Watercooling Radiator bypass idea

[Owenator]

I posted earlier about using two pumps and a reservoir to achieve more flow through the water block (big pump) and less flow through the rad (small pump). The idea is that heat transfer is better in the water block under higher flow and better in the rad under lower flow. My existing cooling loop and my 'bypass' idea is shown in the pic below. I am going to add a bypass loop that will allow me to 'split' the flow to where the block gets more flow that the rad via the valve before the rad return tee. The bypass leg has a valve so I can return to full flow through everything later if I like. I figure this way the cooled water from the rad is going into the water block along with the the 'split' recirculating water. Any comments/suggestions?

O

PS Someone suggested this to me so I cannot take all the credit I just didn't look through my old post to find out who suggested it. Thank you to that person, if it works of course ;)

[WarriorII]

"Theroretically" H20 takes the path of least resistance. By your pics; if you shut the valve to "slow" the radiator flow, it seems you'll loose that flow completely.:eek:

Trial and error. hopefully not too much error.

Maybe just a BIGGER pump, more gph.

Good Luck.
:cool:

[Boilerhead]

I think it is a good idea and worthy of a try.

You would be better off with globe style valves than quarter turn ball valves for this so that you could fine tune the flow through each valve. I imagine that the range of adjustment where the flow path is dramatically altered will be fairly small. You want to maintain high aggregate volume on the leg into the block and the potential for throttling the return circuits while trying to balance the flow for each leg is high.

It would be nice to have some flowmeters to visually keep tabs on what's going on, even the cheapo water wheel kind with a sight glass to see the spinning wheel.

This is of interest to me as I just wrote about the same heat transfer vs. velocity theme on another thread I started at http://forums.overclockers.ws/vb/showthread.php?s=&threadid=55674

A better solution would probably be to increase the size of the rad or maybe add a second one, but adding a bypass cicuit might be a viable alternative for limited space situations such as a fully in case wc deal. It would certainly have merit in a parallel circuit dual radiator setup where the rads were of different types and had different heat exchange qualities. It might be worth looking into for the guy with the 4 rad bong deal with both aluminum and brass heater cores which is posted elsewhere here.

Of course once you get near ambient it's all moot unless you refrigerate an exchanger which is also one of my interests.

Henry

[ButcherUK]

Problem is you end up recirculating hot water. Also the chances of the rad being the limiting component are small. Most comp rads are massive overkill, the blocks are not, so in general just having a higher flow is better than trying fancy tricks to maximise the rad.

[TruckChase!]

Yah man, I think you would end up just sending uncooled water to the cpu, not to mention that both the waterblock and the radiator wouldn't be as effective unless you were able to make sure the water could fill all the passages completely. Air in the cooling system tends to negatively impact performance.

Besides, the flow rate hasn't changed, it's just been redirected to two paths. There will be less water at a lower flow rate being cooled by the radiator. The low flow rate in the radiator wouldn't account for the difference, and the efficiency of the water block would suffer as the flow rate for that drops.

If you want to better cool the water without going to a multiple pump scenario, just get a good strong pump and put two radiators inline. Try, for instance, using your stock radiator and then mounting a danger den low pressure drop cooler mounted inside your case using a pre-existing exaust (or intake depending on your config) fan as it's active cooling.

[ButcherUK]

No, you want two parallel rads, then you get lower flow through the rads while still having high flow in the water block.

[RycheX]

I fail to see the logic in this idea. The idea behind having slower flow through the radiator is to get the water as cool as possible so that the temperature difference between the water and the block is as high as possible. All you are going to do here is raise the temperature of the water going through the water block, thus lowering the temperature difference between the water and the block, and reducing the effectiveness of the system.

The only way that the thermal energy from the cpu is getting dumped out of the system is through the radiator. All you would be doing is trapping the energy in the system and therefore raising the average temperature of the water and the average temperature of the system. This idea just simply doesn't pan out according to the laws of thermodynamics, sorry.

[Hoot]

Do you mean something more like this two-stage process?

73, Hoot

[Owenator]

Thanks for all the responses!

Hoot,

That was my initial idea. I am trying this simpler bypass idea out to see if it has any merits. I should know by this weekend because I am going to do the mods tonight or Friday.

Now, let me say that I am tinkering here and didn't do a first principles analysis of this problem. I do enough of that at work and at home I try to fight the urge to 'engineer' everything, instead I am just trying different things. I have seen a blended two path systems like this in industry it is used to remove heat from fluid systems where you don't want to cool the entire flow rate.

The idea I was shooting for was to improve the heat rejection in my radiator with a slower flow rate while keeping the water block flow high. I will probably buy some flow meters if I can find them cheap. I think I will also me buying another heater core rad to split the flow path to. I wanted to try this first to see if I can get any benefit without spending much $. So far I have the ball valves (I know they're not drawn with the correct symbol in the pic) and tee's so I figured why not. I've got tubing to spare as well.

