Dual Pump Idea
 

http://www.3extremedesigns.com/watercoolidea.gif
Just doing some thinking, and thought that maybe this route might have some benefits? I guess my thinking is that the radiator would keep the reserviour at a certain temp, and the incomming liquid from the cpu block would be cooled by both the radiator, and the already cool liquid inside the reserviour....any thoughts?

It really depends on how you'd like to do it. I personally go pump, CPU block, CPU block, split and reduction to 3/8"ID, GPU and Northbridge, Dual heatercores, re-store to 1/2"ID and recombine lines, pump, reservoir, then around to the other pump. I have two Xeons, and they really don't need what they're getting, but I could afford the extravagance at the time, so I did. Again, it's basically up to you. If you've got dual pumps, put them in a push-pull loop like I did, or create two separate loops, one for the CPU and one for the other stuff. Be warned though. You'll need some room in the case to do this. If you go with a push-pull loop, you've got redundancy, so if one pump dies, you've still got flow. Parallel loops don't give you that nice redundancy. Happy watercooling...

I forgot to mention, If you do the push-pull idea, you get extra head(get it?) All it really means is your loop will be under more pressure, and you can make it somewhat taller than if you only had one pump. I'd put the reservoir at the top, so it can be filled easily, and function as an airtrap.

actually, the route that you have drawn up there looks like a very bad idea, very inefficient. mainly because you sort of have two loops there. there's no guarantee that the water the comes out of the loop with the blocks in it is going to get drawn into the radiator, and likewise, the cooler water out of the rad could be sucked directly back in instead of going to the blocks like it should. and like was said previously, if you're gonna go to the trouble of having two pumps, why not make it redundant? you might as well get something out of it. also, you're cutting way back on the effeciency of the rad (you get the most heat transfer out of the coolant when you have the highest temp difference between the coolant and in this case, the surrounding air). you're gonna be sending a lot of heat back through the cooling loop this way, and your temps are gonna suffer for it.
check THIS thread out, very similar topic.

Yeah, unfortunately your idea is no good at all. It's the worst of all worlds infact.

First off, your cooling will be much worse. That's because cooling is driven by temperature difference. You are feeding your CPU with water that has only been partialy cooled by the radiator. You are feeding the CPU with water that has a temperature somewhere in between what's coming out of the radiator and out of the CPU. The water coming out of the radiator is naturaly much colder than the reservoir water so it makes much more sense to simply feed that into the CPU. The colder the water going into the CPU, the colder the CPU.
The same effect happens to the radiator. Instead of having warm water from the CPU going in which it then cools, it has your sorta warm water from the reservoir going in, which reduces the temperature difference and makes the radiator less effective.
So that's BAD no 1.

BAD no 2 is that you have two pumps and you should put them in series. If you have them in series then you get more pressure and flow which will improve your cooling and give you redundancy if one breaks down.

Edit. Already covered.

Try it. It might work...
 

just thought id add my experience with this idea...http://www.ocforums.com/showthread.p...0&pagenumber=1

this thread might like to read this too: http://forums.procooling.com/vbb/sho...ight=dual+pump

Hmm. I have some problems with agreeing on the points made here.
I have 2x MCP600, and made several test setups. One simular like yours.

You idea is good.A radiator is restrictive. If you have 2 pumps.. why not use it?
Well. I made several setups, even with only the WW and 2x MCP and a load of water.

Extra speed in the water will give you only 0.5 degree. Sorry but true.

In my current setup (best balance between temp and noise) I only use one pump, the WW, the D-teck Z Chip, Double heatercore, single heatercore and a BIX.
The temps with one extra pump are 0.3 degree lower.

I placed my GPU in a seperate loop, and that did help :)

Pumps in series will give better head, but, depending on the type of impellor, als more noise.
In my case, with closed impellor, it makes no sence. Yes my speed goes up, but my temp hardly goes down.

Best way to lower your temps is by optimizing your airflow in the radiator(s)
I have my water temp now 1.5 degree above ambient, with 3 fans at 7v :)
With the fans at 12V it is less then one degree.

"Best way to lower your temps is by optimizing your airflow in the radiator(s)"

jeez, finally
peeps, WCing is ALL about air

What they said. The more surface area of the rad that air is moving through and the more CFM through that surface area the cooler your temps will be. Dropping your water temps will in effect drop your CPU temps nearly linearly. I say "nearly" because I don't have good enough temp probes to be able to tell if it is perfectly linear. I would think it would be but I am sure there's plenty I don't add into the equation.

