Radiators in Parallel

[grudgelord]

While we've discussed various combinations of pumps and even various blocks arranged in parallel I haven't noticed any discussion of serial versus parallel radiators. This isn't to say it hasn't been addressed but I haven't seen it and a cursory search didn't yield any information (I'd actually be surprised if it isn't buried around here somewhere). If this is the case then assume I am lazy and kindly call me a dancing banana and direct me to the discussion or, as it never hurts to revisit our alleged science, discuss it.

The simple question is: Where would be be likely to see better thermal performance, two (identical) radiators in series or in (equal flow) parallel? Unless I miss my mark we'd see a significant falloff in performance of any rad(s) after the first (and I believe as much has been stated by one of our resident rocket scientists) in the series.

For parallel radiators it seems we'd see a more dramatic ΔT at each unit but I'm not sure how you'd handle the math for splitting flow prior to the radiators, applying ΔT, and then re-summing the flow, yada yada yada. Nor am I sure of the real world ramifications of such a configuration. Sometimes theory doesn't translate to practice, so I'm interested in praxis as well as theory.

For the sake of argument let's assume that all components are capable performers at their tasks, no wimpy 80mm rads (assume a pair of decent 120.2 or something) or excessive flow restriction or other such silliness. In fact, lets assume a simple CPU and GPU loop.

Intelligent thoughts?

[Chas_The_Man]

I have 2 Swiftie 120.3's in paralell. The flow through each rad is in theory, equal, slower, and with less backpressure.

I dont know if it makes a difference. I would say it is probably slight. It makes for less tubing in my case.

[serialk11r]

Science/math/wtv tells us parallel radiators will have less restriction...I am working on a erm guide I guess you could say about it.

[Marci]

This isn't to say it hasn't been addressed but I haven't seen it and a cursory search didn't yield any information (I'd actually be surprised if it isn't buried around here somewhere). If this is the case then assume I am lazy and kindly call me a dancing banana and direct me to the discussion or, as it never hurts to revisit our alleged science, discuss it.

Take a dive in the former fount of all watercooling knowledge (relatively).

Go search at www.procooling.com's forums. It's also been brought up in some detail at overclockers.com. A few selective examples...

Not to put too fine a point on it but. . .

There's a lot more going on here than simply looking at the water-air delta-T. If you wish to make things simple, all you need to do is consider the equation for convection. It goes something like this:

q = h * A * delta-T

People are quick to point out that delta-T is "high" entering the first radiator and "low" entering the second radiator. This is overly simplistic. "High" and "low" have no meaning outside of one having a greater value than the other.

Perhaps is would be better to ask a different question.

Say you have one radiator (call is "A") that has twice the internal surface area of another (call is "B"). Now suppose that you have two of radiator "B". Is this enough information to predict the relative performance of "A" against 2 of "B" (in either series or parallel)? Heck no.

Even conducting a test whereby you measure the outlet water temperature versus the inlet water temperature is incomplete.

You could have water 1°C over ambient, but if it's only flowing at 10 gph it will not yield an acceptable chip temperature. You could have water 10°C over ambient and your chip would do fine if the flow rate was 100 gph.

Man I'm getting off track here. Anyway, if you look at the convection equation you can state that the area is the same whether radiators are in series or parallel. So it comes down to a question of "h" and "delta-T". Each of these are variables, ie they change as you travel through the radiator.

"h" is largely a function of velocity. Velocity will always be higher when you plumb in series. Delta-T is simply the water temperature minus the air temperature. Such a statement belies the complexity of calculating what delta-T will be versus total flow rate, total heat load, total air flow, etc., etc.

The gist of what I was getting at before is that you need to consider the effect of higher surface velocity in the radiator (kudos to series plumbing) vs the higher overall system flow rate (kudos to parallel plumbing). With low restriction radiators, overall flow rate won't take much of a hit from plumbing them in series. In this instance (and this is the minority of cases) plumbing in series will actually result in improved chips temperatures. With relatively high restriction in the radiator (most fall into this category) plumbing in parallel will yield enough of an overall system flow rate boost that chip temperatures will be better this way.

The conventional wisdom that parallel radiators is preferred will usually be correct, but not always. Blindly stating parallel to be better is not right. Honestly a little background testing on head loss would let you make the right call probably 95+% of the time.

Really this is no different than the argument about series versus parallel for dual systems. The "best" option depends largely on the restrictiveness of the blocks used. Highly restrictive blocks will do better in parallel while low restriction blocks may do better in series.

I suppose what boggles my mind is that people so readily acknowledge that flow velocity in blocks is paramount yet fail to recognize that the same rules apply to radiators. I guess a partial explanation is that radiator area is practically unrestricted in comparison to block area, meaning that you can compensate for lousy water or air flow with a really big radiator. This isn't the case with blocks where all power will come from a tiny area whether we like it or not.

