Watercooling advice wanted

Druid

Druid

Druid

Associate
Joined
26 Feb 2010
Posts
10
I currently have a self-built watercooling system comprising of the following components :-

Laing D5 Vario Pump
EK Supreme Universal CPU Waterblock : Plexi
1/2 inch Materkleer tubing
Swiftech 2 x 120mm Radiator
XSPC single drive resevoir - clear
Feser one blue Cooling fluid

I shallnt bore you with current setup but what I want to do is expand this system on new PC build and want to know if it is feasable or pushing the limit for the pump and cooling.

New Build will include:-

NEW Gigabyte Gigabyte X58A-UD7 motherboard
Core i7 - 920 D0
ATI 5970 GPU with either EK FC5970 or Swiftech Epsilon HD5970 waterblock
--------

So my current waterblock and system will work with the new CPU but I was wondering if I could also use it to cool

1) The gigabyte chipset through the included chipset cooling. I will need some 1/2 " to 3/8" reducers and a small amount of 3/8" pipe.

and

2) The ATI 5970 using probably 1/2" pipe to the inlet and outlet barbs on the waterblock. However maybe 3/8" to a full cover waterblock will suffice?

I believe I would need 4 Y Fittings to enable me to; slave 2 off the main line prior the CPU and 2 more to slave back into the main line prior to the radiator.
------

My main concerns are
a) The pump will have demand problems or fail much sooner (it is capable of 1200 ltr/hr though - an excellent pump I think).

b) It just wont be able to cool all of the components adequately (although it cools current Core2Duo E8700 , OC'd to 4.1Ghz easily).

I probably havent thought of everything so any advice and solutions are welcome. I dont however want to run a second cooling solution alongside my current one although I do realise that would provide the best results when overclocking.

Finally, I shallnt be overclocking the ATI5970 (well not just byet anyway) but I will OC the CoreI7-920 to at least 4Ghz.

Thanks in advance.
 
Defenitatly will need to upgrade the radiator, the pump should be ok

In my loop i have 3 motherboard blocks (chipset and mosfets), i7 920 and a gtx280 and thats all cooled by a thermochill PA120.4

And the standard waterblock on the gigabyte board isn't very good, how-ever (at least for the EX58 extreme) a company called MIPS makes a replacement waterblock for the standard heatpipe arrangement.
 
Don't do Y splittings if you're asking for general advice. Parallel loops are very difficult to get right, they generally lead to one side of the loop overheating and/or filling up with air. Much better to leave it all in series, or go dual loop if need be.

Your D5 is the best pump commonly in use on these boards. No worries there.

Mips do indeed sell a waterblock that replaces the stock gigabyte one, lets you use your own barbs and I'm entirely certain will be far superior to the default option. Or EK do a full cover set which costs more and probably offers no better performance.

It's going to run hot, probably around air temperatures unless you've got good fans on that radiator.

Take a look inside your cpu block when you take the loop apart, see if it's filled with feser dye. Good odds that it will be. Take the steel plate out while you're at it.

Welcome to the forums :)
 
As Jon suggests, strip your block down and thoroughly flush the radiator out. Feser is terrible stuff and should be avoided. It suffers from dye seperation where it clogs up blocks, rads etc. Distilled/De-ionised water with an additive such as Zerex is best.
 
pastymuncher said:
As Jon suggests, strip your block down and thoroughly flush the radiator out. Feser is terrible stuff and should be avoided. It suffers from dye seperation where it clogs up blocks, rads etc. Distilled/De-ionised water with an additive such as Zerex is best.
Click to expand...

Thanks for the advice. I totally agree with you on the Feser. I have a clear front drive resevoir and I can already see the dye seperating there so a complete flush is definately in order.

The mobo I mentioned at the begining is a brand new release from Gigabyte and I think it would be overkill to remove their current cooling solution and install a new one.
The link to their mobo is here http://www.gigabyte.com.tw/Products....aspx?ProductID=3251&ProductName=GA-X58A-UD7

The problem with running the whole system in series is that I will restrict the flow going through the mobo chipset as it uses 3/8" whereas the rest of the circuit is 1/2".

Maybe I will run the CPU and GPU in series and try a Y splitter to the chipset. I dont think it will matter too much if the flow through the chipset aint too brilliant as I can always put a fan on it also. The main concern then as you have touched on is that it will affect the flow on the main circuit too much.