I setup two rads in a loop this last weekend and have been adjusting the CFM's through the rads and the CPU temps seem to rise and fall pretty much the same as the water temp does. Both temps were taken from the same DMM but different TC's. I did half ass calibrate them so I am not to far off.

so the only way a second pump is useful is to provide more inline pressure or a second circuit of pressure to run more radiators.

i agree, this setup is not going to "re-invent" the wheel but saying its the "worst of all worlds" and a "very bad idea, very inefficient" isnt quite true...

tho i must say, reading those words of encouragement "inspired" me to built it...thanks.

ok, i'm confused, do we have two completely seperate topics going here or what? yes, more air through a radiator is definately a good thing, i think we've realized that from using heatsinks and traditional aircooling, but I still fail to see how the heck having essentially two independant loops (radiator loop and blocks loop) connected by the reservoir, will do anything but hurt temps. please correct me if i'm wrong here.

There is only one topic here. The point of more airflow as I took it was to not bother with a second loop, put the rad and possible the second pump if desired inline with everything else, and make sure your airflow is optimal and your results will be the best it can be as opposed to what is shown above. In other words whats is shown above is unnecessary to achieve the same if not better results.

isnt the whole point of water cooling to get the heat from the restricted space of the CPU out of the case where you can fit lots more surface area to still ultimately radiate heat to the air.

but no matter what your 100% efficiency with a radiater is to hit ambient temp. without something more than air on metal, its impossible to ever ever ever get under ambient. right?

Not necassarily out of the case but that is good to. The goal is to be able to have a device that is as optimal as poosible to trasnfer the heat from the CPU (Usually the water block) to the coolant then to transfer the heat from the coolant to somewhere else (Usually a radiator and into the surounding air). The better the water block the more efficent the heat can be transfered in to the water and the more efficent the rad setup the more heat from the water can be transfered into the air. And then that air is hopefully cooled by convection of the surrounding enviroment.

There is no way to even get ambient temps from anything other than active cooling of the coolant wether that be air on the HS or water in the water block. Something unatural has to be cooling the coolant being applied to the heat exchanger. That could be a ducted A/C cooled air blowing directly on the HS or it could be chilled water running through the water block.

I'm thinking this isn't such a good idea since your mixing the hot and the cold water of the system.

ok, here is the theory in logic steps, tell me where it goes wrong...

after trying to explain this theory of mine to a few people at work, ive come up with this:


lets make the 4 following assumpions:

1) my cold loop will have 2x the flow of my hot loop.

2) a (big) pump adds +1C (50W) to the water.

3) a WB adds +2C (100W) to the water.

4) when 2 equal amounts of water @ different temps are mixed together, the resulting temps is the average between the two.
ie 1 cup 20C + 1 cup 40C = 2 cups @ 30C.

--------------

lets follow 10 (cm3) of water thru my setup:

--start at the water coming out of the rads, lets say its [30C]...5(cm3) goes to a pump [31C] then to WB [33C] then back to res where it mixes with the other 5 (cm3) which is still 30C [30+33=63 /2= 31.5C] then it goes to the second pump [32.5C] then into the rads.
so the rads have 32.5C water going into them.


now lets follow 10 (cm3) thru a normal loop:

--same starting point, out of rad [30C] pump [31C], WB [33C] ....rads.
so the rads have 33C water going into them....

if the 4 assumptions are true, how is the theory false? ( i know the +1C / 50W is NOT true, but it makes the math a lot easier)

-------------

there should be lots of holes in this if the theory is wrong...right? :shrug:

Heres my 2 cents...

I don't think the idea is so bad actually. I don't see why you could not make the water circulate so that the temp in the tank will be on average the same as the water going to the CPU. You just have to figure out the best places to put inlets and outlets.

IMO Pros:
+ You remove the restricting radiator from the CPU-loop giving you better flow in that loop.
+ Less noise than if you would put the the two pumps in the same loop as everything else.
+ You could easily double or tripple the rads without bothering about the restriction to much because the CPU flow will not be affected.

