AIO vs Custom loop temp theory

[NorKris]

Hi

 

Im wondering what is making Custom loops beat Aios. 

lets say you have a setup with the same size rad; Aio vs customloop. and same fans. 
What do you guys think would make the custom beat the AIO

[Den-Fi]

Cold plate size/efficiency. Not much use in rad size if the cold plate can't get the heat away from the CPU.

[NorKris]

My theory is that Flowrate(and the 3x tubesize) is the biggest thing 

[For Science!]

4 minutes ago, NorKris said:

Hi

 

Im wondering what is making Custom loops beat Aios. 

lets say you have a setup with the same size rad; Aio vs customloop. and same fans. 
What do you guys think would make the custom beat the AIO

The little details: radiator material, flow rate, fan quality, cpu waterblock fin properties. But all in all, the difference will not be so big on day 1, however the lack of maintainability may mean that 3-5 years down the road the aio may be dead whereas the custom loop will still be going strong provided it is maintained

[For Science!]

Just now, NorKris said:

My theory is that Flowrate(and the 3x tubesize) is the biggest thing 

Flow rate has very little impact on actual performance, tubing size makes no difference to performance.

[Den-Fi]

2 minutes ago, NorKris said:

My theory is that Flowrate(and the 3x tubesize) is the biggest thing 

It is quite literally the most insignificant thing.

[NorKris]

1 minute ago, Den-Fi said:

It is quite literally the most insignificant thing.

 

2 minutes ago, For Science! said:

Flow rate has very little impact on actual performance, tubing size makes no difference to performance.

Yesterday i did a test: using an AIO pump and connected the Rad, ran some stress tests. then i connected an EK copper Rad of the same size and fin density.  1c improvement.

 

it could be the "Cpu waterblock" but i dunno  

[Den-Fi]

3 minutes ago, NorKris said:

 

Yesterday i did a test: using an AIO pump and connected the Rad, ran some stress tests. then i connected an EK copper Rad of the same size and fin density.  1c improvement.

 

it could be the "Cpu waterblock" but i dunno  

I mean... that's margin of error. Could have literally just been a difference in room temp between your test runs.

The biggest factors will be quality of parts. From machining to materials mass market stuff will be made to the bare minimum acceptable spec. Put that up against a super niche Optimus block and the results may be more obvious if the rest of the loop is up to snuff.

[NorKris]

I ran 7 runs for each rad (240 rad), prime95,  5930x 1,350v 
and did Delta T over ambient 

[For Science!]

1 hour ago, NorKris said:

 

Yesterday i did a test: using an AIO pump and connected the Rad, ran some stress tests. then i connected an EK copper Rad of the same size and fin density.  1c improvement.

 

it could be the "Cpu waterblock" but i dunno  

I wouldn’t expect major differences in performance between an aio and a custom loop at an equivalent radiator comparison, so your results are consistent with my line of thinking.

[NorKris]

nice. still think flow rate matter, cuz loop order matters in  "custom aio loop"

[Blue4130]

16 hours ago, NorKris said:

Hi

 

Im wondering what is making Custom loops beat Aios. 

lets say you have a setup with the same size rad; Aio vs customloop. and same fans. 
What do you guys think would make the custom beat the AIO

The amount of water. Aio has 100ml on average, custom has 3 to 5x that. 

[NorKris]

Ok so the water just cools it self?  AutoCooling 

[For Science!]

13 minutes ago, NorKris said:

Ok so the water just cools it self?  AutoCooling 

No, the amount of water only affects time until equilibrium is reached. 

 

15 hours ago, NorKris said:

nice. still think flow rate matter, cuz loop order matters in  "custom aio loop"

Loop order matters in an AIO only because the pump is phenominally weak and so if it wasn't directly above the CPU block it probably wouldn't even push water inbetween the fins of the block. However there is only one loop order in an AIO, pump/block --> radiator and back.

[NorKris]

6 minutes ago, For Science! said:

No, the amount of water only affects time until equilibrium is reached. 

 

Loop order matters in an AIO only because the pump is phenominally weak and so if it wasn't directly above the CPU block it probably wouldn't even push water inbetween the fins of the block. However there is only one loop order in an AIO, pump/block --> radiator and back.

