running a pipe splitter into cpu/gpu block

fastboy · 31 Jul 2015 at 12:30

most of the loops i have seen seem to use a pipe from radiator-cpublock-gpu block and back into pump/res.

i was wondering why not split the flow from radiator into 2 so each cooling block is fed independant,which is how i think im going to do mine for my first build.

im an engineer/motorcyle service manager by trade before retiring due to a back injury so have a decent knowledge of how things work,i just cant undestand why you would put warm water from one block into another(or maybe the water temp only raises a couple of degrees between blocks over ambient?)

anyone?

Avathar77 · 31 Jul 2015 at 12:32

Close-single loop system will come to its equilibrium with temps over time.

fastboy · 31 Jul 2015 at 13:09

yes but if one method is more efficient than the other then the end temps will differ...otherwise we would all be using cheap pumps and 120mm radiators.

ttaskmaster · 31 Jul 2015 at 13:19

fastboy said:
yes but if one method is more efficient than the other then the end temps will differ...otherwise we would all be using cheap pumps and 120mm radiators.

I work for a water utilities company with several engineers all qualified in Civils and specialising in hydraulics and thermodynamics. We deal with pipes, pumps and coolers on scales from tiny to massive, but the principles are the same.

I challenged them on this before and they gave me some very long, complicated explanations involving numbers, formulae, sketches and reference books, which basically are summarised with - "The temperature evens out throughout the loop".

I'm still going to go pump->rad->CPU->rad->GPU->pump, just because I like how it looks, though.
Not actually considered splitting the flow (and am unsure if the reintegration would spoil the flow rates, unbalance the pressure or cause restrictions on one side), but since I've only just dropped a couple hundred quid on fittings anyway, I may leave that for future plans!! :cool:

fastboy · 31 Jul 2015 at 13:29

this is what i dont understand by people saying it will even out over time,yes each individual loop will have a specific mean temperature..for that particular loop.

adding a different rad or fan would change that mean temp.
putting the rad in a fridge would change that mean temp.
slowing the pump down or speeding up would change that mean temp.

if none of the above were true then why would we not all have 1 fan,120mm rad and $4 pump off ebay..because the mean temp is only viable for that specific loop in those circumstances.

if the single continuos loop is more efficient it should run cooler,if the split loop is more efficient it should run cooler.

i guess all im asking is which would run cooler

D3K · 31 Jul 2015 at 13:29

ttaskmaster said:
I work for a water utilities company with several engineers all qualified in Civils and specialising in hydraulics and thermodynamics. We deal with pipes, pumps and coolers on scales from tiny to massive, but the principles are the same.

I challenged them on this before and they gave me some very long, complicated explanations involving numbers, formulae, sketches and reference books, which basically are summarised with - "The temperature evens out throughout the loop".

I'm still going to go pump->rad->CPU->rad->GPU->pump, just because I like how it looks, though.
Not actually considered splitting the flow (and am unsure if the reintegration would spoil the flow rates, unbalance the pressure or cause restrictions on one side), but since I've only just dropped a couple hundred quid on fittings anyway, I may leave that for future plans!!

Doesn't really work with my understanding of thermodynamics to be honest. If the temp of fluid entering and leaving a block/rad is the same, then no heat transfer is taking place. Delta T = 0 which makes q (heat transfer) also 0.

I can accept that there may be only slight temp differences due to the velocity of the fluid, but there is no way the temps are the same.
It would seem that the temp difference is not enough through practical experience to cause much loss in efficiency from going block to block rather than block rad block, but the fact remains that it is less efficient to some degree.

I have CFD software on my machine, I really must knock something up one of these days.

fastboy · 31 Jul 2015 at 13:41

yes this is what im thinking.

if you have a say pump at 1000 litres per hour but its getting restricted to 500 lph as it hits the cooling block then thats your limit,500lph
now if you split that flow before it hits the block and give each block 500lph then you havent changed the temp of the water as it leaves the blocks but what you have done is give each block the same temp water entering the block..not giving one block hotter water as it is fed from the first block.

couple this back into some thicker pipe as it leaves the blocks as per drawing and your still feeding the res 1000lph(ok there will be some losses due to turbulence and the restrictions on the water being restricted into the res?? but its still better than 500lph??

now ok this is guesswork on the figures because i dont know what flow rates the blocks or connectors are capable of and the may well be capable of 1000lph so its just guesswork by me,maybe someone else could help with that.

the other thing is pressure,the less pressure you can get away with the less chance of leaks too so not sure if that helps with spliting.

kaku · 31 Jul 2015 at 13:45

It's the flow rate of the coolant combined with the rate at which heat is transferred from one "section" of coolant to another which more or less makes any temperature differences negligible.

