Starting off big...

sry, but that is a terrible way to think about it... a 10C delta is absolutely terrible loop set up... your goal in a loop should be to get the delta (from ambient air to water temp) as close to zero as possible... the general rule of thumb that i like to follow is 120.1 worth of rad space for every block you have (and then 120.1 to 120.2 of extra rad space just to be safe)... obviously by changing fan speeds and rads used this rule can be broken, but its easiest to follow... the main way you are gunna effect your delta is through loop set up allowing for maximum flow rate and minimum delta between air temp and water temp... if you would like to post some specifics about your computer, what you plan on cooling, budget, case, ect. im sure we would all be able to help- you out much more

ok... now that (i think) i understand what you are trying to do at the moment, i will try to explain what my ideas on the matter are (if you will excuse my terrible spelling and grammer as i am on a tablet at the moment)... the #1 most important think in water cooling is the delta between the max temp of the water (anywhere in the loop) and the ambient air temperature... this delta needs to be as low as possible for the best performance... there are two main ways to do this

First is the most simple... increasing the cooling compasity of the loop as a whole (eg. increasing rad space)... this can be done through swaping out for a bigger radiator, adding radiators to a loop... the other way to do this is to optimize the radiator space that you do have by increasing fan speed (this will provide minimal improvements if the system is set up correctly in the first place)

the second is to increase your flow rate (this is where the majority of systems can use some improvement)... the general concept is that by increasing flow rate you increase the volume of water that has to absorb the same heat (generally measured in Watts or Watt hours)... this is done by both using a more powerful pump, and decreasing restriction in you loop (most commonly done though minimizing over all loop length and by choosing low restriction blocks)

you should note that both of these obviously have a point of diminishing return... for the radiator space, this point is around 100w per 120.1 worth of rad space (using 1500 rpm fans and decent rads)... for the flow rate it is debatable, but is generaly said to be between a flow rate of 2.0 gpm and 2.5 gpm (this may even be a bit high for your uses)


since you just plan on cooling a CPU and a (670 class or so) GPU you should be just fine with around 120.3 (360mm) to 120.4 (480mm) worth of rad space... this assumes the use of medium speed fans and decent rads (witch it does apear fit your needs)... i would also reccomend ignoring that statistic that you linked to at the moment as it can be rather hard to use no matter how long you have been water cooling and how well you understand it (it still confuses me near to death when i first look at them)... that statistic is also truly only useful for major enthusiasts looking through hunderds of different radiators and nit picking them for their very specific needs (atleast that is the only reason i pay true atention to them)... the mojority of radiators will perform roughly the same (based on tier and cost) at the same fan speeds (of course there are exceptions)

also, since you are planning on a rather simple loop, you wont need to worry much about loop order, but there are a few things to know about it... the general rule of thumb is that the flow of a loop should always go res > pump > rad > "component you want the coldest" and that your loop should be as short as possible... this a very great rule to follow both when you first start and in almost every simple loop... now since you were asking about concept i will go a bit more in depth

as you get more advanced and start building bigger more complex loops with more heat dump it may be advantageous to stray from this set up... the two most notable and different way are changing the loop order and also sacrificing extra length of a loop for more keeping the delta between the max water temp and the air temp at a minimum... these two often fall hand in hand

the most common way to do this is when you have multiple radiators, many components to cool,and multiple pumps (or simply something very strong like an AC pump for example)... there are also some rules that either need to be followed or for one to be very careful when breaking (i will note these later)... I will show an example of a loop when more tube length and radiator movement could majorly improve the efficiency and the performance of the loop

this loop is a loop i proposed for a Cosmos II build i saw a while back (though a few components were changed and generalized)... it consisted of a CPU (3960x or similar), multiple GPUs (dual 680s/7970s or more), one 120.3 radiator, one 120.2 radiator, one res, two pumps (35x or similar), mobo cooling, and ram cooling... you should note that though mobo and ram cooling can do some good in extreme use and cases, they are mainly ascetic for the vast majority of PC users... the loop that i recommended was ordered as such:

