Overclock.net banner
Cancel
Create thread New Forums

What is a minimum acceptable flow rate L/h for a simple water cooling loop?

Jump to Latest
1 - 20 of 22 Posts

Carniflex

· Registered
Joined
·
5,508 Posts
Discussion starter · #1 ·
I'm thinking about making a new system in early summer and at the moment I'm in particular pondering about the pump for the loop. Just 1 CPU and 1 GPU as its a portable system. Up to about 350..375W load, probably single 280mm radiator.

I have been using Alphacool DC-LT pumps for a while with up to 3 of them in a single loop. But I'm not particularly happy with them because of how loud they are at higher rpm's. My current system has 2 of them with plexi tops but pushing them anywhere over approx 2500 rpm makes them too loud even with plexi tops.

So a questions, what pressure/flow rate would you consider to be OK for a simple water cooling loop? I just roughly estimated some numbers:

  • 350W load, 30L/h, 10C water temperature rise when going through blocks.
  • 350W load, 60L/h, 5C water temperature rise when going through blocks.
  • 350W load, 72L/h, 4.2C water temperature rise when going through blocks. (DC-LT + plexi top 2600rpm with zero additional restriction)
  • 350W load, 100L/h, 3C water temperature rise when going through blocks. (DC-LT + plexi top 3600rpm with zero additional restriction)
  • 350W load, 200L/h, 1.5C water temperature rise when going through blocks.
  • 350W load, 400L/h, 0.75C water temperature rise when going through blocks. (Phobya DC12-220 with zero additional restriction)
  • 350W load, 420L/h, 0.71C water temperature rise when going through blocks. (Laing DDC-1T with zero additional restriction)
Alphacool Nexxos 280mm 45mm thick with 1000 rpm fans should be approx 225W for 10C delta T. So - with 350W load I should be hitting approx 15C delta T over ambient air temp. In my impression the CPU/GPU core is approx 20C over the water temp... so that would be putting me already 35C above ambient air temp with the things I'm cooling. Plus the temperature rise from the flow rate. Assuming up to 35C in the summer that puts me already at 70C for the cores (before rise from the low flow rate) and up to 80C if flow is as low as 30L/h and at 55C for the coolant temperature before a temperature rise from going through blocks. DC-LT is rated for up to 65C so I should be fine even with a single 2600 rpm DC-LT? Although it would be probably safer to install it after the radiator.

I have three constraints: (1) noise, (2) size (including fittings, the smaller the better), (3) weight (it is a portable system). Well price is also ofc kind of constraint but the 10 Eur between Laing complete edition (43 EUR) and DC-LT + top (35 EUR) is not a place where to save every penny possible. So if anyone has experience with the Laing and Phoyba DC12-220 I would be happy to hear about them some more. I did search for reviews but perhaps I did not find the right one to make up my mind. I have heard Laing ones are loud and that would be a problem. DC12 looks kind of ok on paper but there is particularly little information available about that - for example I'm not sure in what orientation can it be installed and if it is a reliable pump or develops issues over its lifespan. Also only being rated for 35 C is not realistic - in midsummer it gets that hot already outside, not to mention much hotter inside a PC case.

I'm located in Estonia but normally order my watercooling stuff from Aquatuning.de - ofc if anyone would like to suggest some other store that ships within EU with a reasonable shipping fee and carries a pump that I have not mentioned but should take a look at feel free to suggest it.
 
#2 ·
Quote:
Originally Posted by Carniflex View Post

I'm thinking about making a new system in early summer and at the moment I'm in particular pondering about the pump for the loop. Just 1 CPU and 1 GPU as its a portable system. Up to about 350..375W load, probably single 280mm radiator.

I have been using Alphacool DC-LT pumps for a while with up to 3 of them in a single loop. But I'm not particularly happy with them because of how loud they are at higher rpm's. My current system has 2 of them with plexi tops but pushing them anywhere over approx 2500 rpm makes them too loud even with plexi tops.

