Am I crazy?

[-Dragon-]

I'm wanting to convert my system to a water cooled setup as my next major upgrade for my system. Only problem is the system I want to create sounds a bit... excessive when compared to the average...

My background is I spent 6 years in the navy as a nuclear reactor operator so they gave me a lot of engineering training on industrial cooling systems, so I'm looking at building my computer cooling system from a perspective of overkill = good.

Here's what I envision:

Reservoir as high in the case as I can get it with at least 3 inputs and 2 outputs. The 2 outputs would go down to 1 (likely variable speed) pump each, providing good NPSH for both pumps. The pumps would feed 2 separate radiators, however there would be a cross over between the radiators and the pumps to provide cross flow in case of pump failure/imbalance to still utilize both radiators. Both radiators would feed opposite ends of a 2:3 manifold where the coolant would be diverted along 3 paths, CPU, GPU(s), and Other (NB/SB/Mosfet/RAM/etc...) I envision the CPU would have the largest ID tubing (1/2in, with the other line having the smallest (3/8in), and GPU I haven't decided, maybe 1/2 maybe 3/8, depending if I went dual GPU or not. Then each of the 3 lines would connect back to the inputs on the reservoir.

Pros:

• Redundant pumps, if one fails the system wouldn't necessarily fry itself in a hurry
• Multiple flow paths for higher total system flow from both pumps
• Minimizes problem of water temp rise across multiple components reducing system effectiveness.

Cons:

• Expensive, requires 2x pumps and radiators plus additional tubing and connectors.
• More tubing = more leak potential

I'm not too worried about the expense part since the increased expense pieces are reusable in future systems where fixed cost parts like GPU and MB cooling blocks may require replacement anyway. As far as leaks go I figure I can make the system tight and use water substitute that's non conductive.

What do you think? Too much, or just might work?

 

[Thor]

ok, this is one of the places where KISS applies... (keep it simple stupid) ... the more parts you have the greater your chance of failure, ect.


now then... if it were so easy, would we water cool at all... no, we wouldnt...

while i get why you want dual pump to help flow rates, im haveing somewhat of a hard time understanding you res layout theroy; your wanting to use a res with multi in and multi out, but the outputs will be straight into the pumps, which will put it into two more res's???
really not needed, if you wanted to it would be better to send that output to 2 radiators, and you could use for first res for water mixing between what would other wise be 2 sepreate loops...

the issue with the mutiple flow paths that you have set up is that water like electricity, will always take the path of least resistance... if pump a fails, even thought you have a pump on loop 2, and the water mixes in the res, it wont flow thought loop 1 becuase pump a will induce flow restriction on that loop.
while you may not loose yourvid card, you may fry your cpu.. or vise versa...

there is only 1 way to over come pump failure, and that is to use a single loop, with 2 pumps in parellel basicly put a Y fitting in befor the pumps, split the flow, and then put a Y fitting right after them again and bring it back into a single path... you only end up with 4 extra failure points this way vs dubble a single loop with your proposed method.

now then.. i have never had a loop failure that i didnt cause my self... if you leak test useing the proper methods, this isnt an issue.. use a good quality pump... and it wont fail for a good long time.. tubeing is cheap.. buy extra.. you will screw up the first few cuts... buy nice hardware, and get nice hardware, you get what you pay for..(for the most part...)

 

[-Dragon-]

really not needed, if you wanted to it would be better to send that output to 2 radiators, and you could use for first res for water mixing between what would other wise be 2 sepreate loops...

Read again, that's what it's doing

the issue with the mutiple flow paths that you have set up is that water like electricity, will always take the path of least resistance... if pump a fails, even thought you have a pump on loop 2, and the water mixes in the res, it wont flow thought loop 1 becuase pump a will induce flow restriction on that loop.
while you may not loose yourvid card, you may fry your cpu.. or vise versa...

Both pumps would at all times be effectively feeding all 3 loops, the concept behind water or electricity taking the path of least resistance isn't an absolute, i.e. MORE water will flow through the path with the least flow resistance, but some water will still flow flow through the high resistance paths. In an electrical circuit with a 1ohm and 1Mohm resistors in paralelle, while most of the current will flow through the 1ohm branch, a small amount will still flow through the 1Mohm path. The flow restrictions won't be nearly as severe in my case, and I could use a vice or similar clamp to place a small restriction in the tubing to loops I feel are getting too much flow. Either way if one pump fails all loops would continue to get an aproximately equivalent percentage of the flow they normally got, just all loops would get less flow than normal.

 

[Arcygenical]

Your theory is fairly sound, but it begs the question... Why? Why spend so much money on redundancy when it will barely help your cause. Two pumps, I could see that... But all the parallel loops are simply a bad idea, waiting to happen.

Most of all, you won't really GAIN from this at all. Your temperatures won't be much different in parallel (unless identiacal rads are in parallel).

The ONLY thing I would suggest if you're bent on Parallel is:

Res to Y fitting. Each branch of the Wye goes to an identical pump then an identical radiator, which recombines using a Wye into a main loop for CPU, NB, GPU. A single pump is enough for the loop you described.

 

[mezz]

I am not an expert, but the most logical thing to me would be to have everything be linear except for the two pumps. Just put a Y to split before the pumps, and another after to combine them. You can still have multiple radiators, but with one after the other, along with everything else on 1/2 ID tubing. Having things on separate paths with two pumps like you envisioned would have better flow, but it really complicates things and creates mysterious flow problems when some components are more restrictive than others (which is always).

 

[jackofalltrades]

I believe the best thing to do to eliminate damage due to pump failure is to put a flow meter on it and if flow is lost it would start an immediate shutdown. As far as a flow related issue just use one pump rated to the desired displacment and then go to 2 separate rads to keep temps under control then put the different rads back together into one reservoir. Like everyone said keep it simple.