The reason we don't use a single 120mm rad and a cheap pump is because the flow rate would not be enough, and without flow you'd get "hot spots" in the loop. Simultaneously, the coolant temperature would increase above safe limits causing the pump, tubing, and fittings to fail.

Consider this: if you have a metal ring and you hold it above a candle, the area directly above the flame gets hot, but the heat also spreads through the metal and you can feel it at the top.

If instead you continuously rotate the ring above the heat source, the difference in temperature at any point in the ring is much lower.

fastboy · 31 Jul 2015 at 13:50

and this is my point kaku.
how much flow can the cooling blocks handle,if they are restrcting the flow of the pump then splitting will be more efficient,this why looking at the mean temps of the whole loop is critical.

ttaskmaster · 31 Jul 2015 at 14:11

fastboy said:
if none of the above were true then why would we not all have 1 fan,120mm rad and $4 pump off ebay

The cheap stuff may not have sufficient cooling capabilities (flow, fins, fans, whatever) for your individual rig, though. A 480 rad will cool better than a 120, so we get those. More fans move more air over more fin surface areas.

You're also limited by what you have space for in your individual case. The kit matters more, which is why a custom loop and even some air coolers will both beat AIOs, for example.
But so long as there is a heat transfer taking place, the route isn't so important.

This is why Kapstaad's loop might go Pump-rad-CPU-rad-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-GPU-pump.

OK, I exaggerate - He only buys 8 of each new GPU!

But SLId/X-Fired GPUs generally connect through each other, rather than having seperate branches from separate rads for each one. It could certainly be done, but the temps would not be massively different as what you're feeding in is the same throughout. It matters more that you take that heat away. The nature of a closed water loop just means you then have to cool it back down before it goes back to the blocks.

Think of it as taking heat away, rather than feeding cool in.

D3K said:
Doesn't really work with my understanding of thermodynamics to be honest. If the temp of fluid entering and leaving a block/rad is the same, then no heat transfer is taking place.

The transfer takes place in the rad, which is after the pump, which is after the block. What goes into the block is not as important as how much heat leaves it.

Some temp does drop along that route as well, though it's quite low.
However, some temp is also created, by the friction of the water against the surface roughness of the internal bore of the pipe/tube... and I fell asleep through the rest of the explanation, heh heh!!

D3K said:
It would seem that the temp difference is not enough through practical experience to cause much loss in efficiency from going block to block rather than block rad block, but the fact remains that it is less efficient to some degree.

Not any degree to make a sufficient difference to warrant buying loads of separate loop branches. You can go in pretty much any loop order you want, so long as the res is above the pump to avoid air getting in.

ttaskmaster · 31 Jul 2015 at 14:16

fastboy said:
how much flow can the cooling blocks handle,if they are restrcting the flow of the pump then splitting will be more efficient,this why looking at the mean temps of the whole loop is critical.

If you split, then the 1000l/s flow being restricted to 500l/s by one block will be equally restricted to 500l/s at each of the five, ten, fifty or however many blocks you split it into.
This happens because it's a watertight, airtight, closed loop and the outgoing flow then becomes the incoming flow. If it doesn't then you probably have air getting in.

fastboy · 31 Jul 2015 at 14:26

yes it will be restricted to 500 per block..but each block would be getting the same temp new water at 500lph that left the radiator..not 2nd hand hot water at 500lph for the 2nd block.

couple those 2 outlets back after the blocks into a larger pipe that can handle 1000lph and your back to 1000lph.

again this is only assuming everything else can handle the flow of 1000lph.

think of a bucket with a single hole in the bottom,now drill another and the flow doubles.

for arguments sake:

normal single loop.

water leaves rad at 20c at 1000lph
hits cpu block,drops to 500lph and leaves cpu block at 30c
goes into another block at 500lph at 30c and leaves at 40c
enters rad at 40c and 500lph and cools to 20,cycle repeats.

split pipe.

water leaves rad at 20c at 1000lph
hits splitter and gives each block 500lph at 20c
water leaves each block at 30c
rejoins via splitter at 30c and gives back to rad at 1000lph at 30c.

of course this is dependant on how much flow is restricted by the cpu blocks or any other part of the system after the pump,as i say i dont know thats why i was asking.

