How much better do radiators perform when hot?

JonJ678

JonJ678

JonJ678

Soldato
Joined
22 Dec 2008
Posts
10,369
Location
England
Somewhere on here there's performance curves for a thermochill radiator, I couldn't find them but have found the following graph from here, the delta in question is a 2100rpm 3.4mm H20 model according to here and 25mm thick according to various shops. 3.4mm H20 in sensible units is 33Pa which is pretty high. Testing methodology looks sound, and is better than what I could achieve in a garage so I'm inclined to believe it.

2efn68g.jpg


So, that's the performance at 10 degrees air-water delta. With considerable effort I can probably work out how to scale this with airflow, but I'm unsure what to do about temperatures. It's tempting to say the wattage dissipated is directly proportional to the air-water temperature difference. I know this is untrue, but I don't know how poor an approximation it is.

I'm hoping someone on these boards can point me in the direction of either test data which I can work from, or theory which I can adapt to this purpose. Otherwise I'm going to have to draw a radiator in Ansys, and I really don't want to. The temperature range I'm interested in is room temperature to approximately 80 degrees, Laing ddc rate their ddc up to 60 degrees fluid temperature. I'm yet to persuade anyone to tell me what temperature their radiator will fail at, so I'm stuck assuming it's hotter than the pump fails at.

Thank you.

(p.s. the answer to the obvious question of "why do you want to know" is to check the idea of using peltiers as a means of reducing number of radiators required to cool a computer for application to m-atx. One loop above ambient, without radiators, the second far over ambient with some radiators, pelts joining the two loops).
 
Have you looking at skinneelabs data - there is a semi recent test of all the main triple rads.
He tests at 10c delta, 5c delta and 2c delta.
Not sure if that what you need - but should help answer the principle of the question.
 
Hm. The link to the article I think SS has in mind is here.

While it doesn't answer the question, it does tell me that skinnee labs reckon a constant C/W figure, so wattage dissipated linearly related to temperature delta, is valid. They optimistically state it to eight decimal places then multiply it up to get 2/5/10/15 degree values. So, if they're certain it's linear, I can probably work to the assumption that it is. Really, really good website that one.

Thinking out loud.
Feser 360 with fans at approximately 1400rpm predicts a C/W value of 0.02, so taking a 120mm radiator as a three times this, and an arbitrary heat load of 300W we get a water temperature of 18 degrees (above ambient). Maximum allowable hot loop water temperature is 40 degrees (above ambient). Realistically taking peltier array at COP=2, d=20 and optimistically ignoring inefficiencies suggests a 450W heat load into a 120mm radiator leads to hot water temperatures of 27, cold water temperature of 7. 7 << 18.
 
Last edited:
Not just that, it's using a testing methodology which I agree with and which is described in detail. Fantastic really. In case you missed it in the OP, this pdf is similar though equally compelling reading.

I know you're mostly interested in silent computing so forced convection wont appeal much, but as a passive cooler I imagine you have a keen interest in radiator efficiency with temperature. I think there's definite potential in this idea.
 
Skinnee (and Vapor) are very highly regarded - due to methodology and subject matter.
I'm guessing both have a science/engineering background as there reviews are simular to pear reviewed scientific papers (but thankfull not so dry).

as to efficiency, that I am - and coolant temp has a massive effect on radiant rads, which are my new play thing.
If only the PA120.3 (old) wasn't soooo good with very little airflow, I would have posted some results buy now.
 
Last edited:
How would a radiator perform better when hot? I thought the idea was to keep the radiator running cool

ps. i know nothing about water setups
 
Forced convection (molecules passing energy from one to another)
Is what 98% of the members here use - fans cooling rad, that cool water, that cools the blocks/hardware.

Passive/natural convection (molecules again)
As above but with no, or no direct forced airflow - What me and the other 2% run

Radiant watercooling (waves - infared radiation)
Unaffected by ambient temps - the higher the temp the panel/rad the more effecient energy lossed from the system.
Higher rad temp = higher cpu temp = not ocing friendly
This is what I'm playing with at the minute. It's like and underfloor heating system in reverse.
 
Last edited:
Perform better in terms of watts dissipated, not in terms of the water inside the radiator. If you have a conventional loop you want the water as cool as possible, unfortunately this makes the radiators very inefficient so you need loads of them.

The watercooling community call these things radiators. They're not, in any sensible sense, but nonetheless that's what we call them. SS is playing with an actual radiator, something which gets rid of most of its heat through radiator. The convective things I'm using probably have a proportional relationship with temperature, his will be to the power four. What's the C/W value like for your radiant setup?

Regarding the tec idea in general, I believe I've put together an initial model for the dynamic response. It's very heavily damped and ignores a distressing number of things, but initial results are promising nonetheless.
 
Last edited:
JonJ678 said:
(p.s. the answer to the obvious question of "why do you want to know" is to check the idea of using peltiers as a means of reducing number of radiators required to cool a computer for application to m-atx. One loop above ambient, without radiators, the second far over ambient with some radiators, pelts joining the two loops).
Click to expand...
Interesting plan.

A cold loop and a hot loop with the peltier forcing heat transfer between them.

But by using pelts you roughly double the cooling requirement after they chuck in their heat output. Also increases the power demand for the PSU.

The component list sounds big for a mATX, might have to plan for a larger case.
 
It's just the psu which takes a real kicking. At a cop of one which is probably sensible you'd double the cooling requirement. If your loop used to be at 10 degrees above ambient, and it's now at 30, your radiator is thrice as efficient as before and so you actually need fewer radiators. With some care I'm pretty sure you can halve the number of radiators needed to cool a system, which is ideal for matx. Some calculations suggest a 6 tec device that'll move 250W through 20 degrees will draw 192W, which at a mere 16A doesn't even break atx spec on a 12V rail.

Code seems to be behaving itself.

w1vk9w.jpg


That's the predicted curve for tecs running at 6V, through 0, 100, 300, 100, 0W load. 15 minutes on each, assuming 2l cold water, 0.5l hot water. No temperature control imagined yet, but given that's essentially a 240 radiator I think it's promising.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom