PA 120.3 and MCR320 in parallel

If I put a PA 120.3 and a MCR320 in parallel, would the water flow through each one about equally? I am thinking of adding another rad to my loop and I have a MCR320 I'm not using. I know I could do 2 loops, but then I'd have to have another pump and more holes in the case (externally mounted rear radiators). So I was thinking of just adding the MCR320 to my existing loop. But then I was thinking, should it go in in series or parallel. After some searching here, it seems like parallel would be better.

And yes I'm after that elusive .1C :)

Any advise is greatly appreciated.

]Monty[

Putting them in series is better, in almost all normal circumstances.

Putting them in parallel halves the flow rate to each (at best) and will result in a far greater performance loss than merely putting them in series and dealing with the extremely minor flow-rate loss of the extra pressure drop of doing so. In general, radiators are very low pressure drop items, so putting them in series has minimal overall impact on flow rates. What's of more importance is keeping the flow rates up through the radiators, or at least above 4LPM.

When dealing with multiple radiators, the rules are these:

*) Water-flow series
*) Air-flow parallel

OK, thanks for that Cathar.

OT - Nice bike.

]Monty[

Quote:

---

Originally Posted by montyshaw

OT - Nice bike.

---

Yours too!

Cheers. Although, it ain't pretty no more. It'll be fixed again.

Ouch Cathar.

Is that the source of your injuries right now? What happened?

Quote:

---

Originally Posted by Polizei

Is that the source of your injuries right now? What happened?

---

I was an idiot is what happened. I took a chance on a worn front tyre at one of the world's fastest motorcycle race-tracks, that being Phillip Island.

Rounded Turn 2 at around 85mph (one of the slower corners actually), and the front wheel started to slide. Did everything "right". I know this 'cos the slide lasted a good 20m (~60'). I did everything I could to give the front tyre a chance to re-grip, and it did re-grip for small fractions of a second, but let-go again, and in the end it wasn't to be.

Low-sided the bike, but must've dug some part of me into the track, and before I knew it I had somersaulted with my right-ankle as the pivot point (ie. an 80mph one-legged somersault) which micro-fractured the Talus, and 'cos it happened so fast, after the somersault I slammed the right wrist into the track for good measure, chipping the top off the ulna, and causing multiple cross-fractures to the radius where it joins the wrist. The damage to the radius was fairly serious, with risk of the bone collapsing, so they had to plate the radius to hold it together, and did a bone-graft to put the chip off the ulna back into place.

Am near totally house-bound. Won't be able to walk for 9 more weeks. The Talus is the main weight-bearing bone in the foot, so the usual 6-weeks healing time isn't the case here. 'Cos the bone is such a structurally important bone, the doctors recommend a full 13 weeks (3 months) before it can be considered to be healed completely enough to not risk expanding the fractures through walking/running/whatever.

Wrist, at the moment, is aching. Am on pain-killers. Only have around 50% full mobility in the wirst, but it's still very inflamed and swollen and it'll get better in time as the swelling settles. Strength in wrist is about 1/50th of normal. For example, picking up a cup of coffee is about the limit of what it can do, and even that hurts. Will probably be another 5 months before the wrist gets back to a semblance of normality.

Typing is not an issue though, and I telecommute for work, so my work has been unaffected.

Cathar why are you saying series is better? Parallel has much much less pressure drop because the flow is halved between the 2 rads. Also, those "tests" you mentioned are done with a constant 10C coolant/air temp difference, which means at more flow, MORE HEAT IS ADDED TO THE WATER TO ACHIEVE THE 10C air/coolant delta. There have been a few people who tried series vs. parallel, and parallel got better temps :P
Let me guess, now I'm going to be slaughtered for challenging the god? :rolleyes:

Well, I've tested it myself, but okay.

I'll be gentle. I don't attack people for asking questions, just those who attack me or others without good reason.

Quote:

---

Originally Posted by serialk11r

Also, those "tests" you mentioned are done with a constant 10C coolant/air temp difference, which means at more flow, MORE HEAT IS ADDED TO THE WATER TO ACHIEVE THE 10C air/coolant delta.

