Dual CPU Loop?

Hi,

i'm making a dual cpu loop (2xXeon 5150 2,66ghz), that's my first water cooling so i have many doubts

That's the wc config:

Pump: Swiftech MCP655™ 12 VDC Pump
Radiator: Swiftech Radiator - MCR220 Quiet Power Series Dual 120 mm
Reservoir: Swiftech MCRES-MICRO™ Reservoir
Cpu WB: 2x Swiftech APOGEE 1U-Server (Xeon/AMD64)
VGA: Swiftech MCW60™ Graphics Card Liquid Cooler with MC14 BGA Ramsinks
Chipset: Swiftech MCW30

Click to expand...

I was thinking about the loop but i really dunno

Pump > Vga > Chipset > Y fitting > Cpu1 and 2 on 2 different tubes, to don't make the cpu temps different, correct me if i'm wrong > Y fitting > Radiator > Reservoir > Pump

That's a good loop or it's better to go before on 1 cpu and then on the other?
Also, the tubing will be ~2 meters long 'cause i have to fit it in a proprietary case that i can't change, the pump is good enough?

Also: I already have an AcceleroX2 from ArcticCooling on my x1900XT, is the water cooling going to be better than that or i can save some $ without losing a lot in performaces?

edit: you know a good european shop where i can buy the components listed before? the only one with all the things i need is american and it costs me an additional 70$ of spedition ...

Thanks in Advance, and sorry for my bad english

you should have the two cpus right after the pump so your running the coolest water to them.

i think

Pump>Rad>Cpu1>Cpu2>res

that should be good on a 120.2 as i am not familiar with xeon's temps

if you want vga on it too

a 120.3 rad is key

Pump>Rad>Cpu1>cpu2>Vga>Res

that pump is incrediably strong and should have no problems at all

with a petras top

You will have better overall system flowrate with the Y fitting as you aren't running into the same pressure drop twice. If you have half the flow going through each fluid resistance then you will have 4 times less flow resistance as you would with running them series. That is... negating the pressure drop associated with the Y's. Your flow rate through the radiator will be higher with them in parallel. However the flowrate through the blocks will be nearly halved.

I guess Cathar would be the best person to ask this question.

Thanks for the answers

@EmAn: a 120.2 rad is not enough for all the blocks?
I'll go for that loop than:
Pump > Rad > Cpu1 > Cpu2 > Chipset > Res

And about the Y fitting i'm not really sure, i've read on some forums about that and i have way more doubts :\ any1 here running one of that?

I would never run blocks in parallel with the possible exception of the RamCool from Swiftech (for video RAM and designed to run parallel with an MCW60 GPU block). Your MCP655 can handle a pair of Apogee's with ease along with the MCW60 - all are low restriction blocks. Now, a pair of Storms and two VGA blocks might push the limit but even then I'm not so sure.


For the rad, IMO you'll need a triple to handle the heat load if you intend to OC at all. If it's all going to run at stock speeds I'm sure a 2x rad and some decent fans will keep temps below factory maximums ...

Illidanz said:

Thanks for the answers

@EmAn: a 120.2 rad is not enough for all the blocks?
I'll go for that loop than:
Pump > Rad > Cpu1 > Cpu2 > Chipset > Res

Click to expand...

no a 120.2 is not enougth for that

your loop looks fine

Pump > Rad > Cpu1 > Cpu2 > Chipset > Res

Click to expand...

and the pump will be fine for that as they are low restricition blocks

I too would not run any CPU block in parallel. I would go for Y fitting.

Choosing parallel or series depends on how resistive your blocks are, how much velocity your pump will deliver and how resistive the rest of the system is. You need to find out if the reduced velocity going through the blocks is actually a problem for your system. Here's a nice practical idea how:

http://wc101.com/articles/vs

My quad core, on a Swiftech dual rad, raises the water temperature 4c above 24c ambient while the TAT core temperature is 55c. Presumably a second core will add another 4c for a water temperature of 32c, which is still well below 55c. The heat removing ability of the water will hardly change from one core to the next in series; a difference of only 4c between cores. I don't see you can gain much from a Y even if you can keep the velocity above 1.5 gpm in each segment.

I would go with a triple rad and 1/2" tubing.

i would personally just stay away from going paralell unless it for for ram cooling or something

LandShark said:

I too would not run any CPU block in parallel. I would go for Y fitting.

Click to expand...

Huh? How would you use a Y fitting and not run parallel?

orion456 said:

http://wc101.com/articles/vs

...Presumably a second core will add another 4c for a water temperature of 32c, which is still well below 55c. The heat removing ability of the water will hardly change from one core to the next in parallel; a difference of only 4c between cores.

Click to expand...

That article is bogus. It starts out saying that lower flow rates through a rad are better. Any chart on any WC rad will show that not to be the case. The "more time in the rad" thing is wrong. If, say, 25% of the water in a loop is in the rad at any one time, then no matter how fast or slow the water goes, 25% will be in the rad. If, for example, you double the flow rate, the water does spend 1/2 the time in the rad, but it does so twice as often.

And, no, the water will not heat up 4*c between CPUs. maybe 0.25*c at most. In any WC loop, the delta T between rad in and rad out will be less than 0.25*c. And that includes the heat dump from the pump.

"from one core to the next in parallel" Don't you mean series?

The second CPU may very well raise the coolant temperature by 4C, but if you've got a flow rate of 1GPM, the CPU would have to dissipate over a 1000 watts to create a 4C differential between the inlet and outlet.

billb said:

And, no, the water will not heat up 4*c between CPUs. maybe 0.25*c at most. In any WC loop, the delta T between rad in and rad out will be less than 0.25*c. And that includes the heat dump from the pump.

Click to expand...

You're right, the 4c is the excess heat built up over time and delta T is less than 0.25c. So the only advantge of going parallel is reduced flow resistance and if you use a low resistance block then that advantage is eliminated.

billb said:

Huh? How would you use a Y fitting and not run parallel?

Click to expand...

I think he meant to say he would never run in series. Otherwise it doesnt make sense

I wouldnt use a Y fitting, considering the possible difference in flow between the blocks could lead to poor temps on one CPU. Series would be a better choice.

Thanks for the answers guys

I'm going to build it in series then, i'll post some bench/temps when i'm done, in the next 2-3 days maybe