just a quicky

[rrjwilson]

The pump (reserator) was the part that i wanted to know as this is more crucial at the moment.

I would believe that your system as it stands has masses of flow to your SB and GFX but NB, CPU and MOSFETs are straving. If this is correct my assumption is correct and therefore the fix may just work.

Your system is slightly obtuse to normal but thats just means its fun. With the pump feeding from above its is driving the coolant down the pipe to your splitter but they are at right angles to the original direction of flow so most of the flow becomes gravity driven and therefore bypasses most the other pipes and only the distrubance of the opening will cause flow to those pipes. Reorientating the splitter so the flow must go sideway from a more central joint will provide better flow and hopefully more to the other ones but because the flow has no actual diversion like hitting a wall the choice of direction will rely on other things such as gravity. If this works great, if it does not you need different splitters to help split the flow better.

Famous last words. I have a cunning plan.

6 T-piece splitters (3 for each side - 1-2-4 splits) Use right angle elbows for ease of joining after. The problem is the space this requires is far greater than your current ones. This will suffer a partial problem like your current setup if placed similarly so try to mount the splitters horizontally then have the feeds going to the components vertically. This should create the correct stress on the flow the allow even distribution and no shortages.

 

[agentx]

so i basicly have to remount my splitters so they run horizontal instead of vertical. doing this wont be so bad as i have to send my mobo in (i think it died) so wile its out in rearange the splitters. but if this doesnt work what are these t-piece splitters u mentioned?

 

[agentx]

what if i reversed the main in line to feed from 5

that way the coolant will be forced up through the splitter to all the blocks.

 

[rrjwilson]

what are these t-piece splitters u mentioned?

There are many plastic pipe, clear, delrin and even brass. The brass and plastic work but are just bits. The delrin and clear ones have barbs which mean you can attach other barbs if needed, like the elbow barbs I suggested.

what if i reversed the main in line to feed from 5

that way the coolant will be forced up through the splitter to all the blocks.

This is unfortunately not true. Gravity merely helps your current situation to the bad outcome it isn't the actual cause. Turbulence is the cause of direction change without it there is no change. Having it the other way will be almost as bad a before (yes, almost because gravity is not helping this time), the reason is flow from your pump is the driving force of the system not gravity. I will try to do a simplistic 3D flow model (in paint as I don't have access no more) for all three ideas to try and illustrate the problem for you.

 

[EazzyTarget]

With the pump feeding from above its is driving the coolant down the pipe to your splitter but they are at right angles to the original direction of flow so most of the flow becomes gravity driven and therefore bypasses most the other pipes and only the distrubance of the opening will cause flow to those pipes. Reorientating the splitter so the flow must go sideway from a more central joint will provide better flow and hopefully more to the other ones but because the flow has no actual diversion like hitting a wall the choice of direction will rely on other things such as gravity. If this works great, if it does not you need different splitters to help split the flow better.

I was thinking that in a pressurized system -- like these liquid cooling systems -- that forces like direction of flow and gravity were non-impactive since the back pressure created by block resistance would re-route the liquids through alternate channels.

For example, in my rig the "T Splitter" is fed through one of the side channels of the "T"... not the base of the "T". I believed that because the system is pressurized that the liquid would divert equally to each block...

Is this flawed thinking? Maybe I am cooling the second video card more than the first... on accident!

Thanks.

-EazzyTarget

 

[rrjwilson]

A single split give better flow anyway by nature however the pressure will never outright control fluid dynamics occurring in the tubing.

One card will be getting more but whether it is 90/10 or 70/30 will rely on orientation of the T piece gravity affects everything on or around tthe planet regardless or size or pressure or temperature (well apart from a bose einstein condesate at least).

 

[rrjwilson]

Okily dokily that is my MSpaint CFD plot for you. Plots never show direction only pressure (usually reffered to a static pressure as tests try to imply static rather than dynamic but nevermind that.

Red is high pressure - Opposes flow and forces a change of direction

Orange

Yellow

Green

Cyan

Blue is normal pressure

Light blue

Lighter blue

White is low pressure water - Causes direction change by lack of pressure and want to equalise pressure within the system.

First one shows your current splitter system. You can see the large low pressure area on the lowest split and that this area decreases in size as you move up the splitters. This is because the main flow bypasses all of the splitters because no high pressure is created however, the lowest split has the most back pressure given the flow going to the bottom and so more flow is redirected. The other splits are done by pure velocity change.

The real thing to notice is the pressure and speed differences are fractions of normal measurable tolerances but thye do have an effect.

The T pieces show the changes of direction from different feeds. Where as the top feed has the high pressure to redirect flow, the side feed redirects by relative vacuum in the water. The high pressure redirection is more powerful and as such are what we want to control flow in our loop. Relative vacuum in the water is nice and all and will create a certain amount of flow but it is difficult to maintain and as such is weaker than the high pressure alternative. Being weaker means it does not work as well and is therefore not as favourable.

Here is a 1-2-4 way split that will split the flow evenly (as possible) to every outlet. This is a 2D image and therefore flawed as you can now hopefully read from the plot however, think of these as 3D plots using tubes as boundaries. Thinking of 3D I believe the same 3x T piece setup would work but the middle T-piece would have to be upright in comparison to the flat T-pieces.

This method will split the water to a better standard than your current split. However, this split requires sufficient force from the pump. The critical part is still orientation, the middle T-piece must be fed vertically, the end pieces will feed horizontally. This creates the even as possible results any other way and flow will favour certain tubes.

