Pro/Forums - View Single Post

airspirit · 11-11-2002, 02:36 PM

This may sound stupid (it does to me), but I can't seem to shake the idea.

Background: I'm looking to build a parallel rack style system on a centralized cooling system. Everything will be funnelled through one radiator (23x12x2.5"), and one pump (750-900 GPH, decent head). The coolant exhaust from each case will drain into one large (5') section of 4" PVC where it will settle down to the very bottom where I'll have a funnel attachment that will go to the pump and res.

Idea:

Originally, I thought of pulling ALL water through the rad -> pump -> splits to parallel computer loops -> super-res -> rad.

What if I was to take the rad out of the loop? Go pump -> comps -> res -> pump, with a PARALLEL loop from the rad that draws the heated water from the res, pumps it with a seperate 350 GPH pump through the res, and exhausts it back into the res at a lower point. It would allow much higher GPH through the CPU blocks, without the restriction from a massive 1/2" radiator, and would slow the GPH to the rad allowing it to run more efficiently.

I have a feeling I'd nuke my CPUs if I did this (not all the water would cool). But then, the other half of me says that with a res that will contain over 4 gallons (over 12L) of coolant, not to mention the other half gallon circulating through the system, that the temps would equalize in that massive res.

What do you guys think? Do you think there would be a performance gain? On one hand you have higher GPH through the blocks and lower GPH through the rad, but not all the water (by a long shot) will pass through the rad each time through the system, and on the other you have all the water going through the rad, but a loss in flow rate for the CPU blocks (I'd guess up to 1 GPM each) and a radiator that isn't working as efficiently. I only have one shot at this since I will NOT cut my box, patch, cut somewhere else, patch, and then do the same on the res (which will, even in black PVC, cost upwards of $50).

Ben? Anyone? This is right up your alley. Lets bounce some theory around.

11-11-2002, 02:36 PM	#1
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	Need knowledgable advice This may sound stupid (it does to me), but I can't seem to shake the idea. Background: I'm looking to build a parallel rack style system on a centralized cooling system. Everything will be funnelled through one radiator (23x12x2.5"), and one pump (750-900 GPH, decent head). The coolant exhaust from each case will drain into one large (5') section of 4" PVC where it will settle down to the very bottom where I'll have a funnel attachment that will go to the pump and res. Idea: Originally, I thought of pulling ALL water through the rad -> pump -> splits to parallel computer loops -> super-res -> rad. What if I was to take the rad out of the loop? Go pump -> comps -> res -> pump, with a PARALLEL loop from the rad that draws the heated water from the res, pumps it with a seperate 350 GPH pump through the res, and exhausts it back into the res at a lower point. It would allow much higher GPH through the CPU blocks, without the restriction from a massive 1/2" radiator, and would slow the GPH to the rad allowing it to run more efficiently. I have a feeling I'd nuke my CPUs if I did this (not all the water would cool). But then, the other half of me says that with a res that will contain over 4 gallons (over 12L) of coolant, not to mention the other half gallon circulating through the system, that the temps would equalize in that massive res. What do you guys think? Do you think there would be a performance gain? On one hand you have higher GPH through the blocks and lower GPH through the rad, but not all the water (by a long shot) will pass through the rad each time through the system, and on the other you have all the water going through the rad, but a loss in flow rate for the CPU blocks (I'd guess up to 1 GPM each) and a radiator that isn't working as efficiently. I only have one shot at this since I will NOT cut my box, patch, cut somewhere else, patch, and then do the same on the res (which will, even in black PVC, cost upwards of $50). Ben? Anyone? This is right up your alley. Lets bounce some theory around. __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied