Dual Pumps, Dual Loops, Single Reservoir.
 

I am very intersted in this concept. I believe it can help to increase flow in the whole system, therefore lowering temperatures.

This system is intended for setups with multiple waterblocks. When you have a single pump system, you are forced to either suffer low flow of entire run by running your blocks in series, or by lowering certian parts of blocks by running in paralel. This makes for greater overall flow, but causes lower flow in the paralel runs. By running multiple pumps/loops, you serve to increade flow rates, and also reduce pressure.

What I am thinking of is having multiple loops/pumps to increase flow in those runs. There will be a common reservoir to exchange water flows.

So the first type of setup I am thinking of is one with 3 waterblocks, and a large 2 pass heater core. I want to have one run consist of a pump and the heatercore. The other run will be the second pump and 3 blocks in series. I believe this would give better flow than any other setup with three blocks and a heatercore. Obviously you do not want pumps that will add a great deal of heat to the system, which is not hard because the different loops will be less restrictive, than if you put them all in one loop. I was thinking the new Swiftech pumps would be great for this application.

Thoughts/comments?

BrianW

One thought: You would probably run into the problem of the CPU cooling loop heating the other loop?

Well it is supposed to, that way the radiator dissapates the heat. Maybe I do not understand your idea?

BrianW

Two pumps = better flow rate for each loop as long as you limit your 90 degree bends and such (this is good).

But, I believe your steady state temperature will be the same as a single loop, maybe worst because of the two pumps (not one).

Yeah that is what i thought, but does not matter, with similir temps and higher flow, the resultant chip temps will be lower. And those swifty's only consume 9 watts. Should run really cool.

Also if the rad run is all byself it will be flowing against approx .5 - .75m of head. With the MCP600 that would result in approx. 7-8 lpm. With that flow, I believe the rad would be cooling much more effectively.

BrianW

http://3rotor.homelinux.com/images/c...P03756_sml.JPGhttp://3rotor.homelinux.com/images/c...P03766_sml.JPG
I believe this might tickle your..... ermmmm nevermind... :D


here is some more...

Pretty nice, but I would have two inputs for your res, instead of the t.

I am seriously thinking of hooking this up on my next rig. I already have the iwaki md15rt for the three waterblocks. I will get a MCP600 for the rad loop, and do some flow tests.


BrianW

hi brianw

if you have the space why not run true dual loops?

I want to use this radiator I already have: 12.125 X 7.875 X 2". Do not really have room for any other rads, and that is more than enuff rad for what I will be cooling.

BrianW

Two loops = more pressure, not less. You'd almost certainly get better results if you worked in series (pumps + CPUs). At 100w a cpu heats a modest flow system by less then a degree. With dual pumps the difference would probably be immeasureable and you'd have the benifits of extra pressure per block (due to less tube + running the radiators in parallel).

I am not talking about having the blocks in different loops. I am talking about two loops. Loop one will be one pump, and one radiator. The second loop would be the three blocks in series. Each loop has less components, and there fore less head, than if the three blocks and radiator were in one loop. That is what I meant.

BrianW

Actually, I'm thinking of soing exactly the same.

For me, both Eheim 1250 pumps are submerged into a res. Radiator used is a Silverprop4 with 4 120mm Sunons at 7v
2 loops are used .. just like yours .. One for the RAD the other for the blocks.
Both loops end by dumping water back into the res.

Problem is....the water in the blocks loop may actually circulate within itself, thus making it higher in temp as compared to the Radiator loop. Also, having 2 pumps subermeged may also casue the water temp to rise!!

To counter this, I've decided to place the outlet valve of the waterblocks loop near to the inlet valve of the radiator loop (which is the pump). The same goes for the radiator loop. This way, the warm water from the blocks will be "channelled" into the radiator loop and the cool water coming out from the radiator loops will be channelled into the waterblocks loop.

Although the dual pump dual loop setup will seem to increase flow rates, it might not actually help very much in the overall performace of the setup. Still, it will look very cool and I'm willing to have a go at it :)

Good luck with yours. Post your results k :)

why not just get a little giant pump or basicall a single massive pump and use that? that pump will have plenty to spare, and you could split the flow at the beginning and sent 1/2 to the cpu and 1/2 to the northbridge and gpu. I would think 2 pumps would take up ALOT more room, and be very noisey.

Thats why they are submerged and kept together in the sam container. to reduce the noise and save space. The design I came up with includes both pumps in a res that just slighly bigger than the two pumps combined.