---

Talking about Bill's tests here eh? So if we had to add more heat to maintain a 10C coolant/air temp difference, what would have instead happened if we didn't add more heat (constant heat load)? Did the air-in temp change? If not, what changed?

Quote:

---

Originally Posted by serialk11r

There have been a few people who tried series vs. parallel, and parallel got better temps :P

---

So I'm guessing these people used a known, measured, and fixed constant heat-load across all tests, and monitored their air-in, water-in, and water-out temperatures, and measured the delta's?

Right now for me, I don't think those tests mean ANYTHING...I'm still thinking about it. I still think more flow only slightly increases the performance of the radiator...I don't want to type up the explanation, takes too long. I'll have my mom bring some Y fittings home...I don't have any and the hardware store doesn't sell them for some damn reason, then I'll test it for myself. But definitely, parallel should give the loop overall higher flow.

Quote:

---

Originally Posted by serialk11r

Right now for me, I don't think those tests mean ANYTHING...I'm still thinking about it. I still think more flow only slightly increases the performance of the radiator...I don't want to type up the explanation, takes too long. I'll have my mom bring some Y fittings home...I don't have any and the hardware store doesn't sell them for some damn reason, then I'll test it for myself. But definitely, parallel should give the loop overall higher flow.

---

So what you're basically saying is that you'd prefer to trust your assumptions, over those who have actually tested it.

Parallel will give the loop more flow, for sure. How much more? How much are you waving your hands and simply assuming here? Here's a calculated plot based upon established pressure drop figures:

http://www.employees.org/~slf/curves/pumps/serpar.png

In series, the radiators will have ~5.3LPM flowing through the system and each radiator.

In parallel, the radiators will have ~2.75LPM flowing through them, and ~5.5LPM flowing through the system.

There will be an extremely minor benefit at the waterblocks for moving from 5.3 to 5.5LPM, and a definite deficit at the radiators for dropping from 5.3LPM to 2.75LPM.

The two effects do counter-oppose each other to a certain degree. Which one "wins" depends on a number of factors, but for the general case with modern waterblocks, the performance loss at the radiators outweighs the minor waterblock benefits through the small flow-rate increase.

The differences are small though, I'll grant you that. They are small enough such that many people with 1C resolution on their CPU's won't notice it, but will typically range in the fractions of a degree, but in almost all circumstances should favor the in-series configuration. If people see differences larger than that, something else has changed.

The larger the radiators, the higher the initial flow-rates, and the more powerful the fans, then it is more likely that an in-liquid-parallel config will win, but once again, the differences are never large, and certainly NEVER larger than the waterblock performance differences between the disparate system flow-rates.

Thanks for the little simulation there. No I didn't say I trust my assumptions more than people who actually tested it, I think there isn't really that big of a difference in the first place, and I'm going to test it myself :P And yes, the difference isn't going to be that big but...ah well, I just think tweaking things is cool.

great explanation!! i am going to run some numbers to back that up, be right back

The water is still going through 2 radiators no matter how you do it. So how could running parallel provide less of a drop?

Quote:

---

Originally Posted by SDatl404

The water is still going through 2 radiators no matter how you do it. So how could running parallel provide less of a drop?

---

less flowrate through the rads because the flowrate is halfed, thus less cooling perforamnce.

Quote:

---

Originally Posted by sick_g4m3r

less flowrate through the rads because the flowrate is halfed, thus less cooling perforamnce.

---

I am talking about flow rate drop. Not overall cooling performance. Overall flow rate of the loop should be the same due to the fact you have the same level of restriction in the loop.

Running through the maths of it.

Given a 100W CPU heat load, and even when tripling the Swiftech supplied C/W's for the block, the difference between 5.3 and 5.5LPM is just 0.11C at the CPU die (in favor of 5.5LPM).

The difference in water temps, for 5.3LPM through the rads, and 2.75LPM through the rads, for 108.5W of heat load (CPU + pump heat), is actually 0.11C in favor of the series config. (Use Yate-Loons at 12v).

