Cross-connecting cooling loops

[mad mikee]

Presently (well at least b4 I took the machine apart for the latest round of tinkering) I have 2 cooling loops:

1. CPU block, pump, rad back to CPU (simplifing for this discussion)

2. Same as above but replace CP w/ NB and GPU

In future tinkerings I will be using (2) 2-342 single-pass rads and 2 equal pumps (mcp600 rev2) in the 'loops'. These will be in a separate enclosure (for now) and have 1/2" id tubing everywhere.
4-6 panaflow 120mm M series will be pulling air through the rads. These will be set at 7 volts for lots of QUIET airflow.

I am thinking of interconnecting the 2 loops to create the following flow:

pump1 -> rad1 -> CPU(cascade) -> pump2 -> rad2 -> GPU(MCW-5000) + NB(Z-chip) -> pump1

My reasoning:

A. This will allow me to make maximum cooling use of both rads.

B. Give me some redunancy in case 1 pump fails, there should (hopefully ) be more than enough time to repair / replace the defective pump b4 'castastrophic failure'.

Does anyone see anything wrong with this arrangement?

Your thoughts please :shrug: :shrug:

Thanks MikeE

[Since87]

Originally posted by mad mikee
I am thinking of interconnecting the 2 loops to create the following flow:

pump1 -> rad1 -> CPU(cascade) -> pump2 -> rad2 -> GPU(MCW-5000) + NB(Z-chip) -> pump1

My reasoning:

A. This will allow me to make maximum cooling use of both rads.

B. Give me some redunancy in case 1 pump fails, there should (hopefully ) be more than enough time to repair / replace the defective pump b4 'castastrophic failure'.

I'd suggest just connecting things up in the order that gives the cleanest plumbing layout. The MCP600's consume so little power, the added heat is hardly worth considering.

Your scheme does look good in terms of optimizing heat transfer, but I doubt you can get as clean a layout as you might otherwise. The complexity of the routing may offset the value of the ordering of components.

[mad mikee]

I will have the pumps, rads and fans in an enclosure sitting on top of my 'concoction' (search hint), in order to test some ideas and give me some 'elbow room' in my case whilst I plan out how to build mikeeputer rev3.0 (new case design from ground up to optimize case ergonomics, minimize heat and annoyances, and give me SPACE, and storage, etc). It will look like a piece of furniture to match my maple computer desk, allow me to stick the scanner on top and store other stuff) Very vague resemblance to a rackmount set up. (You would never guess that I don't do lan parties heh )

There will be one inlet and outlet for each loop running from this assembly (a polyethelene file storage case from office depot (no smirking, this is a prototype (as I look for the smirking smilie ) )

I want to get everything but the blocks out of there for now and see how much (or little as the case may be) fannage, etc., I can get away with. Combining loops will give extra capacity for dualies or 64 bit when it is affordable / mature enough for me.

The tubing run from this assembly to the blocks will be 2-2.5 feet at most for each loop. (Imagine 4 1/2" ID tubes running between the case and this assembly ).

The switchover between loops would happen at the blocks themselves, so no big new routing differences / problems

[mad mikee]

Well in testing at least.....

Background / overview:

No I don't have the puter up, I am however testing the cooling subassembly (16"L x13"H x 12"W) for leaks and will put up some pics after I get somewhere to connect the CF card reader, etc.

Slideout motherboard tray will be 1" above cooling subsystem (both removable - cooling subsystem as a unit minus the blocks.)
Most of the filling time is spent getting water into the system (about 30 minutes for 1.75 quarts). However air is gone almost immediately so I can live w/ that.

Inside the cooling subassembly: 2 x 2-342 HC, each fed by an MCP600, gravity fed from res immediately above each pump, pump feeds HC which will feed blocks:

(MCP600(1)->HC(1)->Cascade->MCP600(2)->HC(2)->(GPU/MCW5000 + NB/Z-Block)->MCP600(1))

Air comes in though the bottom (furnace filter in front of horizontal HCs) and exhausted out back by 4x 120mm pana. M1As. Single fill at top (1/4" ball valve, slow fill mentioned above), and single drain at bottom (same thing - connected to both HCs, open top and bottom - FAST easy drain).

Tubing runs should be less than a meter for each (sub) loop.

I like the redundancy aspect, thus the cross-connected loops.

I am trying to minimize the impact of the GPU/NB on overall flow rate/velocity to get the most out of the MCP600s into the cascade.

Some new questions:

1. For the GPU + Zblock, would I be better off running 2 x 3/8 ID in parallel or 1/2" ID in series. I have 2 of those large-bore Ys (2x3/8" -> 1/2") for parallel.

Have option of 2 |_| in parallel or

... /\
|_| |_| (series) (|_| = water making 180 degree turn in block).

2. (for BillA please or whoever? ) Since the pumps are primarily inductive loads (I think?), am I better off putting them both together on their own wire from the PSU or on 2 separate wires w/ something else on with them or ????
(new PSU is PC Pwr &Cooling 510 dlx, so sufficient power should NOT be an issue)

So? :shrug: :shrug: Suggestions /criticisms please?

[mad mikee]

Can you guys give answers to questions in post above please

Thanks MikeE:shrug: