Cooling What order should I watercool my system?

I am asking this question on behalf of a friends who wants to know which order it is best to water cool his cpu and northbridge.

Would you have the water cooling the northbridge first as it doesn't get that warm, and then cool the cpu, OR would you cool the cpu and then cool the northbridge with warm water?

Your answers on a postcard please...

 

I would tend to not bother with the northbridge for water cooling.



Linky: http://www.kustompcs.co.uk/product.php?view=2811

I'm going for one of these new Zalman chipset coolers and just water cool the CPU.

 

I've used a zalman passive sink on my n/b for the past 6 mths and gotta say that it's performance has been limited due to the low airflow inside my case.

I would say if your m8, like me, has very low airflow inside his case (as is common with watercooled rigs) then watercool the northridge after the CPU - cool the hottest first

I would reccomend splitting the flow though as most northbridge blocks will be very restrictive compared to a CPU block so cooling in series would compromise the CPU cooling.

just a few thoughts for ya

 

I originally went cpu => nb => gpu when I did mine, then I thought that rather than try to push the warm liquid downwards, it might be better to have it flow straight upwards throughout the system to the rad at the top.
Consequently, I'm now running gpu => nb => cpu and although my cpu and nb temps are pretty much the same as they were, my gpu is definitely cooler in normal use.

Neither the cpu or the gpu seem to be affecting the nb temp - it stays pretty much the same as my water temp (26degC, currently, while they are at 42 and 31 respectively)

 

That's an interesting perspective... Do you put the rad first or last in this line-up?

 

Nexxo - Last.

It goes Res (lower front) => pump (bottom) => gpu => nb => cpu => rad (top) then back down to the res (as it's cooled).

The way I see it, I've cut out any chance of thermal resistance to the flow.



Seems to work.

 

Hmm, why has it all gone quiet?

Does anyone else have a comment on this setup, or has everyone gone off to try it...

Not meaning to challenge existing ideas, I just tend to try and go with the slightly different approach from the mainstream if I think it might work.

 

I guess the only thing to worry about is the heat put off by the pump (since the water is not going through a rad first.. then cpu, like a traditional setup). But maybe that point is moot, since the pump is not as powerful as others (maybe you need to worry about a pump heating up the water when you go to..lets say a 1250). Whats your experiences with temperature in that setup Pug? Have you tried both ways?(i see you posted the temps for the new setup but not the old) And wheres my stuff!!

edit- doh! ok you said your temps are pretty much the same, so i guess it works well for you

 

i'm thinking of manifolds.

wad do u guys think?

 

I noticed you didn't mention cooling your GFX chip. I'd cool that before I even considered cooling the NB chip, much hotter.

But if that is what you want to do then by all means CPU is first, it is the hottest and needs it much more. And the NB will get along just fine with the water from the CPU.

If you want to avoid (you should) putting all your water through the NB by running is series then by all means split the flow after the CPU and cool that GFX as well. I'd suggest making a custom unbalanced Y, a small tight lint to the NB (1/4-3/8") and a larger line to the GFX of 1/2".

 

pump--->cpu--->gfx--->radiator---->res
id only cool the gfx if i was gonna oc*

i never wc the nb cause its not worth 40 bucks to get a few more mhz *like 1 or 2* outta the fsb to me....nice lil passive system with a lil air over it works just fine.....u need some air flow in ur case to keep temps down for the rest of ur stuff....

 

Hey Pug? How long did it take you to bleed the air out of that rig with the rad at the top ???

 

Not long at all - I tip the rig back on its end to fill the res and, as the rad is detachable, I lay it down so it's at the bottom while I fill, then turn it thru 180 degrees to bleed it.

[Edit] BTW, Edd, thx for reminding me about the temps, the GPU temp is actually incorrect -
I took it apart to take some better pics and noticed my temp probe had come adrift...
GPU temp is actually 38 deg.

 

that one fitting at the top just sitting there or is it going into that ps?