As far as the thermodynamic arguments I believe my existing rad is over sized and at the flow rate it gets it is exponentially less effective. If that is true then the flow path I propose would be attempting to get a greater increase in cooling than the loss due to the uncooled fluid. In other words the flow may be half of what it was but the temperature difference may be three times better so if: Q=m*cp*(Tin-Tout)
where:
heat rejected = Q
mass flow rate of water = m
specific heat of water = cp
Temperature difference = (Tin- Tout)
My hope is that:
Q=1/2*m*cp*3*(Tin-Tout) vs Q=m*cp*(Tin-tout)
or Q=1.5*m*cp*(Tin-Tout) vs Q=1*m*cp*(Tin-Tout)

So I can gain some heat rejection. Of course if my rad rejects heat linearly than I an all wet as it were ;) If my analysis is incorrect please let me know. I am looking for the correct answer and take no offense to all of your comments and suggestions, even the ones I seem to reject at first. Experimentation to me is half the fun of overclocking!

[Les56]

Previously posted :
[url]http://forums.overclockers.ws/vb/showthread.php?threadid=52333;/url]

I use a dual syphon-connected reservoir setup employing 3 pumps to cool a 430watt system(220wPeltier 300w[13.8v,21.5a].,CPU 80w{Duron @1200mh,1.94v],pumps 50w(Eheim 1250,2xAqael600)).Two pumps for cooling and one for the waterbock :
http://www.jr001b4751.pwp.blueyonder.co.uk/P0001036a.jpg

This Heath-Robinson set-up of one radiator ,one copper coil ,and two fan-blown dash plates/trays attains a steady state resevoir temp of 29c with an ambient of 22c. The Eheim1250 produces a 480 l/h flow rate through a Maze2.2 waterblock.

[TruckChase!]

Originally posted by ButcherUK
No, you want two parallel rads, then you get lower flow through the rads while still having high flow in the water block.

Yah that'd work better.. I'm just putting mine in series to reduce complexity since my one radiator pretty much does the trick anyhow.

[SteenkyBastage]

in not so technical terms, here's how i see it...

the idea of a radiator is to take as much of the heat as possible from the water and transfer it into the air (wont go below ambient).

the idea of a waterblock is to take as much heat from the core as possible and transfer it to the water.

the greater the differences in temperature between the water and the core (same for ambient air), the more effecient the transfer of heat.

the faster water moving thru the waterblock at any given time will equate to more heat being able to transfer to the water (altho how much per unit of water depends on the temperature difference). faster flow thru the waterblock will give lower temperature water after leaving the wb. (same ammount of heat, just spread out over more water)

the slower the waterflow thru the radiator will thus allow more air to cool the given unit of water (more time = more air). this also is dependant on the differences in temperatures.

your theory supports the flowrate idea, BUT it doesn't take into account the fact that the water will instead be warm in both the radiator and waterblock, as opposed to what it should be: hot going into the radiator, and cool going into the waterblock. by mixing the flows at any given point (whether it's by using 2 pumps, or by using a bypass) you're hindering both the waterblock and the radiator's temperature difference between the water and core/air.

IMO, the more effective way would be to simply place two radiators in parallel. that slows down your flow thru the radiators, as well as gives twice the total capacity of the radiators. after the water leaves the parallel radiators, it then merges back into a faster flowing single tube which goes to the waterblock.

that's the only way i can think of to not have to combine the hot/cool water at all. if you do combine hot/cold, there is a certain ammount of heat that can never be transferred to air via the radiator, as it will never make it to the radiator.

[Owenator]

Any thoughts on this type of system instead (see pic). I am gathering that the consensus here is that I should really try a second radiator in parallel.

O

[Boilerhead]

I wouldn't even try to use a system such you pictured Owenator, it just adds a further level of inefficiency from what I can see. You are not actively cooling the cpu circuit with the heat exchanger circuit at all, at least it looks like you are ganging three radiators together in that one.

Radiator to radiator cooling using room temperature air is pretty much a waste of time. The only way you could get any practical gains from that setup would be to have the secondary circuit refigerated, but it would still be grossly inefficient.

You'd be far better off to either run the three radiators in series or run two smaller ones in parallel then the large one in series with the two parallel ones.

For failsafe and high flow purposes you could put one pump each on the parallel radiators with low restriction check valves and then if one pump crapped out you would still be running at a minimum of half normal system volume which could save a cpu.

Put a couple pinwheel sight glasses on the parallel circuits and you have visual confirmation that each pump is working.

I think that would maximize the heat exchanging with the hardware you have pictured.

