New watercooling setup for 2cpu rig, open for suggestions/critics

[Cronos]

Hi.
I've been lurking around for some time and learned alot about water cooling.
Great source of information.

I am building new watercooling setup for dual socket machine, specs are:
--2x Xeon s771 Clovertown (Quad) E5320@1860MHz. Currently only mildly overclocked to 2.7GHz. But i hope, with the recent progress with SPD programming, decent overclocking will be possible, to 3.3-3.5GHz.
--MB Tyan Tempest 5000XT
--8x1GB FB-DIMM 667
--Stacker STC-T01
--PSU - single Zalman ZM600-HP. I am planning to install second one, dedicated for CPUs.
--Video is not important for now (7300GT), but later i will get top model from either NVidia or AMD/ATI, depending on which one will be better. Probably 2, for SLI/CF

Here are the main WC components I am planning to get:
--Radiator: 2x PA120.3 (already ordered). Will run in parallel, mounted over the case top with custom-made mounting solution.
--CPU Waterblock: 2x D-Tek FuZion
--Chipset Waterblock - Silverprop Cyclone NexusSX. I want NB water cooling
for higher FSB overclock. Also, 5000X chipset has quite a high thermal output,
32W TDP nominal, 50W estimated at 500MHz, actual load may be higher (as usual, Intel TDP is not the maximal thermal load )
--Pump: 2x Alphacool Laing DDC-Pump 12V Ultra with Alphacool top. Both will be mounted in the 4-in-3 module in Stacker. Question: will they both fit?
They will be ran in the same loop mainly for redundancy, but also for increased performance. For the later, pumps will be separated by some water resistant WC components, namely CPU and/or chipset blocks in the loop.
--Tubing: 1/2" ID (3/4" OD) Tygon Silver Antimicrobial Tubing from Petra.
Want give it a try, i like the idea and look of it.
--No reservoir, just T-line with DD fillport.
--I also want to include Pentek filter in the line, in parallel with Shutoff Ball Valve. For the first several days the valve will be closed so all water flow will go through filter. Then i will open the valve to reduce resistance. Part of the flow will still be going through filter, providing permanent filtering at very little overhead.

That is all, concerning WC setup for now. Later i may be adding custom-made waterblock for memory, as FB-DIMMs runs extremely hot. And, of course, future video cards will be watercooled as well.

[IanY]

Wow. What a beautiful machine!! Congratulations!

Since you like FuZions, go with it. I would prefer Apogee GTs, but I think the lower restrictions of FuZions would serve you well. I would prefer MCW30s on the NB though, but its up to you. I would upgrade to DDC+ pumps instead and consider Petra tops. I personally use the Alphacool dual DDC top and it works well too, but you really should have 18W pumps instead of 10W pumps. You may find that a small reservoir may save you some PITA, especially with a twin cpu setup, but its up to you. Running the radiators in parallel is not the best of ideas, neither is parallel anything. This forum will scream "series" to you

[Movieman]

Cronos: Nice to see you over here!

[Cronos]

Wow. What a beautiful machine!! Congratulations!

Since you like FuZions, go with it. I would prefer Apogee GTs, but I think the lower restrictions of FuZions would serve you well. I would prefer MCW30s on the NB though, but its up to you. I would upgrade to DDC+ pumps instead and consider Petra tops. I personally use the Alphacool dual DDC top and it works well too, but you really should have 18W pumps instead of 10W pumps. You may find that a small reservoir may save you some PITA, especially with a twin cpu setup, but its up to you. Running the radiators in parallel is not the best of ideas, neither is parallel anything. This forum will scream "series" to you

---I was considering MCW30 first, but it is actually low-performance block for low-power devices. It has completely flat base inside. I think for 50W+ load NexusSX with higher surface area base will serve better. Besides, it looks like having better mounting solution. Overall, it looks better
--- I believe "Ultra" is 18W and "Pro" is 10W? I definitely want 18W version.
Also, separating pumps in the loops will result in better performance compared with running them head by head.
--- I chose Alphacool top, though more expensive, for the pre-installed easy mounting for 3.5 bay. As i said, i plan to mount both pumps in the same 4-in-3 Coolermaster module.
---Why running rads in parallel is a bad idea? Two rads in parallel have twice the cooling capacity and lower resistance, compared with single radiator.

[IanY]

Series is what you are after, not parallel.

Parallel implies that you are going to use a Y junction to split. Rad 1 goes to cpu 1 and rad2 goes to cpu 2, then they re-join with another Y junction. Each cpu does not really benefit from twice the cooling capacity.

Series implies rad 1 goes to rad 2, which goes to cp1 and then cpu2. Alternatively, its rad 1 to cpu 1 and then to rad 2 and then cpu 2. In any event, there's no Y junction used and the tubing runs uninterruptedly.

Well, that's just semantics.

Yes, separating two pumps provides better head, but I suggest Petra tops as opposed to the single Alphacool top because its better performing. Like I said, I use dual head Alphacool tops only because I lack space and I cannot squeeze in independent pumps, and I'm now pushing six pumps (and plan more). Yes, buy the 18 watt version, which is called by various names, including "Ultra" and "DDC+."

