I have a new system worth of parts coming in the mail including a swiftech H20-220 compact. I also will be running dual ATI 4850. The huge factor is i am using a Micro ATX case.

Now before everyone brings it up, I understand the restriction on space that I will be having but I was fully expecting that to be a problem....my quest is this. Does it matter what order the water cools the devices?

I was thinking I could save myself lots of space(in tubing) If I cooled the Graphics cards first in the cooling loop. I wanted some feedback from some people with more experience than my 0 experience.

So to sum it up....Do I save space and run the water by the Graphics Cards first, or do i work harder with my space limits and cool the CPU first.

Thanks in advance!

Which case are you using exactly?

GPU first for sure.

CPU first or you might as well put a fan on it.

4850's run fairly hot so that would probably be the best place to start.

In a closed loop system, which component is "first".


Loop order will have very little impact on your overall cooling. Use the order that uses the least amount of tubing and minimizae tight bends.

Tony has the best point...every point in the loop is before and after other components...hence 'loop'.

Typically:

CPU > NB (if applicable) > GPU(s) > radiator > reservoir (if applicable) > pump


And no, there isn't 'super heated water' from one component to another. If you consider 85C of your GPU's stock cooler, you will run ~45-50C (tops) with a good water loop, even with CF or SLI...and even after your CPU and NB. The place to pay the most attention is your radiator and fans. A 3x120mm radiator with good CFM fans is minimum for CPU/NB+GPU. If you are going to run CPU/NB/GPU/GPU you will need more.

A basic rule of thumb is at MINIMUM a 1x120mm radiator space per component. 1.5 per component would be better. So with your Swiftech kit, your 2x120 radiator would probably be maxxed out with a CPU and GPU on it...you could probably squeeze a NB block in there...but adding a 2nd GPU would warrant more radiator heat exchange area.

Good drawing!!

I should have added (and you captured in your drawing) that the only order variable that should be considered a "must" is to have the res be higher than and feeding the pump.

Other than that, in a typical loop - a pump pushing fluid at say 1.5GPM (about the optimum rate) - the fluid is moving through the loop at 4-5 feet per second. With the fluid moving that fast, the fluid temp is going to tend to equlaize across the entire loop no matter what the order. In some of Martin's Liquid Lab Test, the temp difference between fluid entering the rad and exiting the rad (typically said to be the "coolest" water) was less than .5C.

Yes, res or fillport/T-line above and before the pump. Air is much less your enemy with this config since you will fill your system using the one you choose. Thanks, quick Visio drawing I threw together for someone last week to illustrate flow in a loop.

I agree...typically there is this misconception that the CPU 'boils' water because it gets so hot, and then can 'melt' your GPU when it hits the block. Water has completely different cooling properties than air and has a gradual, stablized cooling effect when correctly set up.

The only way I can see water in a loop getting that hot would be if your pump failed, and even the stagnant water would absorb a lot of heat in the tubing and block before shutdown took place. Example: I had one of those older MCP350 pumps that sometimes didn't turn on (the newer 355's fixed this) and would cause the CPU and GPU temps to get pretty hot from the water not circulating after about 10 mins. Once I tapped on the pump, it started up, and the water temp immediately dropped and would never rise above load temps again. Good thing they updated this pump model...its the reason I went to the MCP655.

Could I add this another idea ie:

Res->Pump->Rad->CPU->GPU(s)->Res

This would mean all the heat from the combined CPU, GPU and Pump is dissipated in the Rad.

This only works if your fairly confident that your water temperature out from the GPU(s) is less than 60C, Laing's max operating temperature...

Just suggesting as the res is effectively a heat transfer device and then the pump itself adds a slight temp increase.

Either way, the rad is both before and after the components being cooled, so it is effective in almost any setup. The only real important piece to the puzzle is that your reservior or T-line/fill port opening is higher than your pump. Also, better fans (higher CFM and pressure) on your rad make a difference.

Never put a rad as the first thing after a pump. The best flow characteristics for your coolant is going to be coming right out of the pump so you want the most important component to have that benefit. Rads are the single greatest hindrance to flow of any component in a cooling loop (because of the multiple bends and turns in the pathway).

Rads - at least the most popular ones today - have very little resistance when compared to most water blocks. They will offer the same resistance no matter where they are situated in the loop. If you have a weak pump (flow rate less than 1GPM) then order can make a difference - but you really need a better pump. If flow rate is gretater than 1 GPM (optimum being around 1.5GPM) then order is not going to make any appreciable difference. You can make better performance gains by looking at your rad (number of "sections" not necessarily brand) and fans, restrictions (tight bends and pinches) and block seating, than by worrying about loop order.