res>pump 1>CPU> mobo and ram>120.2>pump 2>GPUs>120.3>res

the reason for this order is that the higher the temp of the water is, the less energy it can absord... with a heavily OCed CPU such as a 3960x or so that can very easily put out a few hundred watts combined with the minimal (yet still important) heat of the ram and the mobo... even with a good flow rate, this can easily raise the temp of the water by at least a few degrees C... even simply the 3-4 degrees that the temp of the water has raised, can make quite a large impact on the GPUs cooling (especaly when they are putting out well over 400W (in some cases i have seen, even up to around 1200W)... this set up is mainly possible due to the use of two very strong pump that kept the flow rate around 2-2.5 gpm even with the largely extended tubing and many components

now that i have explained most of the basics, its time for some of those notes and rules i mentioned earlier:

1) always keep the pump where it can get plenty of water!I cannot stress this one enough... the fastest way to kill any loop is to burn up a pump because it is struggling too much for water... the easiest way to fix this problem is to simply have the res above the pump and to feed water straight from the res to the pump... if you need to have a pump where the is no res to feed directly into it, the easiest way to make sure you are fine is to have the pump physicaly below (closer to the ground than) the component before it in a loop (and make sure it is not a restrictive component such as RAM and mobo blocks, or even a single GPU)

2) keep the component you want the coldest after the largest rad (or only rad)

3) increase rad space and flow rate to increase efficiency and cooling potential

4) there will always be a point of diminishing return



i hope that all of this helped, and if you need any help when you get farther in just feel free to post here and im sure you will find plenty of help
Edited by eskamobob1 - 8/7/12 at 3:50pm

Mathematically speaking, 1 gpm of water = 264 watt/kelvin. In a system with 2 gpm flow, water will heat less than 2 C as it picks up 1000 W from the waterblocks.

Likewise, 50 cfm of air = 28 watt/kelvin. That's the upper limit on what you can dissipate even with the best radiator. 50 cfm is about average for an unobstructed flow rating at 1500 rpm for a 120x120 fan. in practice, hanging the fan on the radiator will cut the flow in about half.

Say, you have 550 watt coming out of the computer, 1 gpm water flow, 6x120 fans blowing 150 cfm total, and an ideal radiator.

Ambient air: 25 C
Air coming out of the radiator: 25 + (550/28)/(150/50) = 31.6 C
Water coming out of the radiator: 31.6 C
Water going into the radiator: 31.6 + (550/264) = 33.7 C
Air out - air in delta 6.6 C
Water in - water out delta 2.1 C
Water in - air in delta 8.7 C

Average temperature across waterblocks (31.6+33.7)/2 = 32.6 C

If the radiator isn't ideal, or either the water flow or the air flow is too high, it won't equalize water & air temperatures and you get one extra term (water out - air out).

In this example, the air out - air in term is the largest, so the setup would benefit from getting more & better fans.

It also follows that 2 gpm is past the point of diminishing returns in most practical cases: in this system, with a radiator setup as large and expensive as it is (2x360), even the flow of 1 gpm is not the bottleneck.

Temperatures of the CPU and the GPU will be higher, depending on how good water blocks are and whether they are mounted properly. Every degree off air-water delta is a degree off the system.

For a CPU, core-water delta may be 30-40 C, so there isn't much point pushing air-water delta to the ridiculous extremes. For a full-coverage GPU block, core-water delta can be under 10 C and air-water delta is more interesting.
Edited by hamster3null - 8/7/12 at 4:50pm

I think you've missed the part where I said that, if the fan hangs on the radiator, it won't do 50 cfm, it will probably do about half. Look at the bottom chart here: http://img524.imageshack.us/img524/552/summary20b.png So 150 is _six_ fans blowing at 25 cfm each.

I was assuming 1 gpm flow.

It will equalize in a state where water out temp is greater than air out temp.

No, these calculations won't tell you if your system won't equalize.

If it's a CPU, one extra degree of delta might mean 40 C above ambient vs. 39 C above ambient. Not a big deal. If it's a GPU, it might be 14 C vs 15 C, bigger difference percentage wise.