Btw i as sarcastic about that Autocooling stuff

 

Water does not flow in the Fins in a radiator??

Yes AIo's has only 1 "Loop order" but i have made an  AIO-Custom Loop (3 Blocks/Pumps, 3 rads, 1 res) 

[For Science!]

3 minutes ago, NorKris said:

Btw i as sarcastic about that Autocooling stuff

 

Water does not flow in the Fins in a radiator??

Yes AIo's has only 1 "Loop order" but i have made an  AIO-Custom Loop (3 Blocks/Pumps, 3 rads, 1 res) 

Water does not flow thorugh the copper "fins" of the radiator, only down the core tubes. However essentially all radiators are bi-directional, it is equivalent whether you use one port as an inlet or another as the inlet. 

[NorKris]

On 2/28/2021 at 10:50 PM, For Science! said:

Flow rate has very little impact on actual performance, tubing size makes no difference to performance.

I did some more digging. and found this from our German youtuber that has his name on that pc case every1 is buying. And i dont think its so wrong to asume that a aio has 1,5 L/H flow rate? 

 

 

 

i 

[For Science!]

1 hour ago, NorKris said:

I did some more digging. and found this from our German youtuber that has his name on that pc case every1 is buying. And i dont think its so wrong to asume that a aio has 1,5 L/H flow rate? 

 

 

 

i 

His absolute measurements are likely incorrect, a normal pump flow has 4 liter per minute. His conclusions from the video are not wrong though, flow rate has minimal impact on performance.

[bowrilla]

On 3/1/2021 at 12:12 AM, NorKris said:

nice. still think flow rate matter, cuz loop order matters in  "custom aio loop"

It has little influence. As long as it flows and is enough to overcome some restrictions like jet plates (if your block has these) then it's basically fine. 5-7K more or less is not a big thing unless you want to push to the absolute limit.

 

I did my own testing (because der8auer's data raises some questions; his flow rate is unexplainably low at all times for the kind of pump he has, the reading must be either wrong or has some unit mishap, it's not GPM though and GPH is unlikely either) and the results (for my loop) are pretty obvious: flowrate has minimal influence. Between 100% pump speed and 25% pump speed (range of ~4x the flow rate between 168.9l and 41.5l) the maximum corrected difference was ~6K at 100% fan speed. Dropping the fan speed by half to 50% reduced difference to 3K. 

 

Some random Windows process popping up in the background and hogging some resources sometimes has a bigger difference. Keep in mind, my data was the result of 30mins Prime95 runs. A very very unrealistic load. Under every day loads the difference will most likely be smaller.

 

See data and discussion in this thread:

 

[GhostRoadieBL]

On 2/28/2021 at 1:43 PM, NorKris said:

lets say you have a setup with the same size rad; Aio vs customloop. and same fans. 
What do you guys think would make the custom beat the AIO

Fluid composition will have some impact, alcohol/glycol mixtures in AIOs isn't as capacitive as most open loop systems. Mostly due to the additives which make the loop not evaporate through very porous and non-insulated tubing is also making it less thermally conductive.

There's also run size to take into account, every extra inch of tubing and mL of reservoir adds convection exhaust to the loop (it's not a lot but it's a factor) 
Another factor is component quality (99% copper vs 85% alloy copper), copper vs aluminum, fin weld quality and coverage, block quality and flow optimization of the blocks. 

Basically it's cost of components and maintenance after purchase has the largest impacts to performance. If you want easy no refilling for 5+ years go AIO but expect a degree or two difference in common day tasks, If you don't mind filling your loop and can afford the extra an open loop will give better temps but comes with more work and cost. 

[NorKris]

9 hours ago, For Science! said:

His absolute measurements are likely incorrect, a normal pump flow has 4 liter per minute. His conclusions from the video are not wrong though, flow rate has minimal impact on performance.

I don't get it. How can u say minimal?  With low flow rate the temps vent up by 6c. that's more than Alum vs Copper, More than Good cpu block vs a "OK"-one. 

