Sign in with
Sign up | Sign in
Your question

A question on single vs. double loop systems

Last response: in Overclocking
Share
April 23, 2012 6:25:48 AM

So I read the thread about the experiment where single loop vs double loop systems were compared and the single loop system won out in pretty much all tests. First, has anybody found any evidence to support a counter claim to this testing? I'm just asking out of curiosity.

Second, my concern which I couldn't find many answers to involved having uneven temps between components. If you were to run a loop that went from a big radiator, through the components in series, and than back to the res, the components would be exposed to varying temps of water. For example, if water is 35 C coming out of the rad, leaves the cpu block at 38 C, enters the chipset, leaves this at 40 C, enters gpu #1, leaves at 42 C, enters gpu 2, leaves at 45 C, you in effect are getting different rates of heat transfer for each part as the rate of heat transfer is dependent on the difference between water temp and component temp. Now this might not be an issue in actuality, because if you have a total delta T of 8 C from atmospheric Temp to chip temp, you likely only have a difference of 2-3 degrees from chip temp to water temp which would mean that the diff in temp from component to component is very minimal.

If however, the varying temp distribution of water to chip between various water blocks is significant, would it make sense to run a loop where you have something along the lines of:

Res 1 -> Pump 1 -> Rad 1 -> Cpu -> Chipset -> Res 2 -> Pump 2 -> Rad 2 -> Gpu 1 -> Gpu 2 -> back to res 1

And by doing this you are basically re-cooling the water between chipset/cpu and processor...



Thoughts? :pt1cable: 
a b K Overclocking
April 23, 2012 8:23:20 AM

keep in mind that my delta T is not in check on my system, but the system is clocked, and I would need 6 to 10 rads depending in the fans used, to bring it in cheek. well let just say that that is probably is not going to happen any time soon, so I found that roughing my system, pump, rad, processor, mother board, rad, GPU, rad, memory, and HDD, than back to pump. gave me about a hour more game time while the system was loaded. does that answer your question? the only reason to worry about the routing is in this sort of situation, otherwise if your system was built to have a proper delta T or wattage removal for the heat being created, this should not concern you. because the system will reach it's equilibrium eventually and temps will be constant throughout the system.
m
0
l
April 23, 2012 12:54:48 PM

toolmaker_03 said:
keep in mind that my delta T is not in check on my system, but the system is clocked, and I would need 6 to 10 rads depending in the fans used, to bring it in cheek. well let just say that that is probably is not going to happen any time soon, so I found that roughing my system, pump, rad, processor, mother board, rad, GPU, rad, memory, and HDD, than back to pump. gave me about a hour more game time while the system was loaded. does that answer your question? the only reason to worry about the routing is in this sort of situation, otherwise if your system was built to have a proper delta T or wattage removal for the heat being created, this should not concern you. because the system will reach it's equilibrium eventually and temps will be constant throughout the system.


Sort of, but I have no idea what you mean by your first sentence.
m
0
l
Related resources
a b K Overclocking
April 23, 2012 1:27:34 PM

Quote:
So I read the thread about the experiment where single loop vs double loop systems were compared and the single loop system won out in pretty much all tests. First, has anybody found any evidence to support a counter claim to this testing? I'm just asking out of curiosity.


I don't know if anybody has spent the money on it yet, if that's what you mean ;)  However, it does make sense since the more heat you have in the loop, the more heat transfer you will have if the environment is at the same temperature.

Quote:
Second, my concern which I couldn't find many answers to involved having uneven temps between components. If you were to run a loop that went from a big radiator, through the components in series, and than back to the res, the components would be exposed to varying temps of water.


This doesn't matter too much, since the properties of water barely change below boiling.

According to a steam table program we've been using for my senior design project, here are the thermal conductivities at certain common temps:

@ 30 C - 0.6150 W/mC
@ 40 C - 0.6286 W/mC
@ 45 C - 0.6348 W/mC
@ 90 C - 0.6730 W/mC

Keep in mind that this is extremely low - Copper is ~400 W/mC and Steel is in the 50s. What you really care about is the heat capacity of water, which is on the order of 4000 J/kgC. You won't be getting much of a temperature change unless you're adding more heat than the loop can really handle.