IMO BIG con:
The bigger the temeperature difference between the watertemp in the radiator and the ambient temperature, the better cooling effect. By putting the rad in the same loop as the CPU, after the CPU, you will have the hottest water of all the system running through your rad.
With your idea, you will not have the hottest water running through the rad, but it will be mixed up with the tankwater... So as I see it you will loose some cooling effect on this setup... But hey... You can get double rads to compensate for that.

good points pelle76 (IMO)...

if you go to the thread i linked to over @ OC, you will see the layout of the res i made. (where the ins/outs are) you'll also notice that the water going into the "hot loop" is coming straight out of the rads (coldest place) this way, the "mixed temp" water is going back into the rads, and the "coldest water" is going to the CPU...(well, pump then cpu)

i understand the possible "gains" in temp arnt earth shattering here, but were fighting for every -1C we can get...right?

had a thought for you.

but what if... you take the line out of the resevior to a pump, then you go to the PC, do all that cooling, then you come out, and you have a Y. but not like you think, on the 2 inlet side you have one from the PC. then the other comes straight from the resevior. after the Y is larger piping than before(2x maybe?) followed by another pump that is stronger as you wanted, sucking not only the water through the pc, but sucking additional water to fill the larger line, like if your PC had 3/8" piping, go to 3/4" piping after the Y. that way the pump fills the line with resevior water, and you can use the second larger pump to blast it through the radiator.

this gets the hot and cold water mixed in a smaller area(no chance of developing hot and cold spots in the resevior) you get the hotter water temp in the radiator atleast partially. and with the larger pump coming second, sucking additional water from the resevior, you get the 2xflow you were talking about.

Yes it is. the water is only the transport medium. The heat still needs to go in the air. (excluding some great setups with a big watertank underground)

And a cascade Silver :rolleyes: ..... but it's not made anymore :cry:
sure, good pump (mcp600 e.g.), a good radiator (heatercore e.g.) and short tubes. and bla bla. (see 90% of the posts)

For the rest AIR. loads of fresh air.
If you can not keep your water temps down (not talking peltiers here) a good block is a waste of money.

The number of times I see ppl waising money on better blocks and better pumps...
and all the time they have the radiator sucking the hot air out of the case.... :D
and then they wonder why the temp is high :p

a) the water in the 2 setups wouldn't be at the same temp. it is the product of the rad(s) ability to shed the constant heat input (from block+pumps) - we know that (assuming the series was delivering a reasonable flow rate in the first place) the (better) flow in the rads in the 2 pump system will have virtually no effect on the rad's C/W (from billA's tests) - for the purposes of the excercise lets take it as nil. therefore the 2-pump system's water temperature will be higher by whatever temperature is required to allow the rad to shed the 2-pump systems higher heat load (block + 2 pumps not block + pump.) in your example the 1 pump system has 150W input and the 2 pump 200w input - therefore the temperature (above the air flowing through the rad) of the water would be 1/3 as much again as teh 1 pump. using your figures the 1 pump system has +3C, and the 2 pump system +4C - the only way your cpu temps could be more favourable in the 2 pump system is if the increased flow in the block loop reduced the block's C/W by enough to be worth more than 1C over water temperature

pauldenton, ok... i can start to see the logic behind that...

so the best idea would be to try to use as small a pump (watts) in the rad loop so as to match the GPM flowing thru the block loop. and since the rads are parallel, the resistance is low (vs WB resistance) so i would want a good GPM pump with (less) concern about head.

now what if we took a sample where the block loop had 2x the rad loop flow... :rolleyes:

like i said, i do plan on redoing this with almost all the same parts (-2nd pump, -res) and then ill be able to put some real life experience behind all this thoery.

im not trying to be a TROLL here, just trying to help us get that last 1-2C off our temps...

thanks for the input (feel free to add more!)

Ares, think you could draw a pic? im having a hard time "seeing" what your saying...

sure, heres the idea. seems possible, need the 2nd pump to be more powerful than the first for a few reasons.

1. thats the only way its effective.
2. if the 2nd pump isnt pulling more water than is coming from the PC, the water could flow back thorugh the Y pipe and bypass the radiators.

essentially your using a venturi design where flowing water at lower pressure sucks water from the resevior with it. pulled through by the 2nd pump, and shot through multiple radiators.

