Open Loop Orientation Doesn't Matter.

[KhaderKh]

Hi,

 

An unexpected experiment resulted in an expected results.

 

I recently did an open loop cooling for my 9700K 5.2ghz in the traditional EKWB way.

 

Pump - Radiator - CPU - Pump

 

Which makes sense to have the CPU get cool water from the radiator.

 

Yesterday I opened the loop and cleaned everything and by mistake I assembled the loop like this

 

Pump - CPU - Radiator - Pump

 

Literally zero difference if not better cooling in the 2nd method but within the margin of error.

[jaslion]

49 minutes ago, KhaderKh said:

Hi,

 

An unexpected experiment resulted in an expected results.

 

I recently did an open loop cooling for my 9700K 5.2ghz in the traditional EKWB way.

 

Pump - Radiator - CPU - Pump

 

Which makes sense to have the CPU get cool water from the radiator.

 

Yesterday I opened the loop and cleaned everything and by mistake I assembled the loop like this

 

Pump - CPU - Radiator - Pump

 

Literally zero difference if not better cooling in the 2nd method but within the margin of error.

Yeah that is pretty normal. You have one item introducing heat so the water will pretty much always be at the same temp no matter the config.

 

The moment it starts mattering is when you have multiple heat sources as the heat from one will travel to the other so if you have lets say CPU -> GPU -> 360 -> 240 rad config the cpu's heated water whilst not much hotter will end up at the gpu raising temps a little. Having it pass through either of the radiators before it reaches the gpu would result in cooler water thus a bit lower temps.

[rafe_28]

Why should it be elsewise? There isnoke heat source, the cpu, o where is the water gonna heat up on its way?

[Eigenvektor]

JayzTwoCents made a video about that some years ago:

 

[jaslion]

15 minutes ago, rafe_28 said:

Why should it be elsewise? There isnoke heat source, the cpu, o where is the water gonna heat up on its way?

In the end it matters little to nothing as the water reaches it's equilibrium in a fast moving environment either way.

[bowrilla]

While there is a delta between radiator in- and outlet, that delta is pretty low and it depends on the flowrate. The lower the flowrate, the higher the delta. The actual delta depends on the specifics of each loop however expect something in the range of ~0.2-2K difference in water temps. That delta does not directly result in the same difference on your CPU block. Of course, having a little bit colder water will result in a more efficient heat transfer between CPU block and water, however the bottleneck is not really the transfer between block and water but both between Die and IHS and between IHS and block (smallest contact area in both cases). 

 

The impact of flowrate on the delta between in- and outlet does not mean, that slower flowrates are better (or vice versa for the blocks). Overall the total amount of heat energy being transferred is about the same. Basically (and very simplified), if you'd trace a water molecule through the loop it might remain in contact with the radiator surface for a shorter amount of time each round when flowrate is high, BUT it will do more rounds in the same amount of time as it would with a lower flowrate. The total amount of energy that is being dissipated through the radiator will be basically the same.

 

Of course, the type of flow (i.e. turbulent vs laminar) has an impact but our regular pumps will usually not easily make flow switch between those states. It is more dependant on the design of the specific component.

 

The difference in CPU (or GPU) temps will be marginal and within the margin of error. I'd dare to say a random Windows background process will have a larger effect on your temps and regulating/controlling your ambient temperature will be more beneficial in real world applications.

[rafe_28]

Well i actually tested this the other day myself with gpu temps and had a 3 C increase in hotspot with the pump at 40% over the pumpat 100%. So it latter quite a bit. The average core temp barely changed tho.

 

Edit: tested with a msi 5700XT gaming X (198.8W average durikg testing) on a 420mm ekwb CE. So thats a really t h i c c rad and the dekta should therefore be even higher right? Cuz ot can handle sudden temp changes better

@bowrilla

Edited June 4, 2021 by rafe_28

[bowrilla]

20 minutes ago, rafe_28 said:

Well i actually tested this the other day myself with gpu temps and had a 3 C increase in hotspot with the pump at 40% over the pumpat 100%. So it latter quite a bit. The average core temp barely changed tho.

Values like that have no meaning without values for ambient and coolant temps. A difference of 1-2K for temp spikes is entirely negligible and within margin of error. Even more so if you're comparing absolute temps and not relative temp differences.

 

24 minutes ago, rafe_28 said:

Edit: tested with a msi 5700XT gaming X (198.8W average durikg testing) on a 420mm ekwb CE. So thats a really t h i c c rad and the dekta should therefore be even higher right? Cuz ot can handle sudden temp changes better

Thick radiators scale well with fan speed but they also usually need more static pressure to properly work. Thinner radiators (with the same fin density) work better at lower fan revs but don't scale up. Which, for most people isn't that big of a deal since most people prefer their fans to be quiet.

[rafe_28]

11 hours ago, bowrilla said:

Values like that have no meaning without values for ambient and coolant temps. A difference of 1-2K for temp spikes is entirely negligible and within margin of error. Even more so if you're comparing absolute temps and not relative temp differences.

I sadly have no coolant temp sensor but ambient stayed between 24-24.2°C duribg my testing. And it wasnt just rabdon Spikes. I measured the average temp using gpu Z after thr temos started climbing  

[Luggage]

On 6/4/2021 at 7:29 PM, bowrilla said:

While there is a delta between radiator in- and outlet, that delta is pretty low and it depends on the flowrate. The lower the flowrate, the higher the delta. The actual delta depends on the specifics of each loop however expect something in the range of ~0.2-2K difference in water temps. That delta does not directly result in the same difference on your CPU block. Of course, having a little bit colder water will result in a more efficient heat transfer between CPU block and water, however the bottleneck is not really the transfer between block and water but both between Die and IHS and between IHS and block (smallest contact area in both cases). 

 

The impact of flowrate on the delta between in- and outlet does not mean, that slower flowrates are better (or vice versa for the blocks). Overall the total amount of heat energy being transferred is about the same. Basically (and very simplified), if you'd trace a water molecule through the loop it might remain in contact with the radiator surface for a shorter amount of time each round when flowrate is high, BUT it will do more rounds in the same amount of time as it would with a lower flowrate. The total amount of energy that is being dissipated through the radiator will be basically the same.

 

Of course, the type of flow (i.e. turbulent vs laminar) has an impact but our regular pumps will usually not easily make flow switch between those states. It is more dependant on the design of the specific component.

 

The difference in CPU (or GPU) temps will be marginal and within the margin of error. I'd dare to say a random Windows background process will have a larger effect on your temps and regulating/controlling your ambient temperature will be more beneficial in real world applications.

It’s perhaps within margin of error but for even the potentially 1-2K gain - with 3 rads and 2 GPUs I put one rad between each component, because why not, it will at least not perform worse than a random order.

Also if you just want to optimize for one thing perhaps take into account that CPU block are more dependent on good DeltaT than GPU blocks, in general.