2 (water), the temp. differences (use R) at the inlet and outlet are:
Rin = T1 - t2
Rout = T2 - t1
The dT' or LMTD is:
(Rin - Rout) / Ln(Rin / Rout)
For an air/water system as discussed the water is the "hot" fluid and the air the "cold" fluid. The outlet air temp. will never exceed the inlet water temperature as the air or cooling fluid is the poorer heat conductor.
Basically if the temperature difference between the air and water into the system is small, t2 will be very close to t1, thus Rin is low, and thus your LMTD value is also small. So for a given amount of heat that you want to eject from the system (i.e. the heat produced by the processor core) the area you would need becomes huge.
Shifty is correct, in that a cooling water circuit is not an efficient way to remove heat from a console/PC, regardless of your fancifulness in radiator design. It boils down ultimately to temperatures of the water and that of the air, in which case the difference will always be small and thus imply low heat transfer performance of heat ejecting into the air at the radiator.
You could on the other hand run the cooling water circuit hotter, using a lower water volume, thus allowing for a greater thermal driving force for heat transfer at the radiator/fans end of the circuit. The only problem is that the cores you're trying to cool also run hotter. In such a case you'd ask yourself whether it's worth it then, as you could simply use heatpipes, heat sinks and fans, and save all the added complexity.
Otherwise you run large water volumes in your circuit, keep your cores cooler, but suffer shitty heat transfer at the radiator end because your temperature approach to the ambient air temperature is small (thus requiring a large radiator with lots of surface area to get the required heat out of the system).
Hope that clarifies
I didn't assume a 10 degree rise across the radiator ... I assumed a 10 degree rise of the water temperature across the block, and a 10 degree drop of water temperature across the radiator. Those are guaranteed in the steady state (ignoring rounding errors and heat transfer through the pipes, mobo etc). There is no alternative, for a steady state each second 100 Joule has to be transferred to/from 2.5 grams of water at those points ...
The thermal resistance of the CPU->water and water->air transitions are a separate issue (and not relevant to it's capability as a "transfer medium").
Last edited by a moderator:
I've a question regarding water cooling.
The water circulation is handled through a pump right?
Is the water only present in one phase in the device? I found the idea interesting to have the water in two phases in the design. Having water in a gaseous state would save you the pump
The water would move to the radiator by itself (in gaseous phase) where is it could condensate into liquid, if the design is properly design and take in account gravity (as well as capillarity) it would then return the pool of of liquid water.
It would have nice benefit as as long as two phases are present and that volume is constant only pressure with the device would varies (if memory serves right).
I believe that the thing is unpractical, you would want to avoid air into the device, it has to handle pressure, etc. So I don't think it's workable / worse the cost.
Anyway there is something I don't really "get" ( not exactly) to me the benefit of water cooling is not the that it cools thing better by it-self but that it moves "heats" more efficiently.
It will take longer to for water to reach the same temperature as air or metal but it will also take longer to cool down. Clearly the main benefit to me as far as cooling is concern is that in fact the water "retains" warmth better and so it will while you're moving the heat away from the device you cool. But I see no benefit in "cooling as such", once you have successfully move the heat away from the device you wanted to cool let say to a radiator it will take longer for the water to cool down too, either way you will spend more power cooling it down or will need a big radiator, or both a big fan cooling down a big radiator. Lavoisier 101 energy conservation.
Clearly to me water as some other liquids are not good at cooling by themselves as the extra time they take to warm you will pay them they will have to cool down (and you want it to cool down), it's that they serve really well as "media" for heat/energy.
For next generation system, I would say the odds are in the 99.999%, out of many servers farms with big money how many use water cooling? Almost none, why? because it's not a magic bullet, it's only a temporary storage for the heat/energy your chip is dissipating in the end you still have to dissipate the heat. it's only useful in extreme case when you have to take away the heat from a really hot chip (like some IBM mainframe).
In a perfect world you would want to air cool the heat pipe/radiator which contain any material which can be present in two phases at the intended temperature you want your device to run
The water circulation is handled through a pump right?
Is the water only present in one phase in the device? I found the idea interesting to have the water in two phases in the design. Having water in a gaseous state would save you the pump
The water would move to the radiator by itself (in gaseous phase) where is it could condensate into liquid, if the design is properly design and take in account gravity (as well as capillarity) it would then return the pool of of liquid water.
It would have nice benefit as as long as two phases are present and that volume is constant only pressure with the device would varies (if memory serves right).
I believe that the thing is unpractical, you would want to avoid air into the device, it has to handle pressure, etc. So I don't think it's workable / worse the cost.
Anyway there is something I don't really "get" ( not exactly) to me the benefit of water cooling is not the that it cools thing better by it-self but that it moves "heats" more efficiently.
It will take longer to for water to reach the same temperature as air or metal but it will also take longer to cool down. Clearly the main benefit to me as far as cooling is concern is that in fact the water "retains" warmth better and so it will while you're moving the heat away from the device you cool. But I see no benefit in "cooling as such", once you have successfully move the heat away from the device you wanted to cool let say to a radiator it will take longer for the water to cool down too, either way you will spend more power cooling it down or will need a big radiator, or both a big fan cooling down a big radiator. Lavoisier 101 energy conservation.
Clearly to me water as some other liquids are not good at cooling by themselves as the extra time they take to warm you will pay them they will have to cool down (and you want it to cool down), it's that they serve really well as "media" for heat/energy.
For next generation system, I would say the odds are in the 99.999%, out of many servers farms with big money how many use water cooling? Almost none, why? because it's not a magic bullet, it's only a temporary storage for the heat/energy your chip is dissipating in the end you still have to dissipate the heat. it's only useful in extreme case when you have to take away the heat from a really hot chip (like some IBM mainframe).
