I have read this thread with interest and it convinced me to join the forum. This is my first post, so I thought I would make it a good one.
I would like to add some more science to the question of the importance of loop order (yes, I am alive and doing science). I read Nordar’s sciency post and he did a good job with the formulas, however there is a better equation for the job. So really this is just an extension of his post. I would like to introduce that equation, everyone likes equations right?
The equation is one of my favourites, mostly because it is quite easy, and is called the Steady Flow Energy Equation (SFEE). It basically states that for a steady flow, the energy in to a control volume is equal to the energy out. Here it is in all of its glory:
Q - Wx = m(Δh + (Δc^2)/2 + gΔz)
Q = heat transfer in or out of the system in kW
Wx = work done on or by the system in kW
m = mass flow rate in kg/s
h = enthalpy in kJ/kg
c = speed of the flow in m/s
g = gravitational constant, 9.81m/s2
z = height in m
Δ = a change in the value (out – in)
Scary? Don’t worry; we can make some assumptions to make it nicer:
1.The control volume in this case will be the water block (any water block) from the pipe inlet to the pipe outlet.
2.No work is done, Wx = 0. There is no work done by the fluid, this term is used for pumps, turbines etc.
3.No change in speed, Δc = 0. As the mass flow in and out must be equal and the flow is not compressed, the volumetric flow (m3/s) is the same. For an equal pipe inlet and outlet diameter this leads to equal inlet and outlet speeds.
4.Negligible height change, Δz = 0.
5.Water is the working fluid. I don’t have steam tables (explained later) for various coolants so I am going to make this simple for myself.
This reduces the equation to:
Q = m(Δh)
Much nicer! Incidentally Nordar, you almost derived this. Good work for one who said “I am certainly no expert”.
The Q, or heat load, can be put in easily. I am going to use 200W for no other reason than it is a nice round number. Remember that the equation uses kW, so we have to use 200 / 1000 = 0.2kW.
For the mass flow, I have used 2 GPM (US gallons). Again, no other reason than it is a nice round number. To convert GPM into kg/s we do the following:
2 * 3.79 = 7.58 litre/min
7.58 / 60 = 0.126 litre / s
As 1 litre = 1 kg for water, m = 0.126kg/s
This gives us Δh easily as 0.2 = 0.126 * Δh, so Δh = 0.2 / 0.126 = 1.59 kJ/kg.
This is where things get a little trickier as Δh does not easily give us ΔT. To get ΔT we have to use steam tables. Steam tables are thermodynamic tools that are used to calculate the properties of water at various points. I wont bore you with the details (mostly because I just use an Excel Add-in for most of my work).
Using the inlet conditions of 1.01 Bar (atmospheric pressure) and 30˚C (another round number) gives a temperature rise of 0.379˚C for a 200W heat load
. Quite small I think you will agree.
A note on the inlet conditions I have chosen: Although the temperature and pressure at inlet do affect the temperature rise these effects are negligible within the range that we see in PC water-cooling and therefore I have just used arbitrary values. The inlet conditions become of greater importance at higher pressures and temperatures.
Just to add a pretty picture and some more scienceyness to the post, I plotted the temperature rise for a range of flow rates and heat loads (attached)
This graph shows a couple of interesting things:
The temperature increases are, as previously stated, quite small and therefore normally the loop order will not make a significant difference. Having said that, if you had two GTX 480’s with full cover blocks you would see a temperature rise of 2.2˚C for 1 GPM, 1.1˚C for 2 GPM (assuming 300W for each card). For someone chasing every single degree it would be unwise to put these before the CPU, especially in the unlikely event that they have a low flow rate (if you can afford 2 480’s then you can afford a decent pump!).
Also worth noting is the effect of flow rate on water temperature rise. The gap between 1 and 2 GPM is very significant, so moving from 1 to 2 GPM will half the temperature increase you see across each component. You get a similar result moving from 2 – 3 GPM, although the effect on water heat rise is less. This shows that flow rate is important up to a point. It is good to ensure that you are getting a decent flow rate, but adding more and more pumps to the system leads to diminishing returns. Please note that I am speaking purely about water temperature rise here, I make no reference to effectiveness of heat transfer, turbulence etc.
This leads me to my final conclusion. Yes, loop order can
be important. If you are chasing every last degree and you have a weak pump and / or high heat loads, put the CPU first. However for the majority of us loop order is unimportant as far as water temperature increase is concerned. As the graph shows, the temperature increases are very small, and better results can be obtained by increasing your flow rate. Therefore in most cases it will be better to build a cleaner loop, thus increasing your flow rate, than it will be to change the component order and maybe lower your flow rate. Also, I don’t want my system to resemble a plate of spaghetti!
I am working on the effect of a water temp increase at inlet on the temp of the CPU, it is a little trickier and not something we are usually bothered with in my work so I don't know how to do it off the top of my head.
Please feel free to add any questions, comments or random abuse.
JohnEdited by GingerJohn - 9/6/10 at 11:40am