So a questions, what pressure/flow rate would you consider to be OK for a simple water cooling loop? I just roughly estimated some numbers:

  • 350W load, 30L/h, 10C water temperature rise when going through blocks.
  • 350W load, 60L/h, 5C water temperature rise when going through blocks.
  • 350W load, 72L/h, 4.2C water temperature rise when going through blocks. (DC-LT + plexi top 2600rpm with zero additional restriction)
  • 350W load, 100L/h, 3C water temperature rise when going through blocks. (DC-LT + plexi top 3600rpm with zero additional restriction)
  • 350W load, 200L/h, 1.5C water temperature rise when going through blocks.
  • 350W load, 400L/h, 0.75C water temperature rise when going through blocks. (Phobya DC12-220 with zero additional restriction)
  • 350W load, 420L/h, 0.71C water temperature rise when going through blocks. (Laing DDC-1T with zero additional restriction)
I don't know where you got those numbers, but they are wrong - way wrong!

The temperature of the water in a look doesn't vary by more that 0.5°c anywhere in the loop, typically 0.25°c

Alphacool Nexxos 280mm 45mm thick with 1000 rpm fans should be approx 225W for 10C delta T. So - with 350W load I should be hitting approx 15C delta T over ambient air temp. In my impression the CPU/GPU core is approx 20C over the water temp... so that would be putting me already 35C above ambient air temp with the things I'm cooling. Plus the temperature rise from the flow rate. Assuming up to 35C in the summer that puts me already at 70C for the cores (before rise from the low flow rate) and up to 80C if flow is as low as 30L/h and at 55C for the coolant temperature before a temperature rise from going through blocks. DC-LT is rated for up to 65C so I should be fine even with a single 2600 rpm DC-LT? Although it would be probably safer to install it after the radiator.

That's wrong too Core temps will be 50-70°c under load no matter what cooling you use.

I have three constraints: (1) noise, (2) size (including fittings, the smaller the better), (3) weight (it is a portable system). Well price is also ofc kind of constraint but the 10 Eur between Laing complete edition (43 EUR) and DC-LT + top (35 EUR) is not a place where to save every penny possible. So if anyone has experience with the Laing and Phoyba DC12-220 I would be happy to hear about them some more. I did search for reviews but perhaps I did not find the right one to make up my mind. I have heard Laing ones are loud and that would be a problem. DC12 looks kind of ok on paper but there is particularly little information available about that - for example I'm not sure in what orientation can it be installed and if it is a reliable pump or develops issues over its lifespan. Also only being rated for 35 C is not realistic - in midsummer it gets that hot already outside, not to mention much hotter inside a PC case.

I'm located in Estonia but normally order my watercooling stuff from Aquatuning.de - ofc if anyone would like to suggest some other store that ships within EU with a reasonable shipping fee and carries a pump that I have not mentioned but should take a look at feel free to suggest it.
Click to expand...
Flow rates of about 1 GPM (~225 L/H) are typical. Beyond that there's diminishing returns.

Pressure? Pressure doesn't matter, only flow rate matters. The only time you use pressure is to calculate flow rates from a pump curve.

A single 280 radiator would be to small. OK for daily computing, but even with 1 GPM+ and high speed (loud) fans won't cool enough for benchmarking or gaming.

I've run a thick (75mm - 2 x 120) rad that had tight fin spacing at 1.2 GPM with 190 CFM fans and it wasn't enough. That was two video cards (67 watt TDP each) and a single core (104 watt TDP) CPU!
 
Save
Reply Quote Rep+ Rep+
#3 ·
A 280 is enough but you will have a higher than normal delta.

The minimum acceptable is .5 GPM,the baseline is normally 1 GPM,easily achievable with any pump on the market today.

A rep begging avatar? Thats a new one......
 
Discussion starter · #4 ·
I got these numbers by taking into account that 1 J/s is equal to 1W. then 4.186 Joule = 1 calories and 1 Calor is the amount of energy needed to increase temperature of 1 gram of water by 1C. Now assuming constant load of 350W and when you know the flow rate then you can estimate the temperature rise as a result of having to soak up that 350W per second heat load. Ofc these are not accurate numbers but they are more a rough estimate based on assumption that the temperature has stabilized. If I have missed something then ofc I am willing to learn - I do not deal with thermodynamics and heat normally. I'm myself more a solid mechanics and numerical simulation guy.