LOL. 0.11-0.11 = 0.00C.

There'd actually be no difference between series/parallel for the proposed config.

Running the fans faster would favor parallel by 0.01C or so, and vice-verse for slower fans.

Okay - I'm going to change my stance now and say that there'd be no discernible difference between series/parallel. Go with whatever is easiest.

Okay then. Running some more scenarios.

2 x single fan radiators => In liquid-flow-series is very marginally better (~0.1C)
2 x dual fan radiators => No preference for either
2 x triple fan radiators => In liquid-flow-parallel is very marginally better (<0.1C)

In the past I had only tested 2 x single-fan rads, which explains what I saw in my testing, which indeed was <0.2C differences in favor of flow-series.

Quote:

---

Originally Posted by Cathar

[injuries]

---

Ouch...
Get well soon :)
I see you found some use for the extra time, making all these tests - well done :up:

Quote:

---

Originally Posted by Cathar

Okay then. Running some more scenarios.

2 x single fan radiators => In liquid-flow-series is very marginally better (~0.1C)
2 x dual fan radiators => No preference for either
2 x triple fan radiators => In liquid-flow-parallel is very marginally better (<0.1C)

In the past I had only tested 2 x single-fan rads, which explains what I saw in my testing, which indeed was <0.2C differences in favor of flow-series.

---

even with the flow halfed to the rads its only that little difference? i was expecting 2-3C

Quote:

---

Originally Posted by sick_g4m3r

even with the flow halfed to the rads its only that little difference? i was expecting 2-3C

---

It's because there's just so much cooling power on offer with two x dual-rads. Yeah, the radiators perform worse with less coolant flow, but because the overall C/W is so low, a 5% performance difference (which is what the difference between 2.75LPM and 5.3LPM is), doesn't equate to much of an actual temperature change because the radiators are holding the water temps to around 2C above ambient anyway. 5% of 2C is 0.1C.

At half the flow, the pressure drop is much less than 1/2 but then the rads aren't exactly the biggest restriction in a loop so...
In series the second rad gets cooler water (but only cooler by a tiny bit) which means that it will cool worse (but by a tiny bit). :P doesn't really matter.

wait, i have a question.

say i have a D5 and 2 MCR320s, a fuzion, and a mcw60.

how would one calculate the expected flow of such a system?

the pressure drops of the parts can only be found at certain flows, and when you cross them with the D5 pq curve, the flow is not the same.

so i am very confused. how do you find the flowrate given a pump and the parts?

Quote:

---

Originally Posted by sick_g4m3r

wait, i have a question.

say i have a D5 and 2 MCR320s, a fuzion, and a mcw60.

how would one calculate the expected flow of such a system?

the pressure drops of the parts can only be found at certain flows, and when you cross them with the D5 pq curve, the flow is not the same.

so i am very confused. how do you find the flowrate given a pump and the parts?

---

Add up the pressure drop vs. flow curves...sorta hard to do but you could probably get an approximate equation for each and then add. Then as for tubing...there's theoretical pressure drops for tubes, so I'd take however many feet you're using, and add a little more on top of that because of bends. Hope that makes sense...

Quote:

---

Originally Posted by serialk11r

At half the flow, the pressure drop is much less than 1/2 but then the rads aren't exactly the biggest restriction in a loop so...
In series the second rad gets cooler water (but only cooler by a tiny bit) which means that it will cool worse (but by a tiny bit). :P doesn't really matter.

---

Are you saying that the greater the flow the higher the resistance gets? I have a hard time choking down that water cooling pumps could push enough water for you to even reach noticable loss in flow due to radiator configuration.

Saying the other radiator gets cooler water is like saying half of a larger radiator gets cooler water. If you didn't need the water colder after it left the first one why have a second? :shrug:

Quote:

---

Originally Posted by serialk11r

Add up the pressure drop vs. flow curves

---

how exactly does one add these up? And then after they are added up for all the parts, do you find the intersection with the PQ of the pump?