In a perfect world you would want to air cool the heat pipe/radiator which contain any material which can be present in two phases at the intended temperature you want your device to run
Is this post serious? I can't detect tongue-in-cheek.I've a question regarding water cooling.
The water circulation is handled through a pump right?
Is the water only present in one phase in the device? I found the idea interesting to have the water in two phases in the design. Having water in a gaseous state would save you the pump
The water would move to the radiator by itself (in gaseous phase) where is it could condensate into liquid, if the design is properly design and take in account gravity (as well as capillarity) it would then return the pool of of liquid water.
For those not familiar, that's what a heatpipe is.
It actually may explain why a heatpipe is measured as being hotter than a tube of water. A thermal probe on a heatpipe can induce cooling at that point. If condensation occurs there, that part of the heatpipe will be pegged nearer the boiling point of the fluid.
Thanks. Nice to know I wasn't wrong.
Which leads me to the next query, which is what water-cooling fans are actually experiencing, especially in the case of console mods like the one Almighty linked to. Those temp readings are definitely lower than I'd expect, and he claims that google shows the same temps being air cooled are almost twice as high (we could really do with links to that info). That mod hasn't got a huge radiator, not clearly larger than the default heatsinks, although I can see that the fins are all copper. So how can it be possible that the heat extraction is working more efficiently? Is there some expensive aspect to make it more efficient? Or is it not possible at all and what we're seeing are false results?Yes it was I've no clue about how heat pipe are done.
And Yes I messed up badly with my "phase graph" if two phases are present T&P vary along the leg of the "y" that "separates' the three phases. My bad. Still shameful mistake (I take it after +15 years without think of it) but it changes nothing on water cooling I failed to see the benefit especially in a tiny form factor, every bonus you get from the water taking more energy to "heat up" you lose as the same statement is true when you'll want to cool it down. Water cooling uses water in only one phase right? (it looks like it by the photo I've seen here and there).
One question is whether the standard mass-manufactured unit is as well-assembled as a custom refit. Was there a control example where the regular air cooler was removed, then the TIM reapplied and then the cooler seated as well as the water cooler?
Graphics cards show very significant temp differences based on how well the factory applied the TIM and fastened the cooler.
Graphics cards show very significant temp differences based on how well the factory applied the TIM and fastened the cooler.
Last edited by a moderator:
Water cooling is fun, nifty, etc. but I can tell you why I believe you won't see it in a consumer toy (aka console): lawyers.
"You want to put water inside consumer electronics?"
"Don't you know there's a reason why hair curling irons have a warning not to insert them into any orifice?"
Also, I've always used monster cpu air coolers that got respectably close to my couple H2O set-ups. H2O also means finding ways to cool all the ancillary components that may no longer have the benefit of a closely coupled air displacement system ( fan )
"You want to put water inside consumer electronics?"
"Don't you know there's a reason why hair curling irons have a warning not to insert them into any orifice?"
Also, I've always used monster cpu air coolers that got respectably close to my couple H2O set-ups. H2O also means finding ways to cool all the ancillary components that may no longer have the benefit of a closely coupled air displacement system ( fan
)
Bong cooler!!!!
First thing I though was fan-assisted bong...which it apparently is?
Yep, Bong cooler... big, ugly and expensive but unlike a normal radiator they can actually cool water to below ambient temperature!
Prophecy2k
Veteran
My point is that you're still assuming a 10°C temperature rise/drop in the water circuit. Which would mean that your inlet temperature to the block will be at least ambient temperature, and outlet temp. 10°C above that. If the desire is to run your core cool, then you'd want to run your water circuit as cool as you can get it, therefore a smaller temperature rise/drop and larger volume of water in the loop.
Also, your statement in bold is not what i was intimating. Water's thermophysical properties alone make it a relatively bad heat transfer medium. Are you sure you understood what i was saying in my previous posts?
Last edited by a moderator:
I'm assuming a lot, I'm assuming it's justified to use a fixed specific heat of 4 joule/gram °C, I'm assuming there is no heat loss anywhere else in the system but the radiator, I'm assuming the pump doesn't heat the water up etc etc. What I'm not assuming is a 10 degree rise/drop, it's a design parameter not an assumption.
Once the system reaches steady state (and it will, at what water temperature above ambient is an open question) the block will put 100 joule into every 2.5 grams of water which goes through it, and the radiator will take it out. Which gives 10 degrees, as designed.
I don't happen to think that a liquid which can transfer 100 Watt with a flow rate of 2.5 ml/s with just a 10 degree temperature increase is a poor transfer medium.
Adding more fluid won't do anything.
The water only rises 2-3c when going through the water block, Some times it's as low as 1-2c
No matter how big a radiator you will always have a water temps that's 4-5c above ambient due to the air that's being pushed by the fans is at ambient temperature -/+ 2-3
Back in my overvolting/overclocking craze daze I could trounce my water cooler with air...granted I would have to put my CPU on a ledge outside my window in January when it was -9 C, but it whooped on the H2O (which was way to clumsily built at the time to hoist out a window).
Back in my overvolting/overclocking craze daze I could trounce my water cooler with air...granted I would have to put my CPU on a ledge outside my window in January when it was -9 C, but it whooped on the H2O (which was way to clumsily built at the time to hoist out a window).
You put the same water loop in -9c ambients and watch it whoop that big ass heat sink.
Water can't get down to below zero, air can. I saw my P4 run at slightly under freezing temperature of water on air cooling when my box was in -20C
Similar threads