As far as radiator size goes - bigger ones weight more. My aim is to stay under 12kg and the lighter the better. I am willing to accept higher than 10C delta if this saves me couple extra kg. When I finish the case design (its a custom one) and figure out its weight with components included I'll then reconsider the radiator space if I still have some weight budget left to be staying under my target weight budget. My current rough estimate is 11.7 kg for everything.

Now returning the the subject of the present post - I take that approx 0.5 GPM is minimum acceptable which is approx 110 l/h. Meaning DC-LT would be firmly in area below minimum.

So .. About Phobya DC12-220 - anyone knows about this one? The number of reviews seems rather limited and none I have seen seems to really like to talk about this particular pump much. Even the manufacturer site is rather shallow with only three pictures and basically saying that its a pump. What makes me suspricious in particular is the claim that its rated for up to 35C - that is not really a realistic expectation in my opinion for a PC component so I kind of suspect it might be unreliable?
 
#5 ·
Quote:
Originally Posted by Carniflex View Post

I got these numbers by taking into account that 1 J/s is equal to 1W. then 4.186 Joule = 1 calories and 1 Calor is the amount of energy needed to increase temperature of 1 gram of water by 1C. Now assuming constant load of 350W and when you know the flow rate then you can estimate the temperature rise as a result of having to soak up that 350W per second heat load. Ofc these are not accurate numbers but they are more a rough estimate based on assumption that the temperature has stabilized. If I have missed something then ofc I am willing to learn - I do not deal with thermodynamics and heat normally. I'm myself more a solid mechanics and numerical simulation guy.

As far as radiator size goes - bigger ones weight more. My aim is to stay under 12kg and the lighter the better. I am willing to accept higher than 10C delta if this saves me couple extra kg. When I finish the case design (its a custom one) and figure out its weight with components included I'll then reconsider the radiator space if I still have some weight budget left to be staying under my target weight budget. My current rough estimate is 11.7 kg for everything.

Now returning the the subject of the present post - I take that approx 0.5 GPM is minimum acceptable which is approx 110 l/h. Meaning DC-LT would be firmly in area below minimum.

So .. About Phobya DC12-220 - anyone knows about this one? The number of reviews seems rather limited and none I have seen seems to really like to talk about this particular pump much. Even the manufacturer site is rather shallow with only three pictures and basically saying that its a pump. What makes me suspricious in particular is the claim that its rated for up to 35C - that is not really a realistic expectation in my opinion for a PC component so I kind of suspect it might be unreliable?
Click to expand...
Its a rebranded Jingway.
A generic pump that isnt quiet and doesnt have the same performance as Xylem pumps,it is cheap and is reliable tho.....for the price.
 
Discussion starter · #6 ·
Quote:
Originally Posted by B NEGATIVE View Post

Its a rebranded Jingway.
A generic pump that isnt quiet and doesnt have the same performance as Xylem pumps,it is cheap and is reliable tho.....for the price.
Click to expand...
Thank you. That is substantially more information than I managed to find on my own so far trying to google for it.
 
#7 ·
This is some testing of the Jingways.

http://skinneelabs.com/jingway-pumps

Your numbers for watts and water heating are good too, but you know that
smile.gif


I have a small system running at the moment that cools a small socket Intel chip and a 980ti with a single PE240 rad. While gaming with 1400rpm fans the water gets up to 15c over ambient but that is still good low hardware temps. With one layer of 2000rpm fans the coolant temp gets down to 10c over ambient.
 
Discussion starter · #8 ·
I decided to double check my rough estimate about the temperature rise as a result of thermal load and look at it in a more granular way. In addition I started wondering if with low volume loop its possible that the "same water volume" passes through the block twice per second (as that would mean that the water gets to give away some heat when going through the radiator). Turns out that one would need quite spectacular flow rate for that to be the case even with a relatively small/simple loop of 2 blocks, 1 radiator and 1 meter of 10mm ID tubing.



Seems like indeed, that the suggestion of aiming for at least 0.5 GPM is well founded and above 1 GPM the rewards for increasing the flow rate are not particularly impressive anymore. If the spreadsheet I whipped up quickly does not have any serious fundamental misunderstandings in it I would say that as low as 0.25 GPM is still not catastrophic although ofc not really desirable point to be in either. My current setup should be sitting marginally above 0.25 GPM (2x DC-LT @ ~2500 rpm) and while the temperature for the GFX card is not very good it is sort of acceptable.
 