 

Just sitting there. (PSU is still aircooled)

The hose runs down the back and then along the bottom to the res at the front.
You can see it a bit better in this pic

 

Since I am planning to wc my rig as well, I was just wondering if anyone has put a splitter coming out of the pumpline/radiator line so both the CPU and GPU get its own flow of cool water... that way, you aren't sending heated water to the next block in your line..

Also, I only plan on cooling my CPU and GPU for now.. If not using the splitter, I might go GPU first (9700 pro) since it idles at around 110F and the P4 2.4ghz idles at about 100F. Any thoughts on how strong of a pump I would need to split the flow?

I currently have an aquarium pump rated at about 220GPH @ 3' and approx 190GPH @ 4'. My case a Lian Li PC82COM. Radiator will be mounted externally, probably sitting on top of case or on the right side in back of mobo. Since my pump is rated ~190GPH @ 4', does that mean 4' of tubing would allow for a max of 190GPH in ideal conditions?

I'm guessing I might exceed the 4' of tubing so would splitting my flow be a bad solution?

 

LCM - to address your Qs in reverse order, the 4' refers to the head of the pump, in other words, it maintains that flow rate at 4' above the pump.
Taller cases (unlikely) and bong coolers may require a more powerful pump to sustain that flow rate, as the head will exceed it.

As regards splitting the flow, you have to bear in mind that water will flow (like electricity) through the path of least resistance.
So, unless you run a separate pump on each line, you run the risk of stagnant flow or dead water in one of your blocks, while the other receives the full flow (or most of it, at least).

Very hard to get right any other way, hence most people run in series.

 

Sweeeet!! I was thinking the pump output referred to height ABOVE the pump. If I can keep my system height between 0-1 feet, my pump will crank out 400-450gph!

In my current situation with the 9700pro idling between 104-110F and my CPU running approx 10F cooler, would it be wiser to cool the GPU first or only if I plan on OCing?

Most likely I will OC my CPU w/the watercooling setup. ATM, it's running at the default 2.4G with stock Intel cooler but I've had it pumped up to 2.88G with a ThermalTake Dragon orb. With the P4 OC'd, temps are about 10F higher but then again, the Dragon orb fan was run at 5000/7000RPM, the Intel HSF a mere 2500RPM.

 

Not sure how a Radeon 9700 would affect temps (without having one here to try), so all I can say is experiment and let us know.

Regards the manifold approach to split flow - I spoke to my (new) fittings supplier today on this,
Apparently there are load balancing manifolds available on the market which will guarantee flow to both blocks but the price is higher than most people would prefer to spend on their entire cooling system!

 

Whoa there, l33t_c0mput6r_m0dz:

Flow is determined by many other factors besides the amount of head (ft. of water above pump level). There's tubing diameter, block resistance, rad resistance, number of sharp bends/angles in the tubing, whether or not you include a flow sensor/meter...

This leads us to the age-old question... Does Size Matter?

At some point the pump has to push so hard that it converts its energy more into heat than flow. This is why big pumps are not necessarily a good thing. Experience of various modders has suggested that an increase of flow beyond, say, 3-5 GPM (11-19 l/m) does not lead to a significantly increased cooling performance in most systems, and neither does switching an Eheim 1048 for a 1250, for example... You're better off designing a good circuit, with a short path (no tubes goin' up'n down all over the place) few bends, and fewest blocks possible (you could cool the NB passively, with the aforementioned Zalman sink). Just forget about HD and PSU cooling. I agree it looks very l33t, and I have been tempted, but it will offer you more complications for absolutely no return in performance. Go for a Rad and block with lowest resistance (e.g. D-Tek Customs' Coolercore, and Sprir@l respectively, although Paul Vodrazka's home made blocks are apparently even better :link:).

So as usual it's not the size that counts, it's what you do with it!


By the way, forget parallel setups. Too many complications (you have to keep flow identical in parallel paths, and from the above you can guess that is actually quite tricky), and there is no real gain in performance.