I would take exception to your comments about the MCW30 but its your choice Its been doing well on my old Tyan Thunder NF2200Pro for quite a while now, and that's also a hot running chipset.

[Psychotic Emu]

Personally I would say that you have very much the right idea Cronos. I'm not saying I have more experience in watercooling than you Iany, because I certainly don't But from my little experience I think Cronos is right and I'm not going to lie and go against my feelings/hide them.
I like the alphacool top for the mounting options and think that the performance difference is minimal.

Also would have to agree (although I could do with a little more time to think about this) that parallel is ideal for rads since the flowrate will drop, allowing greater cooling time (is this crap?) and impedence will actually be approx = to using 1 rad. It will be very important that they both have similar resistance tho.

Having said all that MCW30's are a simple design but for a loop with this many components I think that is ideal. Also I don't think chipsets are that needing of the best cooling.

My thoughts, hope I don't cause problems

[Aerou]

if you plan to watercool the chipset, take the most non-restrictive piece you can find, in this case, the MCW30 is the best

[Cronos]

I was actually considering splitting and re-joining rads outside the case, having only single inlet and single outlet going outside.
Having two rads in serial head-to-head will result in twice the pressure drop compared with 1 rad.
Having them in parallel should actually reduce the pressure drop, but there is additional pressure drop from two Y -junction. It is hard to quantitatively estimate the difference, but i feel parallel will have less drop.

As to performance, i have been doing some simple (though very speculative ) math which shows two rads in parallel should have (slightly) better performance compared with two in serial, at the same flow level in the loop.
Provided flow is higher for parallel, the winner should be clear.
But, only experiment will tell.


Here is the math: we may assume the temp. drop of the water going through rad is proportional to inlet water temp. and inversely proportional to water flow:
dT=a*T/F, a -unknown coefficient.
Total temp drop depends only on waterflow and total power inside the loop.
Lets assume T1 -water temp. for serial, T2 -for parallel, F -flow (equal for both)
Total drop for serial:
dT1=dT1' (first rad) + dT1" (second rad)
dT1'=a*T1/F
dT1" =a*(T1-dT1')/F=a*(T1-aT1/F)/F
dT1=dT1'+ dT1"= a*T1*(2-a/F)/F

Total drop for parallel:
dT2=a*T2/(F/2) (flow is F/2 in each rad).

We have equation:
dT1=dT2,
a*T1*(2-a/F)/F=2*a*T2/F,

->T2=T1(1-a/2F)

That is, T2 is lower than T1

[Cronos]

if you plan to watercool the chipset, take the most non-restrictive piece you can find, in this case, the MCW30 is the best

Thats what i heard many times b4. I would agree, but i want to have chipset temp. as low as possible. The server chipset 5000X is more complex and has more power output than desktop chipsets. In my system, the weak point is not CPUs, but chipset and memory. Actually, the memory is what holding me back now, but as soon as proper mod will be fully developed to set 5:4 divider and arbitrary memory timings (none of which can be controlled from BIOS on any s771 server board) the chipset may become the main bottleneck.

Also, i have seen many people complaining about MCW30 having weak mounting.
The NexusSX has more solid mounting clip, which is very important for me - the board has only two hooks around the chipset.

[Cronos]

You may find that a small reservoir may save you some PITA, especially with a twin cpu setup, but its up to you. R

Thank you. I had given this a second thought and reconsidering T-line in favor of reservoir. Unless there is something much better, i will take Swiftech MCRES-MICRO.
What metal fittings should i take for it?

[IanY]

MCRES is perfectly good. There are several other options, but that's just a reservoir, so I tend not to lose sleep over that lol

If you insist on metal fittings, I would use only the Swiftech metal fittings. They tend to be expensive but at least you have assurances that its a perfect fit.

[IanY]

Personally I would say that you have very much the right idea Cronos. I'm not saying I have more experience in watercooling than you Iany, because I certainly don't But from my little experience I think Cronos is right and I'm not going to lie and go against my feelings/hide them.
I like the alphacool top for the mounting options and think that the performance difference is minimal.

Also would have to agree (although I could do with a little more time to think about this) that parallel is ideal for rads since the flowrate will drop, allowing greater cooling time (is this crap?) and impedence will actually be approx = to using 1 rad. It will be very important that they both have similar resistance tho.

Having said all that MCW30's are a simple design but for a loop with this many components I think that is ideal. Also I don't think chipsets are that needing of the best cooling.

My thoughts, hope I don't cause problems

Psycotic Emu,

You should always openly speak your mind and openly state your position. You should never have to supress your feelings/thoughts, ever. Nobody here is any superior to anyone else. The discovery process is lifelong and we are all just learning.

Like I said, I have relied on Alphacool tops for some time now and I do not own a single Petra top. Feedback from this forum show that its the pump equivalent of sliced bread... a really good thing. Therefore, my recommendation is based on the usage experience by many of my friends here.