#9 ·
Exactly.

Pump requirements are massively overrated by most.
 
#10 ·
Couldn't agree more. Nice to see some sensible discussion
wink.gif


Temperatures clearly aren't the same all the way round a loop like a lot claim, if that was the case then there is no heat transfer! It's just with the overrated flow the difference becomes small as shown above.

A loop does reach an equilbrium of sorts but that's for the average water temperature - which will go up until the rad is able to lose as much heat as the components are putting in.

I'm probably below 0.25gpm with my single DC-LT at reduced speed. I think having the cpu before the gpu helps make this work okay (the gpu being the highest wattage but generally easiest to keep temps down).
 
#11 ·
You're digging waaay to deep in this
smile.gif

Dont get me wrong, it is GOOD to work out & know these things, but dont get too fixated on numbers.. you'll only drive yourself crazy and the end result is never what you calculated anyways.
Also, loop order isn't that important as long as you keep the reservoir or T-line before the pump. Eventually the loop evens out and it no longer matters if the CPU puts more or less watts as the GPU into the loop, you just need a radiator that gets rid of it all. Don't look at it as a "snapshot", look at it as a movie, dont look at the individual frames. Trying to work the plot of a movie by looking at the individual frames (24/sec) will drive you crazy.
smile.gif


Of course, if you want to really optimise, you need to put the lowest heat producing components first in your loop... but we really are talking fractions of percentages here.... and even the most efficient waterblock only picks up 1-2 degrees maximum on each pass... and that is a snapshot.. dont think in snapshots
smile.gif


Also, watch out with minimum speeds... while most of the time our loops are not that restrictive.. blocks, barbs, fittings, hose diameter, 90 degrees couplings, etc it all adds up.. tiny amounts maybe, but still.

And with flow below 0.25 GPM you may find yourself in the situation that you cant get rid of even the smallest airpocket or bubble or fail to have enough "velocity" for your CPU block to work at its top efficiency.

For filling, draining & purging it can be handy to have near/over 1 GPM... then, once you loop is "clear", you can drop to a more meandering 0.5 GPM.

But i wouldn't let it drop below 0.25 , it doesn't make me comfortable... i mean i like an idyllic, meandering stream/river like the next guy, but i dont like a pond where the surface gets broken by the gasbubbles rising from the rotting fauna & flora at the bottom of the pond. But hey, that is just me
smile.gif
 
Save
Reply Quote Rep+ Rep+
#12 ·
Quote:
Originally Posted by RnRollie View Post

You're digging waaay to deep in this
smile.gif

Dont get me wrong, it is GOOD to work out & know these things, but dont get too fixated on numbers.. you'll only drive yourself crazy and the end result is never what you calculated anyways.
smile.gif
Click to expand...
Pretty much this.

Also,you dont need to have a res in front of the pump but it is good advice as it makes it easier to fill without dry running the pump.
 
Discussion starter · #13 ·
Oh I do not plan to spend weeks fretting over 1C difference between one and another option. I started the thread for figuring out what is the minimum acceptable number for a real flow rate (for baseline I'm just assuming atm very roughly 10% drop flow drop per block and 5% per radiator) and I ended up with the same approx 0.25 GPM real flow rate. Which means that I could get away with as little as 1x DC-LT pump, however, something little stronger would be clearly desirable. Also DC-LT (at least the 3600 rpm version why seems to be no longer in production) is loud as hell at full speed.

Now I'm already thinking about weight and price and size. Atm I'm looking at


The EK SPC-60 looks like a well balanced very compact thing and the max 50C temp rating is approximately what I would end up with if the temps hit 35C in the midsummer for few days. I know EK has also quite reasonable shipping fees as I have ordered couple of pieces from them directly after Aquatuning stopped providing their stuff. Phobya DC12-220 is cheap and on paper with good stats, however, temp rating of only 35C is a bit worrysome and I'm not particularly fond of their mounting solution. And then there is couple of Laing's which are also compact enough to fit where I'm thinking of putting my pump. D5 variant I'm not really considering, I just threw that one in there for reference.