I guess the best way is to try the main circuit in series, check temps then run the Y splitters in and see what the effect is.
 
Druid said:
The problem with running the whole system in series is that I will restrict the flow going through the mobo chipset as it uses 3/8" whereas the rest of the circuit is 1/2".
Click to expand...

It doesn't actually work like that. If everything is in series then every part of the loop gets exactly the same flow rate, regardless of tube size, block shape, whatever.

When you introduce Y splitters thats when different parts of a loop will definately get different flow rates. Flow will prefer to go through the path of least resistance and so is affected not only by tube size but also the number of blocks/rads etc in that sub-loop. You could quite easily get lower flow in your notional 1/2" loop if you bung all your cpu/cpu etc blocks into that sub-loop and only have a single chipset block in the other, regardless if the smaller loop uses 3/8" tube. Only if the two sub-loops are indentical will the flow be exactly halved in each. But as you say, if you are willing to test it out for yourself then you will be able to find what works best for you, which is always good.
 
Last edited:
Bubo,

thanks for the advice. Yep, trying to think back to my old Physics days where Pressure x Volume is a constant (well almost) and P1V1 = P2V2 etc.

Larger pipes are better because of less flow resistance (i.e. friction loss) and I believe there will be a slight loss of performance if I connect 1/2" to 3/8" in series piping as a result. However, I agree with your "path of least resistance" comment and I guess that pushing the flow through a CPU block and a GPU block loop will meet with more resistance than the chipset loop.

So I guess series loop makes more sense.

As for the triple rad, I will bow to everyones superior knowledge. However I will try it with the double rad first (its mounted sideways on the back of my desk about 2ft away from the PC for cooler airflow) and see what kind of temps I get before blowing more cash on a triple.

Just got to wait for some memory to arrive and then I shall put the rig together and let you know how I got on.

Just a couple of last questions,
1)what is the best additive for de-ionised water for keeping system running cleanly? Zerex was mentioned but I dont think its available here in the UAE.
2)what is the best dye that wont seperate from the fluid or clog up my system? or best to just leave it clear? 3) any preference on the GPU waterblock, EK or swiftech ? They both look pretty good but have not been able to find any comparison reviews.
 
Last edited:
P1V1=P2V2 is more to do with gases but it is in the same general ballpark. It is much easier to think of a water loop as an electrical circuit. Pressure is equivalent to voltage, flow is equivalent to current. The pump is a battery and each block and length of tube is a resistor of varying magnitude. If you did basic circuit theory in physics then it may be easier to think of it in those terms as to why flow/current would vary in different parts of a loop.

Yes you get slightly higher flow losses if you use 3/8" instead od 1/2" but it really is miniscule compared to the pressure drop across your blocks. To keep things simple I would go with all 3/8" tube if you are planning on using your default mobo chipset block. Get a replacement top for the D5 pump if you do this however, as it has plastic molded 1/2" barbs by default

As long as everything in your loop is copper/brass/nickel plated then all you need to add to DI water is a biocide, e.g. PT Nuke or any small piece of silver in your res, silver kill coils are made for this purpopse.

Edit out that shop link or you will incur the wrath of the ocuk overlords.
 
Last edited:
Great advice. Thanks.

Yep Boyles law is for gases which dont suffer from frictional loss like fluids do. Can't remember fluid formula (its been a long while!).

I guess to run a 2 loop system I would have to be able to check the pressure differential at the output end of each loop and have an adjustable pressure valve on the inlet of one loop to make it the same equivalent pressure as the other loop. Far too complicated methinks!

Didn't know I could get a 3/8" replacement top for D5. How about I just have 1/2" pipe then a 1/2"to 3/8" reducer to 3/8"pipe? or

Run the loop in series with 1/2" to CPU then GPU with output from GPU with a 3/8" to 3/8" pipe to chipset and 3/8"to Rad. Or is mixing different size tubing a bad idea?

What will be the best series order? PUMP -> CPU -> GPU -> Chipset -> RAD -> RES -> PUMP?

Would you add a dye or just leave it clear? I am not a UV, sidepanel, aesthaetic freak but I do have a front panel resevoir ( could always stick a bezel on it though).

link deleted ..oops. Thanks

p.s. Sorry for all the questions. I should probably search the forums better but I usually end up going around in circles.
 