Big ass pump = more bigass heat and noise :)

Quite coincidental aint it georgeteo that the day you ask this question in CMF, the exact same thing comes up here. :D

It would be amusing to see if the concept works overall and if any significant performance does result from it.

The only way to really know would be to try it out. But overall, I still am of the belief that there may be a degredation in performance if not it'll be approx the same overall.

The pumps I plan on using are the MCP600's. The MCP600's only consume 8 watts of energy each. Even with three, that is a great deal lower than any large pump I know about.

Assumptions:

A system with a large 2 pass heater core (12.125" X 7.875"), and three blocks in series would have x flow and y pressure drop.

A system with say two pumps, one pumping coolant from the res to the heatercore back to the res, and the other pumping through the waterblocks(in series), would have higher flow and lower pressure drop in each run, as opposed to being all run in series.

Proposed Idea:

From the flow charts posted here, and my experimintation with my heatercore, I estimate approx 8-9 lpm. The blocks I plan to use are as follows: Cascade , Swiftech GPU block, and Swiftech NB block. Any one care to estimate flow in that loop with a MCP600? I may even consider three pumps. One for the rad, one for the cascade, and one for the GBU/NB.

Based on the fact that these pumps dump nominal heat into the water, can anyone tell me why the dual loop, dual pump system would not cool better?

Ah yes...it mua
Just want to gather more infomation :)

0.02 cents.
 

I've had a similar ideia , but no time to poke around for now.
two thoughts.

how do you make sure all the water go through the radiador loop. and do not recycle hot water to the blocks loop.

how can you state that the full system will be cooler, if all the water is not recycled through the radiator(s).

I believe temps would be similar to, maybe a bit lower than a normal all in series setup.

Here's why: The Radiator loop would be flowing much faster than the waterblock loop. Three water blocks in series is more restrictive than 2 feet of head I believe, although do not know for sure. With the radiator loop running faster, it would ensure that enouph water is being circulated through the radiator loop.

Another note: Even if coolant temps were the same as a traditional setup, the additional flow of this system would ensure lower cpu temps. Is it not true that with constant coolant temp, an increase in flow should ensure lower cpu temps?

As far as ensuring that the water will flow through both loops and that heat cycling will occur can be solved in the reservoir. The return for the waterblock run could be orientated to ensure that the outlet for the radiator run is pulling water from the return of the waterblock run. Vice versa for the waterblock run.

BrianW

Are u testing out your setup anytime soon??

Right now I only have one pump that I can test with. An Iwaki MD15RT. I do plan on getting at least one mcp600 in the next month or so. I will try and test the flow dynamics. And if that proves even moderately successful, I will hook it up to my system and test it out.

BrianW

Still , my questions presist. You cannot make sure.

Are you also assuming the higher flow loop will interfere with the lower , "hogging" most of the water ? That raises another problem. Water has a tendancy of going through the less restricted loop. The radiator loop pump will interfere with the block loop pump. Some water will be recycled through the same loop. Not necessarely a bad thing in the radiator loop.

The switched inlet - outlet will take care of most problems, but wont eliminate all of them.

may i suggest a simple design?
Take that box that #rotor did, make the inlets and outlets as you stated, but insert a wall between the pumps, with perforated holes ( or similar ). There will be communication between the two chambers , but it will be reduced due to the added restriction. Most of the water from the block loop will flow through the radiator loop and very little (if any) will be recycled throught the blocks loop.

i might give it a try once I find out more information :)
Good luck with yours...and hope to see some results soon...
http://www.geocities.com/geo_sin_98/res.jpg
this may explain what he is trying to say

When you state that water will take the least restrictive path, I believe that to be a small effect when both loops have a pump in them. While it is true that if you splith the flow after the pump that water will be affected by the different restrictions. I assume that after the pump, the loop will be on its own, until back in the res. I can not prove it, as of yet.

BrianW

Yes.

One loop consisit of only the RAD so its:
res>pump>rad>res

The other loops consist of the blocks:
res>pump>cpu>nb>gpu>res

to ensure that the water doesn't stay in the loop, I'll simply place the end of the blocks loop near to pump of the rad loop, and vice-versa for the other loop. A devider will be placed in the reservoir to ensure that the hot water coming out of the blocks loops doesn't mix with the cold water coming out from the rad loop.