Quote:

---

Originally Posted by SDatl404

Are you saying that the greater the flow the higher the resistance gets? I have a hard time choking down that water cooling pumps could push enough water for you to even reach noticable loss in flow due to radiator configuration.

Saying the other radiator gets cooler water is like saying half of a larger radiator gets cooler water. If you didn't need the water colder after it left the first one why have a second? :shrug:

---

Yes. I'm sorry, I'm afraid I can't help you if you did not notice that from the various charts, etc. that are available to the public. I can't help you either if you don't see how 2 radiators would be useful.

sick g4m3r, like I said you must approximate. Plug in different equations y=ax^2, and check the values to see if they approximately line up...in the end its still an approximation.

Quote:

---

Originally Posted by serialk11r

Yes. I'm sorry, I'm afraid I can't help you if you did not notice that from the various charts, etc. that are available to the public. I can't help you either if you don't see how 2 radiators would be useful.

sick g4m3r, like I said you must approximate. Plug in different equations y=ax^2, and check the values to see if they approximately line up...in the end its still an approximation.

---

lol... You claimed that the water coming out of the first radiator was colder, like it's a bad thing. This makes no differnace on the total cooling of the system. It's still passing through 2 radiators. Wether at the same time or one after the other it's all the same. Unless so one can logicly prove other wise. All I said was that if the water being cold enough after leaving the first why do you need a second?

The way to "add them up" is easy enough.

For two radiators in series, you just add together the two PQ curves for a single radiator. Add all the PQ curves together (rads plus block + tubing) and see where that intersects the pump PQ curve.

For two radiators in parallel, it's a little trickier. Resistance is proportional to the flow rate squared. At half the flow through each radiator, there's a quarter of the pressure drop, so what we do is take the curve for a single radiator, and divide each point by 4. We then add that curve to the curves for the block + tubing, and see where that intersects with the pump. There is the added restriction of the 2 x Y's though (the Y-split/join itself, and the tubing-Y-barb interface's), and that all needs to be factored in.

I reflected all that in the curves that I presented above.

Quote:

---

Originally Posted by Cathar

So what you're basically saying is that you'd prefer to trust your assumptions, over those who have actually tested it.

Parallel will give the loop more flow, for sure. How much more? How much are you waving your hands and simply assuming here? Here's a calculated plot based upon established pressure drop figures:

http://www.employees.org/~slf/curves/pumps/serpar.png

In series, the radiators will have ~5.3LPM flowing through the system and each radiator.

In parallel, the radiators will have ~2.75LPM flowing through them, and ~5.5LPM flowing through the system.

There will be an extremely minor benefit at the waterblocks for moving from 5.3 to 5.5LPM, and a definite deficit at the radiators for dropping from 5.3LPM to 2.75LPM.

The two effects do counter-oppose each other to a certain degree. Which one "wins" depends on a number of factors, but for the general case with modern waterblocks, the performance loss at the radiators outweighs the minor waterblock benefits through the small flow-rate increase.

The differences are small though, I'll grant you that. They are small enough such that many people with 1C resolution on their CPU's won't notice it, but will typically range in the fractions of a degree, but in almost all circumstances should favor the in-series configuration. If people see differences larger than that, something else has changed.

The larger the radiators, the higher the initial flow-rates, and the more powerful the fans, then it is more likely that an in-liquid-parallel config will win, but once again, the differences are never large, and certainly NEVER larger than the waterblock performance differences between the disparate system flow-rates.

---

? Would having two PA120.2 in series double the heat dissapation?

Quote:

---

Originally Posted by lilflippy

? Would having two PA120.2 in series double the heat dissapation?

---

Good point. Not quite. The water discharge from the first radiator is cooler, thereby decreasing the performance of the second radiator in the series. The significance of this depends on the flow-rate. At 5.3 LPM in the scenario, we're talking about a ~8% reduction in the performance of the 2nd radiator. That would translate into roughly a ~0.04C increase in the system water-temperature for the scenario, or not really enough to sway which was preferable in any significant fashion.

