Completed Flow-Rate testing of my cooling system!

[Jimbo Mahoney]

I've been off work for the past two days with tonsilitis, but I recently received all the bits for my cooling system, so I figured, what better to do with a couple of days than leaktest and flow test the system?



First, for a laugh, a pic:






Now some specs:


Two AQX 50Z's in series @ 13.8v (bench PSU isn't in the shot) - Suction side is 5/8", outputs are all 1/2"

NexXxos XP CPU block

Silverprop Fusion HL GPU block

Custom-made Northbridge block (Thanks Geoff!!)

Heattrap MOSFET block (3/8" tubing because it's so small)

Silentstar Dual HDD cooler

Renault 18 Radiator (not heatercore) w/ DIY Aluminium Shroud and 4 x 120mm Akasa Amber fans.



I'll give figures in lpm to save confusion between US Gallons and UK Gallons, but for reference, US Gallon = 3.785 litre, UK Gallon = 4.546 litre.

Here's a graph of the performance of the NexXxos XP with different flow rates:

(Data from the Procooling review - 71 Watt heatsource).





I applied a Trend Line to the original data to extrapolate theoretical delta Ts for those of you lucky enough to be able to push more than 7 lpm (1.85 US gpm) through it!

Next, a graph of the loss of flowrate due to each component:



The only shortcoming of this graph is that I have not done a flowrate with the pumps alone, so I've used the manufacturer's figure (@12v). I'm running the pumps @ 13.8v, so that 11.3 lpm could be realistic.


Here's some relative flowrate losses:

NexXxos = 39%
Silverprop HL = 13%
Custom Northbridge = 0%
MOSFET block = 8%
Silentstar Dual HDD = 9%
Car Radiator = 4%

Total losses = 42%


With everything all hooked up in series, I get 4.8 lpm (1.25 US gpm) through the whole system. In theory, I would get 4.2 lpm (1.1 US GPM) if the pumps were on 12v.

Do you think splitting the MOSFET and Silentstar into their own branch would increase the flowrate much? They don't need much flow anyway. I'll give it a go if I get a chance.

Hope some of this is interesting / useful for people with similar components.

I also want to extend my thanks to the hardcore Procoolers like Cathar and BillA for doing all the proper calculations; all the people on Proforums who helped with the system component choices.

[Jimbo Mahoney]

OK, I hooked up a parallel flow.


Basically, it puts the MOSFET and HDD coolers in their own branch, allowing the water to take path of least resistance.

This increases the flow through the whole system to 5.7 lpm (1.5 US gpm) BUT it's quite surprising just how little water flows through the restrictive section (the MOSFET and HDD blocks). There wasn't even enough 'oomph' to bleed that section

This is the first time I've witnessed first hand the benefit of having everything in serial flow. The water really does want to flow through the less restrictive path!

In this case, that probably isn't too bad because the MOSFETS and HDDs are the smallest heat loads, perhaps along with the Northbridge.

Just tested - they are getting 2.2 lpm (0.6 US gpm). I think that's plenty for them.

Cool!

[Jimbo Mahoney]

Actually, I'm going to run it all in serial.

Bleeding the system is SO much easier if the flow is high through the whole thing. I get LOADS of air bubbles stuck in the lower flow branch if parallel.

Running in serial will cost me between 0.5'C and 2'C on my CPU temp (according to my calculations), but I'll sacrifice that for an easier to bleed system.