Last edited:
The D5 top will also give a small 8-12% increase in flow and reduced the noise a little.
EK-D5 x-top is what you are looking for.
But a 1/2 to 3/8 converter is a much cheaper option.

Res before pump is the only real rule with loop order.
shorter loops are also better, but this sometimes requires 45 or 90 bends that can start to push the price up.

Avoid dye - use colour tube
Or if you must see you res filled with colour - Audi G11 (blue) G12 (pink/red) G12+ (purple) 5% mix with water - so 50ml in a 1 litre loop.
If you ask nicely, and have your own small bottle, a Audi dealer will give you 50ml free.
 
Okay, copy all that thanks. I think I will stick with the 1/2 to 3/8 converter & clear fluid as the middle eastern audi dealer will probably charge me 1/2 a camel for some Audi G11. After spending all my booty on this rig all I can afford to give him is a chicken wing!

As for the order I thought maybe I should cool the components in the order of component importance as well as needs the greatest amount of cooling for OCíng i.e. CPU - GPU - chipset ?

edit: just read the review on the EK D5 ver 2 top - pretty impressive!
http://www.skinneelabs.com/ek-d5-xt...ht=ajaxSearch_highlight+ajaxSearch_highlight1
 
Last edited:
Q=vA where Q is the volumetric flowrate, v is fluid velocity and A is the pipe area.

As you can see the flowrate will remain constant regardless of area since you're not consuming or producing fluid, and the only thing that changes is the reynolds number which is a measure of turbulence; but going from 1/2" to 3/8" diameter the difference is so small it can be ignored.

Welcome to OCUK :)
 
to the op, i have the mips block on the ud7, it's pretty restrictive, but would be a million times better than the stock waterblock, which is aluminium (you want to keep well away from copper/nickel/brass)

hope that helps
 
twist3d0n3 did you compare the stock cooling temps with the MIPS before you fitted it?

A little confused about your comment. If the stock waterblock is aluminium then surely thats good as it is keeping away from copper/nickel/brass?

Speaking of which whats problem with copper or nickel? EK make lots of waterblocks with copper/nickel e.g.EK 5970 full cover block

Why is the stock waterblock on the NEW gigabyte UD7 so poor?

Are there any reviews/tests of the performance of the gigabyte stock cooling versus other products (e.g. MIPS) anywhere? I am a big fan of real world tests and supporting data.
 
Last edited:
What twist3d0n3 was trying to say is mixing metals within a loop isn't ideal.
With a little thought its fine - the german and american watercooling seen is awash with mixed metal loops.
As long as the different metals are not touching each other, and you use antifreeze/corrosion inhibitors in your coolant it's fine.

Copper is a better material for the transfer of heat - so all the best blocks/rads are copper.

nickel/brass are simular to copper so galvanic corrosion doesn't happen - but normal corrosion does (it's just very slow)
 
Druid said:
twist3d0n3 did you compare the stock cooling temps with the MIPS before you fitted it?

A little confused about your comment. If the stock waterblock is aluminium then surely thats good as it is keeping away from copper/nickel/brass?

Speaking of which whats problem with copper or nickel? EK make lots of waterblocks with copper/nickel e.g.EK 5970 full cover block

Why is the stock waterblock on the NEW gigabyte UD7 so poor?

Are there any reviews/tests of the performance of the gigabyte stock cooling versus other products (e.g. MIPS) anywhere? I am a big fan of real world tests and supporting data.
Click to expand...

as shadowscotland says, i'm merely pointing out that having similar metals is less of a headache, and personally i try to limit it to copper/nickel/brass.

unfortunately, i haven't tested to see the difference with and without the waterblock on the heatpipe array, mainly because i had all the gear come through at the same time and thought there wasn't much point setting it up then pulling it apart again just to fit a few extra bits. my temps aren't the best for IOH, but i believe this is down to the woeful TIM pre-applied to the mobo. when i've got a few spare days, i'm going to strip my rig down and get everything perfect.

the MIPS, albeit very good, is very restrictive though, i can tell you that much
 
ok. I'll stick with the stock gig cooling for now and see how it goes.

I'd be interested to see how you get on as we have the same motherboard.

I think I will go for the new top on the D5 and run 3/8" pipe with compression fittings so going to have to wait a wee while longer before I start the build.
 
You must log in or register to reply here.
Back
Top Bottom