When your pump is running at ~1000 L/hr (4.4 gal/min), your convection heat transfer is able to carry away a lot of the heat conducted through the block, which is why the temps on the other side stay low. Since blocks are made of copper, you're able to get rid of a lot of heat very quickly thanks to that 400 W/mC thermal conductivity.


If however, the varying temp distribution of water to chip between various water blocks is significant, would it make sense to run a loop where you have something along the lines of:

Quote:
Res 1 -> Pump 1 -> Rad 1 -> Cpu -> Chipset -> Res 2 -> Pump 2 -> Rad 2 -> Gpu 1 -> Gpu 2 -> back to res 1

And by doing this you are basically re-cooling the water between chipset/cpu and processor...

The order doesn't have much of an effect in loops because the cycle is short enough, so everything is close to equilibrium. We're also only talking on the order of hundreds of watts of heat - not much to something like water with an incredibly high heat capacity.

If you had something like an industrial size loop, then cooling the water in between would be more efficient, but most likely because you would be boiling the water at some point and would need to condense it.

m
0
l
April 23, 2012 3:41:29 PM

boiler1990 said:
Quote:
So I read the thread about the experiment where single loop vs double loop systems were compared and the single loop system won out in pretty much all tests. First, has anybody found any evidence to support a counter claim to this testing? I'm just asking out of curiosity.


I don't know if anybody has spent the money on it yet, if that's what you mean ;)  However, it does make sense since the more heat you have in the loop, the more heat transfer you will have if the environment is at the same temperature.

Quote:
Second, my concern which I couldn't find many answers to involved having uneven temps between components. If you were to run a loop that went from a big radiator, through the components in series, and than back to the res, the components would be exposed to varying temps of water.


This doesn't matter too much, since the properties of water barely change below boiling.

According to a steam table program we've been using for my senior design project, here are the thermal conductivities at certain common temps:

@ 30 C - 0.6150 W/mC
@ 40 C - 0.6286 W/mC
@ 45 C - 0.6348 W/mC
@ 90 C - 0.6730 W/mC

Keep in mind that this is extremely low - Copper is ~400 W/mC and Steel is in the 50s. What you really care about is the heat capacity of water, which is on the order of 4000 J/kgC. You won't be getting much of a temperature change unless you're adding more heat than the loop can really handle.

When your pump is running at ~1000 L/hr (4.4 gal/min), your convection heat transfer is able to carry away a lot of the heat conducted through the block, which is why the temps on the other side stay low. Since blocks are made of copper, you're able to get rid of a lot of heat very quickly thanks to that 400 W/mC thermal conductivity.


If however, the varying temp distribution of water to chip between various water blocks is significant, would it make sense to run a loop where you have something along the lines of:

Quote:
Res 1 -> Pump 1 -> Rad 1 -> Cpu -> Chipset -> Res 2 -> Pump 2 -> Rad 2 -> Gpu 1 -> Gpu 2 -> back to res 1

And by doing this you are basically re-cooling the water between chipset/cpu and processor...

The order doesn't have much of an effect in loops because the cycle is short enough, so everything is close to equilibrium. We're also only talking on the order of hundreds of watts of heat - not much to something like water with an incredibly high heat capacity.

If you had something like an industrial size loop, then cooling the water in between would be more efficient, but most likely because you would be boiling the water at some point and would need to condense it.


First off, the testing that I spoke of has been done and there's a link to it in the post above yours. Now what I'm about to explain doesn't matter if what rubix said holds true, which I believe it does. While the heat transfer coefficient does not vary much, this does not fix my theoretical dilemma. The actual heat transfer is dependent on the difference in temperature between the GPU block (which is dependent on the GPU temp) and the temp of the water. The total water cooling system is a complicated transient system however when it reaches a somewhat steady equilibrium, there will be CONSTANT temp differences between the water block and the water. Imagine this scenario:

-Water leaves radiator at 35 Degrees
-CPU waterblock is at a steady 50 Degrees
-Waterblock heats water up to 42 Degrees
-Next (GPU) waterblock is 57 Degrees
-Waterblock heats water to 49 Degrees
-GPU #2 Waterblock is at 64 Degrees
-Water leaves gpu #2 at 57 Degrees
-Goes back to res.

