Table of Contents
Introduction
Definitions
What are we essentially doing when we watercool our CPU?
A simple model of the watercooled CPU
Which coolant should I use?
How important is the coolant flow rate?
Introduction
This section is written to provide answers to frequently asked questions, to provide
solutions to common problems, and to give some more background information. It also tries to
give an explanation why things work, which might involve
a bit of very basic physics. Please note that answers and solutions found here are the
result of what I think is true. This is not always equal to the view established by other
'authorities' on the PC watercooling subject.
If you think that what I am telling here is just wrong, please drop me a mail
Back to top
Definitions
This section explain some of the definitions used throughout the document. Not everything is accurate
to the last decimal, but it is not meant to be either. For more accurate definitions see any highschool
physics book. I recommend to skip this part, and come back when needed.
Absolute temperature: Kelvin
The Kelvin (letter: K) is the basic absolute temperature measurement unit. The coldest possible temperature
is defined as 0K. The scale of the K is equal to that of the degree Celcius, which means that an increase
of 10K in temperature is equal to an increase of 10 °C in temperature. The only difference is the
offset: 0 °C is equal to approximately 273K. 20 °C is equal to approximately 293K.
Energy: Watts, Joules
An absolute amount of energy is given in Joules (letter: J). A joule is approximately the amount of
energy required to lift 100 grams over one meter, to give you an idea about the order. For a more accurate
definition, see your physics book.
A Watt (letter: W) is defined as one joule per second. Thus, a 100W lightbulb 'uses' 100J per second (energy cannot be used. It
can only be transformed from one form to another, or to mass). With this energy
you can also lift 1 kilogram over 10 meters each second.
Specific heat
The specific heat is the amount of energy it takes to increase the temperature
of a kilo of matter with 1K. For example: plain water has a specific heat
of 4.18 kJ·K-1·kg-1. This means that if we use a heater element of 4180W to heat a
kilo of water (approximately one liter), that the temperature rises 1 degree every second. Now, if we
used a heater element of only 1045W, the temperature would rise 1 degree every 4 seconds.
Adhesion and cohesion
The larger the adhesion of a substance is, the better it sticks to a surface. The larger the cohesion of a substance is, the better the molecules
of the substance stick together.
Thermal resistance
Thermal resistance (letter: Ø) is used to quantify how well heat can be transported. It's unit is K·W-1 (Kelvin per Watt).
Thus, if we have a heatsink that has a thermal resistance of 0.2 K·W-1, and we put 10W of heat energy into it, it's temperature
will rise 2K above ambient. The thermal resistance of an entire cooling system is equal to the thermal resistances of it's separate parts
added together.
Back to top
What are we doing when we watercool our CPU?
Well, the answer of this question looks obvious, but it takes a good look to see what we are really doing:
making a giant HSF (Heatsink and Fan) combo. The task of a standard heatsink equipped with a fan is to absorp the
heat generated by the processor, and pass it on to the air that the fan blows over the heatsink. We do exactly
the same in a watercooled system. We absorp heat from the processor with a waterblock, and we pass it
to the air by using a radiator and a few fans. The only difference is that we have two parts now, and we have
a way of transferring heat between the two parts (the coolant circulation). Because the part that passes the heat to the surrounding air
can be much larger, we have made a more efficient HSF combo.
Once we have realised that we have not created a bit of magic, but just a complicated and large HSF combo,
it is easy to see the optimal operating conditions. Those are the same for both the HSF combo and our watercooling.
One of the very important optimal operating conditions is:
- The temperature differences throughout the system must be as small as possible. A higher temperature difference
over the coolant inlet and outlet of the waterblock raises the average temperature of the waterblock. If
the inlet temperature is 20 °C, and the outlet temp is 30 °C, the average waterblock temperature
is 25 °C (in reality it's closer to 30, but hey, it's an example). Now, if we reduce the temperature difference,
we also reduce the average temperature. If the inlet temperature is 20 °C, and the outlet temperature
is 21 °C, the average of the waterblock is 20.5 °C. Way better!
The same is true for the temperature difference between waterblock and radiator, and the temperature difference
between the inlet and outlet of the radiator. This is true because the amount of energy a radiator can dispense is
directly related to the temperature difference between the radiator and the ambient. A large temperature drop
over the radiator lowers the average radiator temperature, which reduces the effectiveness of the radiator.