OT - Cathar is this another bike accident?!?! :D

Quote:

---

Originally Posted by Cathar

I took a chance on a worn front tyre at one of the world's fastest motorcycle race-tracks, that being Phillip Island.

---

Quote:

---

Originally Posted by Borgod

OT - Cathar is this another bike accident?!?! :D

---

Wow! Mind like a steel trap eh? Can't hide anything from you, can I?

What's the grinning smiley for? Schadenfreude?

So basically the differences are small enough, that it probably depends more on how the tubing is run and the restrictions of the Y's.

One thing I don't understand is if the flow rate through a radiator is lower, why does that mean less performance? It seems to me that the longer the water is in the radiator, the more opportunity there is to dump heat into the air. Won't slower flow make the temp delta (between incoming and outgoing water) greater ?

]Monty[

Quote:

---

Originally Posted by montyshaw

One thing I don't understand is if the flow rate through a radiator is lower, why does that mean less performance? It seems to me that the longer the water is in the radiator, the more opportunity there is to dump heat into the air. Won't slower flow make the temp delta (between incoming and outgoing water) greater ?

]Monty[

---

I would lean logically towards that conclusion too, in our application i suspect that their is not enough heat in order to dissipate through both radiators as the amount of heat that is transfered out of the radiator increases with the difference between it and surrounding air. If we had a higher temperature then we would see a bigger cooling difference if the second radiator was in series or parrallel because it would compensate for the heat that could not be removed in the first pass. This is the same principal as why there is not just one tube but runs of tubes in the radiator, i.e. multiple passes, and since the radiators are geared towards our temperature and efficiencies we would have to have extreme conditions before we saw a benefit.

Quote:

---

Originally Posted by Cathar

Wow! Mind like a steel trap eh? Can't hide anything from you, can I?

What's the grinning smiley for? Schadenfreude?

---

I'm morbidly amused that you had another fall after the seriousness of your last one (if it indeed is a new fall).

Quote:

---

Originally Posted by montyshaw

Won't slower flow make the temp delta (between incoming and outgoing water) greater ?

---

(sigh)

I'll pull out ye olde racetrack analogy.

If I'm a molecule of water, circulating around a water-cooling loop, I'm just going around and around and around. It's just like being a race-car on a race-track at a fixed speed. The time I take to travel through the radiator is like the time spent travelling down the main straight of the race-track.

If the race-track is 5 miles long, of which the main-straight (radiator) is 1 mile long, in the space of 1 hour, how much time per hour do I spend on the main-straight if I'm travelling at a fixed speed of 60mph, 120mph, or 180mph?

The answer is that the speed doesn't matter of course. I spent 1/5th of my time on the main straight. 'cos the straight is 1 mile long out of 5. Even if I travel at twice the speed around the circuit, it'll take me half the time to get from the start to the end of the main straight, but because I'm going twice as fast everywhere else, I'll be on the main-straight twice as often. i.e. No matter how fast I travel, I'll be on the main-straight for 12 minutes out of every hour.

Now common physics tells us that the rate of heat-exchange is proportional to the temperature difference between an object, and something else that's cooler than the object. If I'm a molecule of water, and air is cooling me (by way of the air cooling the metal tubes inside the radiator that I'm flowing through), then the longer I spend in the radiator, the cooler I will get (that's good), BUT, the cooler and closer I get to the air-temperature, the less quickly I'll lose heat (that's bad). The two cancel each other out. So why does higher flow results in better performance?

The more quickly I rush around, the more likely I'm going to be tumbled about (think white-water rapids as opposed to a smooth slowly flowing river). This means that I'm going to get tossed against the cool metal walls of the radiator more often, rather than just cruising along down the middle of the tube, only passing heat slowly to water molecules beside me that are only a little cooler than I, because they are beside another molecule, and then beside another molecule, before we get to the cold metal wall. i.e. water sucks for transferring heat if it's not getting mixed about and thrown against the cold walls.

Hope that explains it in a "simple" manner that is intuitive, obvious, and directly counters the "slower flow is better" argument.