I've said it before and I'll say it again. They only way to be 100% certain is to test the options.

Source: http://forums.procooling.com/vbb/sho...9&postcount=22

Mhh for one i dont think he bases his assumptions on "a semester of mechanical under [his] belts", like i said he's a pro. Do not underestimate him.
For two he maybe has badly formulated his thought, please bear with him he's a technical guy. Ya know, all technical guys have difficulties being clear to other people. (and to suits, but thats another problem)

In this very case he tried to be short and to-the-point, since this very topic has been beaten dead and buried 50ft under a long time ago.
What everyone agreed on, was the basic thermal equation:
q=UAdT
where A is the surface area of transfer. The higher the flow, the higher A is (and the lower U is due to laminar vs turbulent etc..)
That's true for WBs, thats true for rads, thats true for air on a heatsink, thats true for air on a rad, thats true for your car radiator.

Putting rads in parallel will increase the dT, while not killing A too much since the mean water path will be shorter (ie backpressure will be lower) - so resultant flow wont be divided by two (compared to flow in series setup).
The result, a parallel is winning.

Source: http://forums.procooling.com/vbb/sho...8&postcount=39 (later in the same thread)

The last time it was discussed in depth here was a LONG time ago, and the forum database appears to have been pruned of all content beyond a certain threshold, thus a lot of older useful info has been lost.

[grudgelord]

Originally Posted by Marci

Take a dive in the former fount of all watercooling knowledge (relatively).

Go search at www.procooling.com's forums. It's also been brought up in some detail at overclockers.com. A few selective examples...


Source: http://forums.procooling.com/vbb/sho...9&postcount=22


Source: http://forums.procooling.com/vbb/sho...8&postcount=39 (later in the same thread)

Exactly what I was looking for. Many thanks for the links!

EDIT: I hadn't realized just what a Pandora's Box I was (re) opening until I checked these links. The debate over in procooling got rather heated (pun intended) didn't it? I'm surprised these guys weren't duking it out with broken beer bottles in the streets.

[grudgelord]

Originally Posted by serialk11r

Science/math/wtv tells us parallel radiators will have less restriction...I am working on a erm guide I guess you could say about it.

I'd be interested in your findings when you are finished.

[ladderman]

Take a quick look at this.

[eXceededgoku]

^^ wtf.... woah..........

[septim]

looks cool ladderman.
i would maybe add another d5 just for redundancy and more flow...

[nikhsub1]

Originally Posted by ladderman

Take a quick look at this.

Looks like a bloody mess Anyway, if you know the basic principals of what is happening with rads (or anything really) in series vs parallel you should be on the way. Ladderman has his rads in series, as do I. The only way to know for sure what will perform better is to test it both ways.

[ladderman]

Woah Boys.

Thats not my rig, it's done by a company in the UK called Vadim. They make them to order, it costs 5 grand.

I just put it on to show what I thought was parallel rads.

I've just started building my TJ07

[Marci]

Yes, that's one of Vadim's PCs... the rads in it are in series.

I'm surprised these guys weren't duking it out with broken beer bottles in the streets.

Heh - that was saved for the diesim vs CPU testing arguments...

[Cronos]

It can be easily shown mathematically that two rads in parallel should be better
than in series -but the margin is rather small, so that final decision should be based on convenience rather than performance.

[Robump]

I am thinking about doing this, then i may split them into 2 different loops.

Rob

[NaeKuh]

Originally Posted by ladderman

Take a quick look at this.

isnt that a series??? i dont see a splitter.

you connected the output of rad1 to the input of rad2.

Unless im missing the Y somewhere??



The OP is wanting to use a Y, with = length tubing. Its much more difficult to pull off because of space limitations. However in your case it could be done.


BTW: very nice setup tho. But not what the op is looking for.


EDIT: ACK wrote this b4 i saw marci's post.

Originally Posted by Cronos

It can be easily shown mathematically that two rads in parallel should be better
than in series -but the margin is rather small, so that final decision should be based on convenience rather than performance.

heh... if you want to say that... then watercooling for that bit of extra OC, really isnt worth it in a financial standpoint. Not much different 200-300mhz+ then what air can do. But why do we watercool to get it?

I think it would be a very fun project. But the only case i could think it would be possible to pull it off would be a dualy. This way you have absolute precise control of the Y split, and you can assure each tubing will have the same up's and downs, and = pressure will be taken to both rads.

Well, you could always fugly it and do it external too. Maybe make wings for your computer.

But in all, its very hard to pull off precisely.