This is the current work-in-progress draft of the case where this thing would go

The pump in question would go into the front bottom. I would have up to 47 mm height wise to the second GFX card if I would use one (I'm atm planning to stick to only one). I'm not planning to have a separate res this time as the radiator itself would act as one, being directly before the pump intake and located higher than the pump.
 
#14 ·
without a res or T-line, filling and purging is going to be something which requires a LOT of patience
smile.gif
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #15 ·
Quote:
Originally Posted by RnRollie View Post

without a res or T-line, filling and purging is going to be something which requires a LOT of patience
smile.gif
Click to expand...
It should be more or less fine. My current setup is similar to what I'm planning.

Pumps (2x DC-LT currently) right after the radiator at the bottom. In the new setup I'm planning I would just skip the tube res above the radiator. Although I must note it took me a while to get the damn bubbles out that liked to rattle around in there. The setup is light enough to be moving it around while bleeding to dislodge some particularly stubborn bubbles.
 
Discussion starter · #16 ·
I think I'll probably end up going with Phobya DC-12 220 (Jigway DP-400) and throw in a standalone fan controller to be able to drop it somewhat down should it turn out to be a whiner. With just 1 CPU + 1 GPU I should be ending up at approx 1 GPM should its noise level turn out to be acceptable for that and it should be (hopefully) better than going with single DC-LT.

I'm not a fan of the mounting solution (double sided tape basically or two brackets) and having the pump rated only for 35C is not realistic either. I'll do a little more research into it ofc before I order the next batch of parts for my project in few months.

Thanx for the help guys. Got some pretty good hints out of this thread.
 
#17 ·
I have no idea what I get for actual flow rates but at idle, it has to be extremely low (I would think far less than .25gpm but I really dont know). Most people suggest using water temp to regulate pump speeds but I have mine plugged into the CPU header on the motherboard. It typically runs at less than 1000RPM and makes very little noise (slight ticking is all that is audible) but I have quite a bit of radiator surface area relative to components I'm cooling. At idle or while browsing the internet, the radiator fans typically stay off (two front mounted 140mm fans at extremely low RPM are enough to keep it cool).

Edited to make that a bit easier to read.
 
#18 ·
Quote:
Originally Posted by Carniflex View Post

... started wondering if with low volume loop its possible that the "same water volume" passes through the block twice per second (as that would mean that the water gets to give away some heat when going through the radiator)./quote]And what happens when the water passes through loop every 2 seconds?
Click to expand...
Your calculations don't take into consideration the effect of time spent in the heat transfer device (block or rad) and the effect that time has on:
1: Delta T between the heat transfer surface at the beginning of the device to that at the end of the device. The less time in the device the greater the Delta T (of the surface to the water) at the end.
2.) Turbulence.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #19 ·
Quote:
Originally Posted by billbartuska View Post

Your calculations don't take into consideration the effect of time spent in the heat transfer device (block or rad) and the effect that time has on:
1: Delta T between the heat transfer surface at the beginning of the device to that at the end of the device. The less time in the device the greater the Delta T (of the surface to the water) at the end.
2.) Turbulence.
Click to expand...
That is true. My estimate is basically just conservation of energy assuming perfect heat transfer from the heat source to the water. For time I'm considering constant load in time i.e., watts per second against constant perfectly mixed flow in milliliters (cubic centimeter of water) per second. The points you mentioned do make a difference in reality and that is why some blocks are better than others or have the best performance at some given flow rate. The goal here was more of a ballpark figure from napkin corner estimate and not solving the full 3D Navier Stokes system with the exact geometry of the loop.

The "loop time" in the table is the time it takes for the entire content of the loop to circulate once at a given flow rate.
 
#20 ·
Quote:
Originally Posted by Carniflex View Post

The "loop time" in the table is the time it takes for the entire content of the loop to circulate once at a given flow rate.
Click to expand...
...a brave assumption.

Nice work anyway, more people should be doing these types of calculations instead of "One rad face for each block plus one extra."
 
Save
Reply Quote Rep+ Rep+
1 - 20 of 22 Posts
About this Discussion
21 Replies
8 Participants
Last post:
Carniflex
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling