No competition, why only HWlabs?

[Cathar]

Originally Posted by aaronspink

What's the price range for it?

Customised it's costing me $150 Australian, or around US$105.

This is Australia here - we don't have heater-cores or radiators that cost less than ~$60US at an absolute minimum.

It is an actual car-engine radiator despite it looking just like a double-sized heater-core - apparantly for a 1000cc motor, or at least that is what I was told. I'll get them to dig up the official specs, including the car model/year that it came out of.

However, as seems to be the case with cars outside of the USA, the heater-cores and radiators we use typically come from different manufacturers and that again is why these things seem to be hard to track down in the USA.

[Cathar]

Okay, the radiator is from a Honda "Life" 600cc car, which is sometimes called a "City". The radiator is in the book as an "HR920". That's all the information that they had on it as it's apparantly a very old radiator.

Now that I have it, it actually has 13FPI. The core area itself is 1⅛" thick (~29mm). Exactly 300mm wide in the core area, and exactly 240mm high.

Here's a piccy of it. The tanks on it are customised cut-down tanks from a different car radiator. It's set up in a single-pass configuration and let me say that it has exceptionally low flow resistance.

[Etacovda]

looks very interesting. I wonder how much i can get hold of them for here - I'm pretty lucky, i get the camry BA cores for about 25NZD (whats that, 20AUD?), and most other cores for about $15.

[Cathar]

Originally Posted by Etacovda

looks very interesting. I wonder how much i can get hold of them for here - I'm pretty lucky, i get the camry BA cores for about 25NZD (whats that, 20AUD?), and most other cores for about $15.

Second hand though...?

[Etacovda]

Of course, I assumed that Bravo's BA rads were used as well. New, theres no way they could even be made for that, I'd assume.

Out of interest, is the City (they're all called Citys' over here) Turbo rad any larger?

[Cathar]

I posted this at OCAU but thought that the guys here might be interested as well:

Also stuck a pair of Panaflo L1A's onto the radiator box to replace the old ThermalTake fans, for which they all seemed to have some "knocking" noise whenever I tried to run them at lower voltages.

You can see the L1A's PQ curve here:

http://www.panasonic.com/industrial/...pdf/fba12g.pdf

The Panaflo L1A's are quietish at 12V. If held in one's hand they are very quiet, but when mounted or sitting on something, transmit a fair out of bearing vibrational noise. I've mounted them very firmly inside my radiator box and this seems to have stopped most of their vibration. After running them overnight they seem to have quietened down a bit further - presumably just from breaking in a bit.

12v fan test

Anyway, at 12V on the pair of fans, sucking through the radiator, and runningBurnK7 @ 2.7GHz/1.90v, and the two Laing D4's also running at 12V dumping a combined ~32W of heat into the loop, I'm seeing the water temperatures stabilise at 2.3C above radiator intake.

At a rough guess of the CPU dumping ~80-90W of heat into the loop at this level, I'm roughly estimating that the radiator C/W is sitting around the 0.019-0.021 mark, which for a pair of low-speed low-noise fans I feel is a pretty good achievement, and certainly good enough to consider using fans like the Papst 4412FGL's even at lowered speeds for the ultimate in silent but powerful radiator based cooling.

7v fan test

At 7V, the L1A's are very weak, pushing at best about 40CFM and at best 13Pa of pressure at no air-flow. Certainly a whole lot weaker than Papst 4412FGL (about 2/3's the strength). You can see the Papst fan range's PQ curves here:

http://www.dangerden.com/images/fans/papst_spec.pdf

Water temps have stabilised at 3.4C above in radiator intake temperature.

This works out to a C/W of around 0.028-0.031.

So what's this all tell me (apart from that the radiator cools excellently even with very weak fans and very low noise)?

Extrapolating from what I'm seeing it tells me that it should be possible to build a purpose built PC cooling radiator that is:

~16x16cm in size for the core
14FPI
1" thick in the core
single pass
coupled with a single Papst 4412FGL (26dBA fan) sucking on the core via a shroud

and achieve a C/W of slightly less than 0.05, for basically the equivalent performance of the Thermochill 120.2 using a pair of the stronger and noisier 4312L's.

i.e. a silent, compact, and high performing radiator solution that for a water-cooling setup with say a single pump and a hot overclocked CPU running most anything should never really see water temps exceed more than about 4C above ambient.