Deathjester · 31 Jul 2015 at 14:29

Lets assume (And we are assuming here) that the blocks are equally restrictive:

Series:

@20 degrees C

The CPU has a TDP which heats 1 GPM of water by 1 degree C
The GPU has also has a TDP which heats 1GPM of water by 1 degree C

The radiator has a TDP which dissipates heat of 1GPM of water by 2 degrees C

The water temp is at equilibrium.

With the same components:
Pump head = 1.0 GPM
The CPU can heat 0.5GPM by 2 degrees C
The GPU can heat 0.5GPM by 2 degrees C
This recombines:
the radiator then dissipates the 2 degrees C hotter liquid at the same rate previously noted.

As far as I am aware that's how it would work.

(Although the temperature differences are actually much lower).

fastboy · 31 Jul 2015 at 14:55

i guess its hard to work out without actual data on flow rates of the connections and block flow rate,just something i was interested in.

Deathjester · 31 Jul 2015 at 14:57

fastboy said:
i guess its hard to work out without actual data on flow rates of the connections and block flow rate,just something i was interested in.

Yeah, the only thing which will actually make a difference is if the flow rate for the max pump value is increased when the blocks are in parallel instead of in series.
But that is likely to be dependant on the rest of the loop as well.

fastboy · 31 Jul 2015 at 15:04

yep thats what i was getting at,maximising what the pump has to offer by minimising restrictions and maximising the thruput at minimal pressure.

ttaskmaster · 31 Jul 2015 at 15:24

fastboy said:
think of a bucket with a single hole in the bottom,now drill another and the flow doubles.

Yes... until the water in the bucket drains, at which point the head of pressure drops and the flow reduces as the head is replaced with air.

In a closed loop, that flow is pretty constant throughout. However many holes you put in that bucket, it should go through every hole at the same pressure.

fastboy said:
yep thats what i was getting at,maximising what the pump has to offer by minimising restrictions and maximising the thruput at minimal pressure.

Then that is more down to the components themselves, the fittings you choose, the angles you put them at and the bends in your tubes, affecting how quickly the water can carry heat away from the blocks, rather than the number of radiators or loop branches.
You're also restricted by roughly room temperature, as the fans blowing through the rads cannot really blow much cooler than that.

If you like, I can ask the modelling guys at work if they can map it out in their software?
Theoretically we can create a supply or sewer test-network and just set the asset values (pumping station, rising main, etc) to those of a W/C loop...

pastymuncher · 31 Jul 2015 at 15:41

I believe I have some useful input here. My radiators are both mounted in a box on a windowsill for maximum cooling. I have inline temp sensors on the inlet and outlet of my case. My loop comes into the case to the cpu block, onto the gpu block and back out again. Just now typing this my water temp on the inlet is 17.2 degrees and 17.6 degrees on the outlet. The biggest difference I have ever seen is only 1.2 degrees between inlet and outlet and that was after a lengthy gaming session. Normally the difference is in the 0.4-0.7 degree range so I reckon splitting the flow between the cpu and gpu would make little to no difference.

fastboy · 31 Jul 2015 at 15:47

thats all i wanted to hear pasty

and regards the bucket your wrong..if that bucket is constantly being filled up as in a closed loop then 2 holes will still empty faster,we are not talking about pressure but flow rate..2 different things yet inheritantly connected.

Deathjester · 31 Jul 2015 at 16:02

The biggest part of this is that you KNOW in a series loop what your flow is going to be.

But if the restrictiveness of your blocks varies (which it does) then you will get different flow over the components.
If this is far enough out of kilter you will massively reduce the flow over one of the components which could cause excessively high temps.

If the blocks are the same then parallel is fine.