A large temperature difference between the waterblock and the radiator causes the same effect. It is vital that
you understand this if you want to be able to make your own decisions about what is good, and what is not.
Back to top
A simple model of the watercooled CPU
In the definitions section, we said that the thermal resistance of an entire cooling system is
equal to the sum of the thermal resistances of the separate parts. Now, if we want to design a good cooling system, we
must know where the thermal resistance comes from. Below is a simplified thermal model of a CPU with a watercooling system
installed.
| Ambient |
| | | | Depends on the surface area of the radiator and air flow over it
| | Radiator |
| | | | Depends on coolant and flow rate
| | Coolant in waterblock |
| | | | Depends on the waterblock materal and design
| | Bottom of the waterblock |
| | | | Depends on lapping and thermal compound used
| | Top of the CPU package |
| | | | About 0.15 K·W-1 for FCPGA packages (Athlon/CuMine)
| | CPU die (the chip itself) |
|
As we can see, we have only one constant factor, and that is the thermal resistance between the CPU chip itself (also called
the 'die'). It also explains why it is impossible to obtain really low CPU temperatures: say that our processor is a
heavily overclocked Athlon, putting out 80W of heat. Now, if we are able to keep the top of the CPU at 20 °C, the CPU
die will be 20+0.12*80=32 °C. Good HSF combos equipped with highspeed fans can reach a thermal resistance of about 0.2 K·W-1. A
decent watercooling system can reach 0.05 K·W-1. Assuming the rest of the thermal resistances stay constant, switching
from a good HSF to watercooling gives you about 4 °C improvement in CPU temperature. Therefore, a watercooling system without
TEC is of little use if improving the overclock is your only goal.
Back to top
Which coolant should I use?
Short answer: plain water with an anti-bacterial additive.
Longer answer: there are many liquids that you can use as a coolant in your homebrew
watercooling system. Samples are water, oil, alcohol, automotive coolant and mercury. But before we
can decide which one performs best we need to know which properties are required for a good coolant.
The most important ones are:
- Large specific heat.
The larger the specific energy of the coolant, the less it's temperature rises during the
travel through the waterjacket, assuming constant coolant flow rate. The reverse is also true: the
temperature of the coolant in the radiator does not drop much, thus the average radiator temperature
will be higher. This improves the radiator's ability to get rid of the heat.
- Low adhesion
A fluid with a high adhesion coefficient sticks to the metal surface of the waterjacket and radiator, leaving a slow
moving film of the fluid. This prevents efficient removal of heat.
- High thermal conductivity
The heat transfer from waterjacket metal to coolant occurs through two mechanisms: radiation and conduction. Of these two
mechanisms, conduction is responsible for the majority of the heat transfer. A high thermal conductivity allows the heat to
distribute quickly through the coolant.
Now, look at the properties of different liquids:
| Water | Engine oil | Alcohol | Mercury
| | Thermal conductivity (W·m-1·K-1 | 0.58 | 0.15 | 0.17 | 8.7
| | Specific heat (kJ·kg-1·K-1 | 4.18 | 1.85 | 2.40 | Unknown
| | Adhesion | Unknown
|
This clearly shows that alcohol and oil are bad choices. Mercury and water are way better. But mercury is extremely heavy,
expensive, and both mercury in liquid form and in gaseous form are very bad for your health. The winner is clear: water.
But plain water has the disadvantage of being a good soil for unwanted bacteria and algae (veteran's disease, anyone?).
To prevent this, you must add an anti-bacteral additive. Good candidates are copper sulphate (CuSO4, colors your water deep blue), alcohol, or
commercial anti-bacterial additives used in waterbeds. Using distilled water does NOT solve the problem, because your hoses
are not entirely sealed, and allow gases to pass through. Some bactera and algae can use this in combination with light to grow.
Another possibility which also improves the heat transfer capability of water is to use an additive called WaterWetter, produced
by Redline Racing.
Back to top
How important is the coolant flow rate?
Short answer: not. Almost anything ranging from 200L/hour to 1500L/hour is fine.
Longer answer: not here yet
Back to top
| 