[jinsean]

So, I sucked at physics and I think I finally understand the difference b/t rads in series Vs. parallel. Please correct me if I'm wrong.

Rad in parallel:
Pump -> splitter -> Rad 1 + Rad 2 -> joiner to rest of loop

Rad in series:
Pump -> Rad 1 -> Rad 2 -> rest of loop

[Pete]

Vadum charge you for a load of craply built stuff.

[teyber]

http://forums.extremeoverclocking.co...+radiator+idea
good one there, check that out

[serialk11r]

Originally Posted by Marci

Take a dive in the former fount of all watercooling knowledge (relatively).

Go search at www.procooling.com's forums. It's also been brought up in some detail at overclockers.com. A few selective examples...


Source: http://forums.procooling.com/vbb/sho...9&postcount=22


Source: http://forums.procooling.com/vbb/sho...8&postcount=39 (later in the same thread)

The last time it was discussed in depth here was a LONG time ago, and the forum database appears to have been pruned of all content beyond a certain threshold, thus a lot of older useful info has been lost.

Wow that says it all...

[sanhacker]

Just a quick question.
Wouldn't I accomplish the same thing as radiators in parallel if I use a reservoir which accumulates and distributes coolant to my system in a combined reservoir?

I mean dual radiators with a common reservoir for coolant and an "average" temperature for both. Thoughts?

[sick_g4m3r]

I'm going to jump off a cliff to see if gravity affects my current bodily setup (I did gain a few pounds recently).

Radiators are more efficient with a larger dT. If you put hot water through one radiator in a 2way series, it'll cool it decently through the first and the second won't do crap because of the low dT. In parallel, you'll have more efficient use of both because the dT is at the maximum for both, and the flow rate is lower in both (more time for thermal transfer in the rad). It doesn't take a rocket scientist to see that if the theory overwhelmingly supports one choice over another that odds are the one supported by theory is most likely going to do better in practice.

Oh, and if you get them working even EQUALLY well in series, you have a seriously screwed up system, i.e. you're not blowing air over your rads when they're in parallel or something. There shouldn't be any series configuration that does even equally well if all things such as air flow across the rads and mean system water flow/pressure are the same.

Personally, I like pushing the bounds of logic and trying things that are counterintuitive, but only to a point. It doesn't take jumping off a cliff to prove to yourself that gravity works. Neither does it take lighting your car on fire to prove that it is generally combustible or eating nightshade to prove that it is poisonous. If you want to waste the time, that is your perogative and I'll look forward to seeing how your results back up real science and putting this debate to rest, but there is no sense further beating on a dead issue. The only reason someone should even consider doing it this way is for aesthetic reasons, and even that is normally a stretch since, as Ben pointed out, even the oddest configurations can be set up either way.

My $0.02.

this is so true. that parallel idea sounds very attractive

[serialk11r]

Originally Posted by sanhacker

Just a quick question.
Wouldn't I accomplish the same thing as radiators in parallel if I use a reservoir which accumulates and distributes coolant to my system in a combined reservoir?

I mean dual radiators with a common reservoir for coolant and an "average" temperature for both. Thoughts?

I don't see how you would do that. Putting the radiators IN FRONT of the pump? That wouldn't be very friendly to flow nor bleeding.

[sanhacker]

Originally Posted by serialk11r

I don't see how you would do that. Putting the radiators IN FRONT of the pump? That wouldn't be very friendly to flow nor bleeding.

Actually, I'll be using two pumps. One for each loop. Having the coolant in a common reservoir would assure me of an average temp for both loops. NO?

[serialk11r]

Originally Posted by sanhacker

Actually, I'll be using two pumps. One for each loop. Having the coolant in a common reservoir would assure me of an average temp for both loops. NO?

Oh IC what you're getting at, dual loops with 1 res? No that's different. Parallel rads is maximizing flow and cooling power inside a single loop. By splitting flow to 2 rads, you get less restriction, and I'm thinking probably better performance. Its useful for those without the space for huge rads yet want to throw in more cooling power, or those who want sheer overkill

[sanhacker]

Originally Posted by serialk11r

Oh IC what you're getting at, dual loops with 1 res? No that's different. Parallel rads is maximizing flow and cooling power inside a single loop. By splitting flow to 2 rads, you get less restriction, and I'm thinking probably better performance. Its useful for those without the space for huge rads yet want to throw in more cooling power, or those who want sheer overkill

Okay. Overkill it is.
However, I'm thinking if a loop goes down, I have a fail safe system. The likelihood of two pumps going down is well, just say very unlikely.

As long as water is moving in either loop, I think I'm okay.
Obviously there is likely something I'm missing.

I've just decided dual pumps, dual rads and a common res might be the ticket. Wouldn't this give me a common temperature between loops?