I was playing around with an idea similar to this a 'lil while back. But what I found....by working the numbers only (which is NOT a substitute for experimentation ala BillA) is that you make no gains from having more flow through the rad. I have yet to see if it may work the other way around, with more flow through the block (gotta change the Excell sheet I made up to calculate the balance point). The sheet calculates water temp on a pass by pass basis at the block and rad, then calculates the mixed water temp supplying the block...repeat repeat repeat....usually took about 4-500 loops to balance out. Result....well with the blocks I had detailed enough info on I got slight losses, never any gains. I'm in the process now of rebuilding my rig with components that I have enough info on to play a little more with it.

Ares, if both pumps were pulling from just the resevoir as initially presented, wouldn't any pump suction already be split among the resevoir inlets?

I'd think the mixing of the waters in a resevoir would be ok unless the resevoir was very large or the inlets were placed very poorly in relation to the outlets. If you think about the flow scenario, basically there'd be 2 pumps putting essentially 3 times the usual flow through the resevoir, so the resevoir will be nicely turbulent unless it is quite a bit larger than what seems typical.

As to comments by others, I'd think the water moving through the radiators twice as fast as the cooling loop shouldn't really make much of a difference, though. The radiator will basically shed just as much heat as delta-temp and airflow will let it, so I'd think little gain unless the radiator was already overkill for the setup, then running twice the flow through the radiator as the cooling loop would at best be less than running two radiators in series.

And here's a musing for all those who love to say that delta-tempature between coolant and ambient makes radiators more efficient, I don't see how that's possible. Here's a thought experiement of why: for a constant airflow through the radiator, a constant flow rate, and a constant ambient tempature, graph your radiator input tempatures x more than ambient, and output tempatures y more than ambient. Wouldn't (x-y)/x be the real measure of efficiency? So for the premise to be correct as x increases y approaches x, so a little postage stamp radiator could cool infinitely super-heated water to ambient in one pass.

I would suppose the best-case scenario is then that the radiator will reach a point of load where it sheds a constant (x-y), which would mean it's all about the ability of the air to remove heat from the radiator. So essentially all water cooling accomplishes is moving the heat from the cooling blocks to a big heatsink (the radiator).

This might be slightly OT...

Its an idea I had based on my limited experience of water cooling, where two pumps might be quite useful.
In this case, one pump is driving the "main" cooling loop (CPU, GPU, rad), and the second smaller pump would drive a "secondary" low flow circuit to cool lower power components such as hard drives, northbridge, mosfets, PSU...

The main loop would be 1/2" or 3/8".
The second loop could use 1/4", which has the advantage of being that much more flexible:)

Excuse the lame paint hack of your wonderful diagram :D

As to comments by others, I'd think the water moving through the radiators twice as fast as the cooling loop shouldn't really make much of a difference, though. The radiator will basically shed just as much heat as delta-temp and airflow will let it, so I'd think little gain unless the radiator was already overkill for the setup, then running twice the flow through the radiator as the cooling loop would at best be less than running two radiators in series.


While i'm not sure if Foxgguy wants to do this for the same reasons I can say why I thought it might work b4 playing with the numbers. It is not the added flow through the rads I thought would add some performance, it was the added PASSES through the radiator. Granted added flow thru a a rad produces practically no change in it's C/W. But what happens if the water makes 2 passes through the rad for every one through the block? Will "re-coolling" ,if you will, the water in the loop add any additional performance? As far as I can tell at this point that answer would be no. I was picturing that added passes through the rad would act like having a second rad of about 1/2 the size of the original in series with the first...which it does. But the gains there appear to be outweighed by the gain in C/W at the block that could be made by having all of the flow go through it. This outcome was based on calcs alone and is not by any means proven with experimentation. What I am working on now is rebuilding my system with components that have known C/W
and flow curves curves to test it for real. The way I am going to set it up will allow me to alter the flow rates through the rad or block which will allow me to see if any gains are available from this method or it's inverse (allowng the block to recieve more flow than the rad). Intuition alone tells me that aproach probably won't present gains either.... I will do some calcs b4 I complete the project to estimate what may happen. Why bother doing it?......just fun to play with ...thats all.

if you guys used some data (and there is enough 'real data' out there to be able to gin up some bogus values for calcs)
and calculated the Log Mean Temperature Difference (LMTD) of your rads (air side also)
the efficacy and cost/benefit ratio of your different scenarios can be easily evaluated

but then there would be no 'discussion' I guess