In this configuration, the setup becomes more like a series loop with 2 pumps.
res>pump#1>cpu>nb>gpu>res>pump#2>rad>res

I really don't know the advantages of having 2 loops as the only way to be certain that its effective or ineffective is to try it out. I will only be able to give this dualpump/dualloop config a spin sometime in early OCT as I'm waiting for my blocks to arrive :)

This is a very true statement,and I think a better way to think of it. Now hear me out on this type of config:

Components:

Cascade

Swiftech GPU and NB blocks

Two Pass Heatercore (12.125" X 7.875")

Iwaki MD15RT ( Cause I already have it)

2 X MCP600

Res 1: One inlet from HC, two outlets to loop1a & loop1b
Res 2: Two inlets from loop1a & loop1b, one outlet to loop2

loop 1a: from Res1>>1/2"ID>>MCP600a>>1/2"ID>>Cascade>>1/2"ID>>Res2
loop 1b: from Res1>>1/2"ID>>MCP600a>>1/2"ID>>GPU>>1/2"ID>>NB>>1/2"ID>>Res2
loop2: from Res2>>1/2"ID>>Iwaki>>1/2"ID>>HC>>1/2"ID>>Res1

What do you think? I believe it would have high overall flow. The block restrictions will be in paralel halving their head with respect to flow through the entire system, and the hc. The Hc is the only part in true series. The blocks are in series with respect to the hc and entire system, but run paralel with a pump each. That pump is also adjustable. So some tuning of balanced flow thru blocks could maybe be realized. Also with two res, I can make the tubing extraordinarily shorter. An added benefit as well. The only reason i suggest 3 pumps is because at only 8watts each for the two MCP600's, little heat would be added to the system due to pumps. The Iwaki consumes a deal more, but does not add much heat to the loop. My Eheim 1250 seemed to get the hc warmer with the machine turned off and no fans.


Brian

If you dont get more flow from using three pumps, I will be very very shocked.

That statement from georgeteo was from an observation I made here in his thread over at CMF.

Regardless, it would be something that I'd enjoy seeing though, just for proof of concept.

Sin22 wrote: There is an added benefit to higher flow. Higher flow will make the waterblocks pump more heat into the water. I believe it has been said that a sweet spot with heatercores is at a certain flow. I do not remember where it was or what the value was. But I believe high flow blocks like the Swifteck models do indeed benefit from higher flow.

Thoughts?

BrianW

Aye I do agree that higher flow is beneficial to waterblocks. BillA's testing has shown us that. However, do look at some of the C/W vs. Flow rate graphs on O/cers.com or Swiftech's webbie.

There comes a point of diminishing returns. Once the knee in the C/W graphs has passed, the decrease in C/W in relation to increase in lpm becomes a linear progression, with a fairly shallow gradient IMO. Then further translating the C/W into real world temperatures. How much more temp difference or performance difference will you be obtaining by jacking up your flow from 5lpm to 10lpm ?

Taking for example the MCW5002 graphs.

That doubling of flow rate resulted in less than 0.1C/W increase in performance. Real world assuming an 80W temp source would be a change in 8deg. A fair amount in theory. in reality though?

I agree that high flow is definitely worth it, but at what cost?


As to the sweet spot for heatercores. I believe the Thermochill evaluation done @ O/cers.com was what you'd be looking for.

To me, reading the graphs there, it would appear that the base trade off would be a 5~7lpm flow rate inside the rad (120mm sized rad) before once more a point of diminishing returns takes place.

But all of the above is my own personal beliefs.

I am interested to see what sort of performance you do get from it. I mean at the end of the day you can make all these theorectical comparisons and comparisons with data that is seen online, but its still doing it yourself that matters. Even if you dont get the performance that you may have initially wanted, its an experience and something that would educate both yourself and those who kept track on your work.

I did something similar with my res . My bleed fill lines connect to the res. The res is a PVC T fitting. The josnson pump is connected to the side outlet. The top line is the air/water exit. 1 of the lower lines is a fill tube while th other 2 are just returns.

If I flush the system with a soap mixture some will run out but most will run through the system. It takes about 2 minutes of running water through the fill tube and letting water out of the escape tube to get all the soap out. It also makes flushing the air out of the system pretty easy as the escaping water carries out.

My point though is the water gets pretty well mixed in the res. The only problem I'd see in using a secondary cooling loop would be the fact you could never get the coolant to ambient. However I could see using a secondary loop for use with some kind of filter.