 

[EazzyTarget]

Ritchie,

Thanks for those great pics! I am however confused, the orientation of my splitter is like your "Side" picture, and while that one doesn't show an arrow, I am assuming it flows from the top like all others.

Mine is setup exactly like this, with GPU1 being cooled by the equivalent of the right hand side outlet of your picture -- which in my rig is cooled better than the second GPU -- which seems wrong (see the picture here http://forum.overclock3d.net/showthread.php?t=23407 -- inflow to the GPU's is on the left, outflow is on the right).

I guess other factors can be the impactor, the deviance is not that large -- only 8 to 12 degrees Farenheit -- so maybe that's all I am seeing in the differential, something like thermal grease, passive air cooling, etc.

Anyhow, thanks again for your post -- very enlightening.

-Eazzy

 

[rrjwilson]

Thermal transfer depends upon a few major variables and some minor ones. Time in contact with the surface and temperature difference between the two surfaces being the main ones. The longer the contact time the more can be transferred (which may lead to equillibrium being reached if no more heat was added). The larger the temperature difference (dT) between surface the greater the transfer that can occur. Watch temperature of hot tea cooling, it will be 80C to start then inside 5mins be 50C then 10mins to 30C 15mins to 20C.

Think of contact time as a shift time for little people to move bean bag to the other side of the river.

Think of the dT giving larger bean bags for larger difference.

Now you'll see the optimum point can be reached with subzero coolant (massive bean bag) and low pump speed.

This is not the end though. Low pump speed means loop probably wont run, so speed must be up enough to go through the loop without problems. You try fitting enough coolers to provide maximum dT to all parts.

Many will argue that the coolant passes through the radiator the same amount of times in a loop per minute so why worry about the heat exchange.

Reasoning behind this is, if the pump does 2lpm and the loop is 1l then the coolant circulates twice every minute with 2 chances of cooling whereas with a 1lpm there is only one pass but double the amount of time in the radiator. Nice try but still lacking, time in the radiator for the first scenario is fast and therefore heat exchange is reduced maintaining more heat in the loop making the next loop cycle hotter from the start. This means an overall balance will be reached but will inevitably be higher because heat exchanging is not happening for long enough at the radiator end. The second scenario allows the water to have more time to exchange creating a larger heat difference ready for the next cycle so overall heat load will be reduced and therefore loop temp to be reduced.

As for your GPU cooling thats actually not to easy to explain without a few years of fluid dynamics behind you. Basically the force of that "little" vacuum area is dependent on one thing, Velocity. With your pump you probably have some real speed going which means the force dragging the water in the other direction is greater. Also like I said in the original explanation, imagine the tube haing all the flow in it. For a 1/2" your talking a normal size vacuum area but make it smaller and the forces are enlarged creating greater vacuum to pull more away.

Well thats the short version without too much stupid detail.

 

[agentx]

ok short of ripping the config apart and retubing everything is it possible to have the master in line at the bottom (were #5 is) making the pump push the liquid against gravity and creating enough pressure to sustain a "good flow" by this meaning at the very least .5 l/min. please calculate if a max head pressure of 2.2m is sufficiant to do this. i have gained flow by doing what i have mentioned above (making #5 the master in) the gfx cpu and sb get better flow the nb is less but the mosfets still dont get any.

 

[rrjwilson]

1. You haven't given me the new placement for each attachment so I couldn't "run" the numbers.

2. I don't know the flow restrictions for your individual blocks so cannot "run" the numbers.

3. I don't have access to my old university specialised computer to do CFD modelling which even on that takes a few hours.

Assuming the pipes are simply moved up one then your results imply the makor restriction is actually within your blocks. Two blocks have recieved the same flow in both your setups the MOSFET and GFX. The MOSFETs receive no flow in either first time because flow wasn't going the right way but now it should receive the full force of the pump and it clearly isn't. GFX has received good flow in both and it is in the middle of the splitter setup indicating large low pressure forming in the pipe on both occasions as being the cause for the flow.

Without retubing with other parts your only option to gain better flow is to attach the feed onto #2 and orient the whole spliiter setup sideway in comparison to its current setup. This will stop gravity taking part and force a mostly even split.

Will the better pump help? Yes but I cannot be sure it will solve the problem just by power.

 

[agentx]

ok well i'll order the pump and see if that helps as the added pressure should be enough. thnx for the help so far, much appriciated .

 

[rrjwilson]

Tis ok mate hope it all works out.

Just difficult to do 3D calculations for flow through something you've never measured in any way shape or form without the thing and a few hundred thousand pounds worth of computer.

Going up should help a bit but I believe (as I said when you first mentioned the opposite feed idea) it will be only slightly better.

Oh and BTW I'm getting increasingly bored of waiting for the Back Ordered status to change on my bits so I can control my WC setup.

 

[agentx]

ya waiting sucks i hate having something thats not 100% complete lol.

i also found a more powerfull pump the koolance pmp-400 it has 15ft of head so im hoping this will do it as its the most powerful pump that still has the fan header connector. should have it ordered in a day or 2.

 

[agentx]

ok i got the pump today and have it installed in my xt now just waiting til thursday when my replacement mobo gets here. (last one died on me while doing the w/c portion of the build )

 

[EazzyTarget]

sustain a "good flow" by this meaning at the very least .5 l/min.

agentx,

In most of your posts, it sounds like you are able to actually measure pressure and flow...

I was wondering how are you measuring these? Do I need a new tool???

-EazzyTarget

 

[agentx]

lol no i cant messure the loop pressure as i dont know of nor have the tool that would be needed. but the reason u see pressure referred to so much is becuase i dont have enough lol