So as you can see in a situation such as this everything is in equilibrium but the gpu 2 is 7 degrees hotter than GPU 1. Now I can't overclock gpu 2 as high as GPU 1 b/c it is hotter. This sucks :pfff: 

Fortunately according to rubix though each component in a properly designed system will only raise temps by around 1 Degree.
m
0
l
a c 324 K Overclocking
April 23, 2012 4:41:35 PM

I understand your perception of what will happen, however, due to the specific heat of water and thermal capacity, it is able to absorb and transfer far more heat in the same amount of time than has the ability to otherwise raise temps by anything other than marginal amounts.

Most loops are operating at a flow rate of .75 gpm (low end) to 2.0 gpm (upper end). Therefore, the volume of water moving through your loop far exceeds the amount of heat in watts being dumped into your loop; thus the water does not have the ability to heat up and neither does the copper, chip or PCB of your component. The calculation of this is considered delta-T and is a combination of flow rate of your pump, block restriction, radiator cooling ability based on total volume, fin/tube design and airflow of fans over the radiator. It also is dependent upon your actual ambient room temperature as you can never cool below ambient using normal air or water cooling.

With stock clocks, you should see high 20's C or low 30's C (again, depending on your actual ambient temperature) and at load, anywhere from mid 40's C to mid 50's C for CPU depending on how much you are loading the system and most likely low 40's C for GPUs. Overclocking obviously adds more clock cycles per second, typically requiring more energy, thus producing more heat in watts as a result.

Also, remember that water temperature is not the same as reported temperatures given by RealTemp, CoreTemp or other CPU/GPU monitoring software. They directly impact water temperature, but calculating your total loop output accordingly is key to maintaining a good cooling delta.
m
0
l
a b K Overclocking
April 23, 2012 4:45:11 PM

Quote:
First off, the testing that I spoke of has been done and there's a link to it in the post above yours

I know what you're referring to - I was saying I don't think anybody has bothered to spend the money to combat those tests.

Quote:
While the heat transfer coefficient does not vary much, this does not fix my theoretical dilemma.

But it's the heart of the theory. The water won't conduct heat very well, and it requires a lot of energy to raise the temperature even 1C.

Quote:
The actual heat transfer is dependent on the difference in temperature between the GPU block (which is dependent on the GPU temp) and the temp of the water.

Somewhat, but not entirely. There is a temperature gradient across the block; the whole block is not at the same temperature as the GPU.

Also keep in mind that measured chip temperatures are not the same as the water temperature.

Quote:
-Water leaves radiator at 35 Degrees
-CPU waterblock is at a steady 50 Degrees
-Waterblock heats water up to 42 Degrees
-Next (GPU) waterblock is 57 Degrees
-Waterblock heats water to 49 Degrees
-GPU #2 Waterblock is at 64 Degrees
-Water leaves gpu #2 at 57 Degrees
-Goes back to res.


First off, the water blocks are not that hot in the flow area, since the whole block never reaches the component temperature. The gradient is actually pretty sharp across the block.

In such a short amount of time you won't get the necessary heat transfer to have the water heat up that much. As discussed before, it takes a lot of energy to heat water.

At steady-state, your rads will be removing as much heat as you put in, which is why the temps are relatively constant. Since the net heat transfer to the water is very small, you don't ever put in the ~4000 J/kgK that you need to raise the temperature by 5-10C.

If something like this ever happened in a loop, you'd reach you temp limits very soon and probably would see boiling of your coolant.
m
0
l
a c 100 K Overclocking
April 23, 2012 5:08:19 PM

Pretty much what they said... you would need to have to water moving exceptionally slow for it to actually heat up even 1 degree between entering and leaving a component. In reality the difference is maybe 1 degree from CPU input to 2nd GPU output - and if you have sufficient radiators then that 1 degree gets dropped before going back in.
m
0
l
a c 190 K Overclocking
April 23, 2012 5:19:49 PM

You need sensors actually in the water itself to accurately calculate the temp and as explained in posts above, after a while, the temperature throughout the loop is balanced,
more water means it takes longer to reach equilibrium, but more rads means an increased ability to dump the heat,
and yes, there is such a thing as too much, although I personally ignore it :p 
Moto
m
0
l
!