As for parallel rather than series, that thought is contrary to popular wisdom in this forum. I am *not* saying that parallel is wrong, just that the overwhelming portion of experts in this forum, including some industry leading experts, will vehemently state that parallel is not good and series is much better.

As for Chronos, he's entitled to his views. He is also probably correct, but I am not taking sides here. Just be prepared for a barrage of flames, and I want *nothing* to do with that.

As far as I am concerned, I would never run two cpus in one loop, so this whole hoopla about parallel vs. series is moot to me. You already know me as the crazed one that insists on multiple independent loops, but that'll require multiple pumps and possibly more reservoirs.

The Clovertown-based 5000X chipset runs extremely hot. Server chipsets tend to do so. I am sure there are multiple compelling products in the market, but the mCW30 has served (no pun intended) me well. Familarity breeds contempt, but I don't fix what ain't broke.

[Psychotic Emu]

I smell an engineer Cronos. Then again I smell quite a few engineers on these forums and I can't say thats a bad thing.
I can't disagree with your theory. I am of the same school of thought but with a clear head I have to say that often the proof is in the pudding and experience counts for an aweful lot.
You can only go so far with theory. Things can be very different in reality.

I must also say that I have relatively little invested in my PC and so am a little out of my depth in monetary terms. Suffice it to say, I would invest proportionally in cooling equipement, and ensure redundancy.

I would also like to congratulate you on an amazing looking system.

[QuietIce]

I also tend to believe in parallel rads and I've been scoffed at, too. But looking back at history even Cathar ran his parallel for awhile then decided his rads weren't restrictive enough to warrant it and went series but the performance was near the same. Search through the Aussie OC forums for that thread ...

[IanY]

I don't think we should be blindly emulating anyone. Nothing beats personal trial and error. With all due respect, whatever is good for Cathar, is good for Cathar. I think you need to find out what's good for you.

Like I said, I wouldn't set up two cpus in parallel or in series. I would go for independent loops. But that's just me.

[Aerou]

exactly IanY, had a lot of crashes with the HE SAID thing ...
there is a similar problem with the stickies here

[Cronos]

Update:
According to manufacturer (Pentek) data, its p-series spun polypropylene cartridges with 1micron rating for inline filters have very low pressure drop 0.6psi/5gpm. For more realistic 2gpm the pressure drop will be negligible. For example, D-Tek Fuzion is rated for 2.2Psi/2GPM.

If true, this means that: the Pentek or similar filters with 1micron or more spun polypropylene cartridges (and similar) can be used in water cooling loops on the constant basis, without any need for removing or bypassing. Good finding, this may greatly simplify loop structure and maintenance.

Updated:
I was told that polypropylene filter should be avoided as it tend to lose filaments, which can be disastrous for pump.

[Cronos]

This project is undergoing huge revision with strive for simplicity.

1) The pumps will be two DDC+ 18W with default top. If possible, i would like to get new revision if only for the sake to have the latest
2) The tubing will be Tygon Silver 3/8" ID - 1/2" OD with Coolsleeves.
3) FuZion will remain as the CPU WB of choice, with some 3/8" fittings.
4) DD Maze4 for chipset.
5) I have already ordered 2x PA120.3 with 1/2 fittings. I planned to use AquaXtreme G3/8 - 1/2" fittings for radiators, and now i will need some G3/8 -3/8" fittings.
I cant seems to find any metal fittings of this size. Are there any available?

5)No reservoir and no filter. Too much hassle. T-line it will be, strategically placed so all the loop can be filled without even running the pump. PA's drain port should be of great help here.

[Marci]

I cant seems to find any metal fittings of this size. Are there any available?

Not seen any anywhere meself...

Also,

..........?

[Cronos]

..........?

Fixed.

Can i just get plastic fittings like this one :
Black Nylon 3/8in. NPT thread to 3/8in. barb fitting ?
http://www.sidewindercomputers.com/blny3nptthto.html

[Marci]

Can i just get plastic fittings like this one :
Black Nylon 3/8in. NPT thread to 3/8in. barb fitting ?

Will require plenty of PTFE to seal... but otherwise should be ok. Would be preferable to order 2 3/8"BSPP to 3/8" Hose fittings from http://www.thermochill.com/acatalog/...cessories.html thus ensuring threads match and needing less PTFE tape to seal than an NPT threaded fitting.

[Cronos]

Will require plenty of PTFE to seal... but otherwise should be ok. Would be preferable to order 2 3/8"BSPP to 3/8" Hose fittings from http://www.thermochill.com/acatalog/...cessories.html thus ensuring threads match and needing less PTFE tape to seal than an NPT threaded fitting.

Looks great, really want this one, but how much will it cost me shipped to USA?
This would be double international shipping for me as i am assembling my parts
in USA and then get them shipped to my country in one package.

[Marci]

E-mail steve@thermochill.com, asking for a shipping quote for sending 2x 3/8"BSP to 3/8" Hose nickel fittings, provide him with your full destination address, he'll email you back with full price.

Place order thru site. Ignore shipping figures. Steve'll manual adjust it once downloaded to match the quoted price.

May take him a while to respond... 48hrs...