|| flow balancing

[nemaste]

i was thinking about hooking up an inline system with the CPU parallel to the MB & video water blocks (which are serial to one another). i know that T-fittings have less flow restriction straight through than they do flowing around the 90degree bend. so i thought about using the straight through connections on the T's for the CPU side & installing a ball valve on the PCU side to force some of the water through the video/MB loop. i thought going back to a straight forward series loop would be a simpler solution (no parallel flow, CPU->MB->Vid->pump->rad->""). After all, simpler = more reliability in general.

i'd like to know if this is practical & if there is some person or page out there can demonstrate a well thought out valve design.

i was also curious if anyone has shown a significant improvement in cooling with parallel config over a series config (identical single-loop systems).

[Cyco-Dude]

instead of using T's, why dont you just use Y's? less resistence plus you can ditch the ball valve thing you got.........

id try it both ways (serial and parrallel) and see how it works out.

[Brad]

he wants to set it up so that the cpu water stays in a straight line, while the GPU water turns 90 degrees. Very good idea IMHO

[nemaste]

cycodude, i'm completely with you on that thought, but none of my local stores have the Y's like the ones Daden is selling (i think that's what you were talking about). those would be more efficient than T's regardless if a control valve is used or not.

my thought of using a ball valve (gargatuan space killer BTW) is to have some level of control over the flow ratio between the two sides.
[edit: i thought my idea of manipulating the bends in the T's would work well too. thing is, i'm not 100% sure if the resistance from the blocks might force me to change the location of the valve or orientation of the T's. for example: it could be that the CPU block alone creates more resistance than 2 90degree bends & the GPU & MB blocks in series (albeit unlikely)]

i just read the Koolance review in the April edition of Maximum PC (r0x0r5). Gordon firmly dinged the Koolance system for not using "point to point water lines", which is jargon for no serial blocks (all blocks in ||). while controlling 2 flows would be a challenge^1, balancing 3 would be a challenge^3 (not to mention i'd likely have to buy more matching 1/2" silicone tubing & add another huge ball valve). do you guys think adding a 3rd || loop to separate my MB & GPU blocks would be worth it?

[jtroutma]

Three channels worth it......... in a word.........No!

That would be overkill not to mention very difficult to adjust. Look at it this way:

You would have three separate channels of water all going to three different blocks with different resistances. Say your NB block has less resistance than your CPU block but more than your GPU block. In that case more water will flow through the GPU block than the other two. On top of that, less water will go to your CPU block than your NB block becuase it now has a lower resistance that your CPU block.
To fix that you would resitrict the GPU block but now you have just changed all the resistances to the whole system and have to start from scratch again!!!

It just like trying to tune a 6 cylinder engine that had 6 independent carbs. you change one and you have to fine tune the others to compensate (a good firend tried to do this once with a Jaguar, he never did get that thing to run right)

IF you had enough time and constantly worked with it, you COULD adjust it for optimal cooling on all units but the hassle would not reap very big rewards.

I would just do two channels; one for the CPU, one for GPU and NB in series w/ ball valve if your CPU block has a lot of resistance to it.

[bigben2k]

It's hard, but not impossible. Actually, the parallel design increases the overall flow.

The problem is that the valves would have to be set to seperate the flow in such a way as to get a specific ratio. It's a problem because you can achieve the same ratio wether the valves are all mostly closed, or all mostly open.

Ideally, you'd leave the CPU loop open, no valve, then put a valve on the GPU and Chipset cooler loops. The more open these valves are, the better overall flow, but less flow to the CPU.

Running the WB in series is far simpler, because there's no adjustment, but then, there isn't as much flow.

The ideal setup would probably be either:
1- a parallel loop, with the GPU and chipset valves barely open
or
2- use T fittings, where the CPU feed flows in a straight line through the T. (as you propose)

1-This would divert most of the pumps flow through the CPU only, with a little bit of relief from the other two loops. You'd achieve near max flow rate through the CPU block. (where max is one loop through the CPU wb only)

2-You'd achieve a similar result as #1, but more flow would be diverted to the other 2 WB than you probably want.

What it comes down to, in terms of flow dynamics, is how each block resists flow. Usually, the CPU block is (by far) the most resistive, as GPU and chipset coolers are often simple open block designs (i.e. no channels), and as such, have very little restriction. The purpose of the valves is to correct or compensate for those flow restrictions in such a way as to divert more flow to the CPU block.

I hope this helps... In short, I'd go with the valves on the GPU and chipset cooler loops. You'd have to monitor the temps, to make sure that you achieve the cooling that you want for those 2 WBs.

[nemaste]

my thoughts exactly bigben. i'd have no ability to "target" the cooling where it is most needed if i used 2 || lines (CPU ||

MB&GPU), where the CPU path goes straight through the Ts, & no valve. if i add a valve to the GPU+MB line, i can divert the

flow to the CPU. if the GPU & MB get too hot I won't be able to divert flow from the CPU. in this case, I would then want to

locate the valve on the CPU line instead. predicting which will be the case would require experimentation or knowing friction

coefficients for the blocks, but i think i have made an initial decision. i plan on using 3|| WBs with valves on the GPU & MB

lines. BTW, the blocks are:

Maze3
Daden "video"
Daden "chipset"

i don't need to go overboard & do anything like experimenting or calculating the flow dynamics to get reasonable results

anyways; as long as i give the MB & GPU WBs enough water to keep them below a certain temp, the system will operate more

efficiently than without valves (max CPU flow available, minimum flow restriction). i think a 3|| system may have significant

benefits considering the WBs very restrictive & the math of flows is analogous to electics; || WBs:|| resistors, series

WBs:series resistors, well sort of (where friction:resistance, pressure:voltage, & GPM:current). i'm not experienced with h2o

yet, so i'm not sure how restrictive my BIX will be relative to everything else:/ i'd need multiple rads to reduce the "heat exchanger" pressure drop anyways. however, since i already have 3 WBs i might as well keep them ||, right?

i have a Digidoc 5 to adjust this system with, so it shouldn't be hard to pull off (i know, it's too late to avoid the DD. i

bought it months ago).

off the topic of my rig, wouldn't that be schweet to make an h2o system that has a PIC based controller that thermally

regulates valves, fans, & pump(s) & communicates real time with a PC?