So how's that for a pointer for a step in the right direction for PC radiator improvements?

What I would REALLY love to see now is a case laid out like the Lian Li V1000 made with a cavity at the bottom to mount two such radiators side by side, being for the "average" water-cooler who would install a single radiator, and for the "nutter overclocker" who would install two radiators, all still in relative silence.

[psychofunk]

Bro......you don't just kick ass, you kick big fat gigatic badunkadunk...

[Cathar]

I ran the L1A's at 9v, which near as I can extrapolate would pretty closely approximate the Papst 4412FGL's ar 12v, and got a radiator C/W of around 0.022-0.024. Performance is clearly dropping away pretty rapidly when using fannage much weaker than that - but this would be expected as 12cm fans don't usually come much weaker than the FGL's anyway.

So yeah - definitely possible to build a radiator that should achieve very good cooling performance in a quiet and compact form factor so long as the radiator is designed to match the fan, and not the other way around which is what we have today where most of us have to get the right fan to match the radiator, and most usually, even when using the dual fan sized radiators they still require somewhat noisy fans to get to what could be considered a highish-end level of cooling.

[Les]

Good stuff as usual.
Think the Papst 4412FGL projection may be a trifle optimistic.



Edit :
Error in radiator simulations - so accustomed to water-flow sims, I, stupidly, shortcut calcs and assummed dP=k*Q^2.
Corrected graph:

[Cathar]

Les, thanks for the graphs.

My thinking was as follows:

For the 300x240x29mm radiator I was personally approximating fairly close to a 10Pa drop at 3m^3/min, or for 160x160x25mm, around about 1.25m^3/min at 10Pa, or about 6.5Pa at 1m^3/min. So maybe 1.1m^3/min with the Papst.

C/W of air at 1.1m^3/min ~= 0.044

Bill's tests of the Thermochill 120.2 appear to imply a radiator efficiency in the very high 90's percent for air-flow rates of less than 1.0m^3/min and flow rates of at least 1gpm.

Assuming radiator efficiency of 90% @ 1.1m^3/min => C/W of 0.048-0.049

Ultimately it comes down to whether or not I was right with the 10Pa @ 3m^3/min assumption/extrapolation in the first place though. Core is thinner and less densely finned than the Thermochills, so I applied some extrapolation there to arrive at the large radiator's pressure drop by using Bill's PQ graph for the 120.2. Bill suggested 10Pa for the 120.2 at 1m^3/min. 2.5x the surface area immediately implies 2.5m^3/min at 10Pa, and I applied an arbitrary 20% bonus due to the more free-air flowing characteristics of this radiator.

[j813]

I thought these new Honda radiators are Alum made? So not all new Honda Cars are not equiped w/ Alum Engine Blocks?

[Cathar]

Originally Posted by j813

I thought these new Honda radiators are Alum made? So not all new Honda Cars are not equiped w/ Alum Engine Blocks?

This particular radiator is apparantly from an early 80's car. Really I tried hard to get more information from the radiator guys, but after I while I got the impression that perhaps I was asking for information that they simply did not know.

[j813]

Stew
What Fan(s) are you(will be) using w/ that?


Thanks for the info

[Cathar]

Originally Posted by j813

Stew
What Fan(s) are you(will be) using w/ that?

Will be using the Panaflo L1A's, as tested/reported on 8 posts above.

I see no need for me to consider using stronger fans. I am seriously considering grabbing a pair of Papst 4412FGL's though just to make the fannage even quieter. As tested, if I ran the Papst's, I am unlikely to ever see water temps of more than 3C above ambient no matter how hard I overclock this CPU and what loads I run on it.

Problem I have now though is that presently the 2 x Laing D4's sitting inside the 1/2" thick MDF radiator box are presently the noisiest things in my system even when run down to about 9v or so. Below that the Panaflo's and the (necessary) 2500RPM 8cm case fan that keeps the system and video card cool takes over as the noisiest thing.