 

Basically same impact as noctua fan vs cheap fan for 50$

[GhostRoadieBL]

36 minutes ago, NorKris said:

I don't get it. How can u say minimal?  With low flow rate the temps vent up by 6c. that's more than Alum vs Copper, More than Good cpu block vs a "OK"-one. 

That was starving the loop if it truly was 1.5 l/h. When I think of that number it's moving a molecule of fluid through the loop on a 20min cycle, that's going to be in the cpu block long past the fluid's thermal saturation point making it more of an insulator than a heat syphon. 

The testing I had in the linked forum post was for finding the minimum pump speed you could run without causing the temps to rise outside of run variance. 

If you starve the loop of flow of course the temps will rise but that's not really going to happen unless you intentionally cause it. I would think (based on the rest of the findings in the video) 1000rpm would be more than enough for the loop pushing the flowrate to around 2+ l/h for a shorter dwell time. 

[NorKris]

4 minutes ago, GhostRoadieBL said:

That was starving the loop if it truly was 1.5 l/h. When I think of that number it's moving a molecule of fluid through the loop on a 20min cycle, that's going to be in the cpu block long past the fluid's thermal saturation point making it more of an insulator than a heat syphon. 

The testing I had in the linked forum post was for finding the minimum pump speed you could run without causing the temps to rise outside of run variance. 

If you starve the loop of flow of course the temps will rise but that's not really going to happen unless you intentionally cause it. I would think (based on the rest of the findings in the video) 1000rpm would be more than enough for the loop pushing the flowrate to around 2+ l/h for a shorter dwell time. 

Ye it was alot of nice info there. Im trying to find out what makes the biggest difference between AIO vs customloop at the same rad size. 

 

does any1 have any idea of what the flowrate of a AIO pump is? 

[bowrilla]

4 hours ago, NorKris said:

Ye it was alot of nice info there. Im trying to find out what makes the biggest difference between AIO vs customloop at the same rad size. 

 

does any1 have any idea of what the flowrate of a AIO pump is? 

The Asetek AiOs do not have jet plates forcing water through the micro fin structure. The result is most likely a lot less flow through the micro fin structure and most of the water just rushing next to it. There will be water in between the fins but it won't exchange much so inbetween the fins the water will heat up but the heat isn't well transfered. That in turn will reduce the surface area that comes in contact with the actually flowing water. Less surface = less heat transfer. The pumps would also be to weak for a jet plate most likely and the construction is further complicated due to the built in pump (with Asetek AiOs at least) to the point at which you can't just place the channels wherever you like. It's not so much the flowrate that's important but the head pressure. The result of large head pressure is potentially more flow but it also means it can sustain flow at higher levels of loop restriction. A jet plate for example means additional restriction. Many pumps can deliver ridiculous flow rates without restriction, but if the head pressure is too low, that flow will collapse quickly. It's similar to fans: static pressure is important to sustain the airflow once you add restriction. A good radiator has ideally more smaller main channels than fewer larger channels as a greater number of small channels in total increase surface area between water and radiator. Smaller channels however increase restriction. The benefit is that (as long head pressure is high enough) the same amount of mass/volume per time needs to go through the channels the faster that mass/volume needs to flow (Bernoulli effect). At a certain flowrate this will create laminar flow. With small enough channels there will not be significant layering so we end up with an even and calm flow of our medium through the channels. The larger the channels get the less is the amount of molecules actually being part of the layer that efficiently transfers heat to the contacting material.

 

This also means that tube diameter will have an effect on flowrate by the way (again: Bernoulli effect): the larger the diametre of your tubing the slower the flowrate. I'd need to run some numbers to come up with an idea of how big that impact is - probably not huge but it should at least be noticeable.

 

Flowrate does play a role BUT only when actually leaving a certain range of flow.

 

I'm personally really tempted to run some more experiments in a more controlled environment with an electric heat source instead of a CPU running some synthetic loads. Alas, at the moment I do not have the extra funds to buy some additional stuff for this. Maybe at a later point.

[NorKris]

Why cant these guys give us specs on their pumps

btw I'm more interested in CM pumps  (they are not asetek right? )