Loop order doesn't matter. Or does it? - Page 2

Very spot on posts guys..
My take on this subject is, as long as you can have the cleanest routing possible loop order will jot matter much, however as long as flow and having the pump fed temps will reach an equilibrium so to speak.

Yeah, and it's worth observing too that this means the water would be 3c warmer, not the next component in the loop (eg. CPU), since say the CPU is at 60c under load, with the water being at 30c vs 33c, that's only a 10% difference in the delta between the water and CPU block, and won't add up to any noticable difference in temps.

Well, it really does not matter for all practical purposes. While "200W" might sound like a lot you would have to take into account a flow rate and energy needed to increase water temp by a 1 C. As rough rule of thumb if you have a normal flow rate any component your water goes through increases the temperature by ~1 C (usually, in reality less than that) so - the "last" components in your theoretical chain of 10 GPU's would have 10C higher temperature than the "first" component. Now ask yourself whats the flow rate in your loop and how much water is it in your loop ? I'll give you an example - I have 9x120mm monster radiator, 2x GPU's a CPU and 5.25'' bay res - there is approx 2 L of coolant in my system, most of it sitting inside a radiator. So assuming a "normal" flow rate of, lets say 180 L / h or 3 L/min it means that each L of water spends about 10 seconds, at most per component and 30 seconds in radiator in my loop. I have only 2 L in there though - so all the water in my loop goes through all components 1.5 times each minute. At the end of a day it really does not matter for all practical purposes.

Now that out of the way, if you have really crappy flow rate and a lot of components I would put CPU first as GPU's take temperature a lot better - if doing so would not increase substantially the length of my loop - if it would then the losses I would take in performance from the longer loop would probably negate any gains I might get from a couple degrees lower CPU temp.

Dunno how often this needs to come back, but it seems to be one of those subjects that needs to be debunked over and over.

Takes 264w of heat at 1gpm to raise water 1c. At 1 gpm the water going through the loop won't change much in temperature (at most 7c, and that's with a lot more wattage in heat than a normal 15A circuit can dish out of the wall). Meaning as soon as you reach your delta t difference (the temp at which your rads dissipate the most heat due to the difference between ambient and loop temp) the water will have equalized and stay at the same temp. Now if you DO care about 1 or 2 celcius, then by all means route the tubing for the cpu first (as most people won't have more than two GPU's in a loop anyhow). However, if you do and you have too much tubing, then you affect air flow and water flow. Which might actually cost you another 1 or 2 c.

TL;DR The differences are negligible when routing the loop, and to go through the trouble isn't worth it 99% of the time.
Edited by Rognin - 10/21/12 at 7:42am

You are correct in a theoretical sense, but for all practical purposes it makes no difference - the specific heat capacity of water is too large for it to make much difference.

In my loop for example, when my pump is turned down to low flow (about 0.5gpm), to maximise the temperature differences in the loop, I see about a 3c difference between the water going to the PC from my radbox, and the water coming back from the PC. This is with very carefully calibrated sensors, so it should be pretty accurate.

Bear in mind, I'm actively trying to achieve the situation with the least possible equilibrium there by turning flow right down.

If you double that flow to a more normal 1GPM, you'd see a 1.5c difference from inlet to outlet.

This is with 3 GPUs and the CPU, dumping about 800W or more of heat into the water.

In actual, practical application, that means no difference whatsoever in component temperatures. Just because the water hitting it is 3c warmer, doesn't mean for example that the CPU will be 3c warmer - the difference between the CPU temperature and the water is still large enough that it's very easy for the energy to get from the waterblock into the water, the only time you'd see a difference in component temperature would be if the delta between the water and the waterblock became quite small - going from say 30c water and a 60c CPU (a 30c block/water delta) to 33c water and a 60c CPU (a 27c block/water delta) makes no difference that you could ever measure in a PC - to even measure the difference would take a carefully calibrated heat source with an absolutely constant and predictable output, and very sensitive and carefully calibrated temperature sensors.
Edited by BorisTheSpider - 10/21/12 at 8:05am

A few others have already given good answers but I wanted to highlight this part because it is an overstatement and thus a false assumption. Water coming out of the rads is not reduced all the way to ambient, unless you have a super efficient huge array of rads that can take out all the heat in one pass, or your room already has a high ambient temp (in which case it will be pretty uncomfortable). Just like the computer components generate a certain amount of heat per unit of time, your rads can only dissipate a certain amount of heat per unit of time. Generally speaking, the dissipative capacity is less than the heat generated when you start your computer, so the water temps will increase at first. The reason they don't just keep increasing without bound is that water (or any substance) will naturally dissipate more heat as the delta to ambient increases--think about how quickly a cup of tea cools down after you pour it. It cools down quickly just out of the kettle and more slowly as it approaches room temp. The same way in reverse, as your loop heats up, the efficiency of your rads increases. This also means that there are sweet spots where everything balances, ie. the total heat introduced at the blocks and removed from the rads (and very slightly all over as a general passive cooling effect) are equal. The equilibrium point changes depending on the load, so you will normally see your coolant temps stabilize at one point when you're not running your video cards, and at another higher point when going full guns. That's what we mean by "equalize": not that temps are completely the same everywhere, but that the loop overall has hit one of those equilibrium points.

So you are correct that coolant temps will always be a little bit lower as measured exiting the radiators and a little bit higher exiting the blocks, but as many have mentioned, the difference is normally less than one degree celsius and is not a major factor in the overall cooling ability of your loop.

Your rad won't remove all of the heat in one pass just like the blocks won't heat up the water much in one pass. It takes time. Turn down the fans on your rads and watch how long it takes the water to heat up. Eventually it will equalize. Then turn them back up and watch how long it takes your temps to drop. It doesn't just happen right away.

Loop order doesn't matter because each individual water molecule is moving so fast it doesn't get a chance to take all 600w - lets say- of heat at one go otherwise certain molecules would be evaporating off in your loop.

After say 10 runs of the same loop, the water should be equilibrated i.e the water will be at a constant temperature until you stop gaming, folding or what ever it is you do.

Look at it this way--

My current loop runs at about 4.2 Liters per minute. There are at most 2 liters (probably closer to 1.5) of water in my loop, which means that every minute my coolant makes at least 2 complete loops, closer to 3. I've got 4 temp sensors in my loop and can tell you that it takes at least 2 minutes for my water temp to show any measurable bump after applying 100% load on both my CPU and GPU. In other words, my coolant makes 4-6 complete circuits around my loop before any measurable change in temperature occurs. (This is backed up by the fact that my 4 sensors always read within .5C of each other, which is within their margin of error.)

You can come up with crazy situations where loop order might make a difference of a degree or two, but changing your order will usually imply adding tubing and therefore resistance, slowing down your flow or requiring more pumps (which mean more heat dumped into your loop), which will negate the marginal benefits you got from rearranging things.