[Roscal]

New BIX III and BI III :

[psychofunk]

Originally Posted by Cathar

2500RPM 8cm case fan that keeps the system and video card cool takes over as the noisiest thing.

With regard to the above you may want to check these out. Super quiet and yes they push no air at all

http://www.nexustek.nl/80mmcasefan.htm

I have only personally tried the 80's but they also have some 120's but I don't really feel the need for quieter than FGL's although you friend that you mentioned earlier may be interested

http://www.nexustek.nl/120mmcasefan.htm

[jlrii]

This has been bugging me for a while as I am doing the same sort of thing (low velocity) with a completely different setup tho'. Is there a boundry layer involved when dealing with air throug a rad? From working in HVAC for quite some time I know there is an air film to deal with when doing load calcs for buildings. I haven't found any info at all on how air movement effects it in a case like this. Since the atmospheres "viscosity" is so low, does that negate the effect with even slight air movement over a rad? Air impingement if you will.

Edit: From what I've read in rad literature the creating "waves" in the fin surfaces is done to increase efficiency, Is this just for increased surface area or to create turbulance?

[Les]

Originally Posted by Cathar

My thinking was as follows:

For the 300x240x29mm radiator I was personally approximating fairly close to a 10Pa drop at 3m^3/min, or for 160x160x25mm, around about 1.25m^3/min at 10Pa, or about 6.5Pa at 1m^3/min. So maybe 1.1m^3/min with the Papst.

C/W of air at 1.1m^3/min ~= 0.044

Bill's tests of the Thermochill 120.2 appear to imply a radiator efficiency in the very high 90's percent for air-flow rates of less than 1.0m^3/min and flow rates of at least 1gpm.

Assuming radiator efficiency of 90% @ 1.1m^3/min => C/W of 0.048-0.049

Ultimately it comes down to whether or not I was right with the 10Pa @ 3m^3/min assumption/extrapolation in the first place though. Core is thinner and less densely finned than the Thermochills, so I applied some extrapolation there to arrive at the large radiator's pressure drop by using Bill's PQ graph for the 120.2. Bill suggested 10Pa for the 120.2 at 1m^3/min. 2.5x the surface area immediately implies 2.5m^3/min at 10Pa, and I applied an arbitrary 20% bonus due to the more free-air flowing characteristics of this radiator.

Agree with all.
Updated graph - gone the whole hog on PD (29mm v 50 mm thick) ie 5.8Pa @ 2.5m^3/min :



Edit :
Error in radiator simulations - so accustomed to water-flow sims, I, stupidly, shortcut calcs and assummed dP=k*Q^2.
Corrected graph:

[BillA]

very nice Les
and good calcs too Cathar, anyone should understand

[HammerSandwich]

I'm confused by the double curves, Les. Is that a 29mm rad and a 50mm rad in each size?

[Les]

Originally Posted by unregistered

very nice Les

Yes, I thought was rather pretty

Originally Posted by HammerSandwich

I'm confused by the double curves, Les. Is that a 29mm rad and a 50mm rad in each size?

They are example radiators
The radiator P/Q curves are actually Kryotherm calculated for 14 FPI radiators with 0.7 and 0.95mm fin thicknesses
Have modified graph to include details.

[HammerSandwich]

Thanks, much clearer.

[Les]

Made a bish of the radiator simulations
Have corrected.

Some, less pretty but more typically Les, graphs :

[Cathar]

Les, from 2nd hand information given to me, the Thermochill 120.x rads are 24FPI and 32mm thick in the core, with 8mm in-between each flat tube. Tubes are 2mm wide (so 10mm from tube center to tube center).

In comparison, the Honda radiator is 13-14FPI, 29mm thick with 10mm in-between each flat tube. Tubes are 2mm wide (so 12mm from tube center to tube center).

Just wondering how that all fits in with matching BillA's data to the air-flow PD's you're calculating with the other rads. I assume that you've probably already made some adjustments?

[Ruiner]

Hmmm. So how long until someone markets a rad of thickness/fin density of that honda unit, but in PC sizes (single and double 120, with shrouds and 1/2" barbs)?
Has anyone researched using watercooled motorcycle radiators?