Henry

[Owenator]

Henry,

Actually the smaller device isn't a radiator but a water to water heat exchanger (HX). I do think that is adds complexity and I am not yet convinced if it will work any better. I could make a water to water Hx that is basically like having two waterblocks back to back. I actually have a few spare water blocks so I may actually try this if I can find a small cheap inline pump. All of this is in an effort to explore radiator vs water block flowrates which is more experimental than practical. I am not really unhappy with my existing setup just looking to explore some ideas.

O

[Diggrr]

I think you might be a bit disappointed using the two waterblock interface. Waterblocks work great with a high temperature differential (hot athlon vs. room temp water) but I don't think the temp transfer would be as good with both loops near the same temp.
I'm leaning towards the common reservior idea instead. This would guarantee good temp transfer from loop to loop.

Of course, this is just an opinion, your mileage may vary. :D
It's nice to see an experimenter on the loose.

[Boilerhead]

I completely agree with Diggrr on this. The best heat transfer will occur when there is only one coolant circuit being actively cooled.

Common reservoir would undoubtedly cool as well as a parallel rad system, but you are still duplicating hardware resources without the security of being able to have the CPU circuit pump fail and still have active pumping on that circuit.

I have no idea how much cooling you might get out of convection alone, especially on OC. Might be a worthwhile project to underclock and simulate a pump failure to see how high CPU temps get. Hmm.....

Henry

[Rack]

http://www.hardforum.com/showthread.php?threadid=273754
My thoughts on the bypass system.

Strangely enough, I came up with something similar to Hoot a few
days ago, but I was going to take advantage of the water's thermal gradient (or try) pulling colder water to the block off
the bottom of the res, and dumping the hot water on top for the rad to skim off, and then having the rad water go back to the bottom of the res cooled down for the waterblock yet again. Maybe a nice tall transparent tower. Hell, I was even going to try and make it a whirlpool to promote separation, but that's a project for later. I got a fun project tonight, reading a Morgan/XP's internal diode. Should be fun. :)

I'll have to give this thread a thorough read, but the thread seems to be light on hard numbers.

[RycheX]

You have to remember that anytime you have any kind of interface in a thermal system, efficiency is going to be lost. The more interfaces you introduce, the quicker your efficiency is going into the pot. You really want to keep the entire thermal path as short and as simple as possible. The best solution that has been presented (by far) is to have two radiators in parallel.

[Owenator]

Originally posted by Rack
http://www.hardforum.com/showthread.php?threadid=273754
My thoughts on the bypass system.

Strangely enough, I came up with something similar to Hoot a few
days ago, but I was going to take advantage of the water's thermal gradient (or try) pulling colder water to the block off
the bottom of the res, and dumping the hot water on top for the rad to skim off, and then having the rad water go back to the bottom of the res cooled down for the waterblock yet again. Maybe a nice tall transparent tower. Hell, I was even going to try and make it a whirlpool to promote separation, but that's a project for later. I got a fun project tonight, reading a Morgan/XP's internal diode. Should be fun. :)

I'll have to give this thread a thorough read, but the thread seems to be light on hard numbers.

Now reading a Morgan/XP's internal diode sounds like some info we could all use! I was wondering if anyone was looking into the internal diode reading.

As fas as hard numbers, I think half my temp issues are due to my in socket thermistor. I need more accurate measuring equipment for temp and flow and then I can begin to 'tune' the system. I think a couple of flow meters and either a digital thermometer or a DigitalDoc should do the trick.

I think I need to stop speculating and get down to some engineerin' analysis of this project that would be more informative. Still tinkering is so much fun !

O

[Rack]

...me! The project last night pretty much consisted of hooking the wires to the right pins on the back of the motherboard, oh and putting the new Morgan in and installing windows on a fresh partition. I had soldered up the reader a few weeks ago. Here are my preliminary results:
morgan 1GHz@1GHz, 1.85V, MC370 w/38cfm delta
room: 17C
normal idle: about 35C, kinda forgot about that one.
reg52 idle: 17 1/2C
k7burn load: 45C
sensor: MAX6657 remote, POR defaults
program: SpeedFan 4.03 beta 3

[Owenator]

Originally posted by Rack
...me! The project last night pretty much consisted of hooking the wires to the right pins on the back of the motherboard, oh and putting the new Morgan in and installing windows on a fresh partition. I had soldered up the reader a few weeks ago. Here are my preliminary results:
morgan 1GHz@1GHz, 1.85V, MC370 w/38cfm delta
room: 17C
normal idle: about 35C, kinda forgot about that one.
reg52 idle: 17 1/2C
k7burn load: 45C
sensor: MAX6657 remote, POR defaults
program: SpeedFan 4.03 beta 3

Cool! You should write it up for the front page! I could sure use a way to read the internal diode temp when I get a morgan chip. Getting temps from in socket thermstors always leaves me wondering how accurate they really are.

O