Cathar....are you assuming a perfect splitter? i.e., a Y-splitter that has zero restriction

Since I don't think one exists....granted they're not very restrictive, but considering there is restriction, that would lower the flow of the parallel setup slightly, giving a super-slight advantage to series (and getting closer to your <.2C difference, rather than the calculated ~.1C).

Maybe it comes down to whatever is easiest to tube?

BTW, I don't think I've said this yet, but I REALLLLLY like your wholistic system simulator program you have there....:)

Quote:

---

Originally Posted by Cathar

Hope that explains it in a "simple" manner that is intuitive, obvious, and directly counters the "slower flow is better" argument.

---

It did for me :)

I'm learning some thing new here everyday.:D

^ wow excellent explanation!


so if i wanted to add the pressure drops of a fuzion, 2 MCR 320s in series, 1/2" tubing, and a MCW60:

.5GPM~.4PSI drop
1GPM~ 1.8PSI drop
1.5GPM~ 3.45PSI drop
2GPM~ 5.5PSI drop


so with these numbers of flow vs pressure drop (PSI), how would I intersect this with the D5 PQ curve, of flow vs pressure (not drop)??

Quote:

---

Originally Posted by Cathar

So why does higher flow results in better performance?

The more quickly I rush around, the more likely I'm going to be tumbled about (think white-water rapids as opposed to a smooth slowly flowing river). This means that I'm going to get tossed against the cool metal walls of the radiator more often, rather than just cruising along down the middle of the tube, only passing heat slowly to water molecules beside me that are only a little cooler than I, because they are beside another molecule, and then beside another molecule, before we get to the cold metal wall. i.e. water sucks for transferring heat if it's not getting mixed about and thrown against the cold walls.

Hope that explains it in a "simple" manner that is intuitive, obvious, and directly counters the "slower flow is better" argument.

---

I think this is a key point that people fail to understand. Your great analogy makes it a lot easier to understand.

cathar, nice explanation, thankyou :) i knew all about the time spent in the rad being the same, but never thought of the water rapids :)

Quote:

---

Originally Posted by sick_g4m3r

so if i wanted to add the pressure drops of a fuzion, 2 MCR 320s in series, 1/2" tubing, and a MCW60:

.5GPM~.4PSI drop
1GPM~ 1.8PSI drop
1.5GPM~ 3.45PSI drop
2GPM~ 5.5PSI drop


so with these numbers of flow vs pressure drop (PSI), how would I intersect this with the D5 PQ curve, of flow vs pressure (not drop)??

---

It's somewhat involved.

Download this tubing/fitting pressure drop calculator. It's what I use.

Using established measured flow/p.d. results for your various items, manually calculate the total system pressure drop for at least 5 points in the range from 0.5gpm (2lpm), to 2gpm (or 8lpm).

We then model a curve against our plot points (via any decent least-differences curve modelling software) using the formula "a*flow^b + c*flow". Model for a, b, & c. This formula is accurate enough for modelling flow-resistance, and is typically what is used when you see the smooth line PQ graphs for various things (but NOT for use with pump PQ curves).

Plot our curve onto a graph, and then verify that the curve matches our calculated data points to a fair degree of accuracy. There are always going to be some inherent small amounts of error, primarily derived from the margins of error in the measurements of the pressure-drops of the blocks/radiators.

What this does is gives us a smooth curve that fills in the gaps between the data points.

We then plot that curve over the PQ curve for the pump, and determine the point of intersection.

^ thanks a ton cathar! Thats the direction I was heading, but what I dont understand is how can you plot this when one curve is pressure drop, and the pump curve is attainable pressure

Thanks Cathar. Great explaination!:up: Now I just feel dumb ;)

]Monty[

I did it!!

It wasnt pretty but i did it. I used a QuadReg to get the components' curve, and exponential reg to get the D5's curve.


And in the end I ended up with (VERY rough):

1.55GPM @ 3.63PSI of pressure.

1.55 GPM for a D5 on two MCR320s, a MCW60, and a Fuzion is not bad at all!!!

That seems too good, but I think I did it right.

BUMP

sorry, i just want to know if these calculations seem right. Seems a little too good.