OK guys...you have convinsed me..!!!!

[[NH]Naughtyboy]

As the topic says....I have been convinced to sell my BIX:es and buy a HeatCore.

I have a question though.
I was planning to use 2xEheim 1048 for my WC...using 2xBIX.

How about using those 2 1048:s with 1 HC....is it to much...or can the HC take it..??

And also where would you suggest me install the pump/pums...??...befor or after the rad..or mabye and before and one after..??

WC:
PSU: S-320-12 for the cpu block and my regular TruePower 550w for gpu and rest of my system
Cpu: Swiftech MCW478 UHT 226w TEC
Gpu: Swiftech MCW 50 T 80w TEC
Rad: HeatCore 6 1/8" x 7 3/8" x 2"
Pump: Eheim

I have a P4 2.4b and my plan is to run it above 3ghz. If I´m not totally wrong it will produce aproximatly 92.3 watts of heat @ 3014mhz and 1.7vcore.

Ambient temp will bee about 20C´

[bigben2k]

Well, first, yes the heatercore can handle two pumps. They're typically tested to 25 psi, and that's the equivalent of 17.6 meters of water.

You'll want to put the pumps in series because:
a) they work better that way (thanks to Airspirit)
b) it'll give you more pressure (which is good!)
Unless you're running two seperate loops.

I used to say that the waterblock should follow the core, so that the rad can dissipate the pump heat, but I was wrong, and anyways, the pump heat is relatively insignificant.

What we all found out is that the pump benefits from an unrestricted inlet (thanks to BillA), so I'm inclined to say that the airtrap/res should come before the pump, but that wouldn't be an accurate description of the solution: the pump inlet will benefit from the largest possible diameter, so if it's possible/necessary, use a bigger tube for that particular connection.

If you're ready to go ahead with the two pelts, you might think about getting either the largest (and/or thickest) heatercore that you can fit: you're reaching a very high load of heat.

Fans for those heatercores will be critical: pick wisely.


Good luck!

[airspirit]

I'm torn on this issue. Because of the pelt, you're going to be producing much more than just 93W, you realize that, right?

The more I think about it, the more I think you should be good as you are. I am about to contradict my previous statements on pump arrangement, but this is a unique situation that could benefit from a unique component arrangement.

Here is what I envision as the optimum setup in your situation:

There would be two cooling loops in your system. Each would be fed by one pump, and each would be cooled by one BIX. The trick will be going pump->block->rad->RES. The res would be one large reservoir that would feed both pumps. Since one loop will be producing much more heat than the other, the temps would equalize in the res allowing both to be cooled the best they can be expected.

If you run two pelts in series, you won't get the results you want. Either the GPU pelt will be getting hot water from the CPU pelt or vice versa. One will suffer greatly. If, however, you put them on seperate loops like this, while neither will be getting the coldest coolant possible, they will both be getting much colder coolant than the one that was getting the hot water in series would. In the end, you'll get better overall cooling.

Does that make sense? Sense both radiators would drain into the res, there is no chance of backfeeding if one goes out preserving the component on the good loop. If they are in series both would still get coolant flow, so this is the only drawback to parallel flow, but how often do good pumps go out without some sort of warning (noise, vibration, etc)? This would allow you to do what you want with parts you already have, and this type of arrangement, if there is good drain placement in your loop, will make filling/draining and air removal very easy.

Did that make sense?

[[NH]Naughtyboy]

OMG...this is a nightmare...2 gurus telling me differnt things.
Well one thing I can tell... is that keeping the BIX:s is to late..have already sold them.
The issiue was not just bad cooling...there was also a room issiue...ther wasn´t room in my case to fitt them in a way I was satisfied with.

Originally posted by airspirit
I'm torn on this issue. Because of the pelt, you're going to be producing much more than just 93W, you realize that, right?

Yes I do realize that...it will take care of bothe the cpu heat load and the heat produced by the TEC it self. Right..??


About the loop... my plan is/was like this:

Res-->pump-->rad--->pump--->splitter--->block-->splitter-->back to res.

As you can see I thought of making a splitter out of either copper or polyblocks. The idea is to drill like an Y shape inside...1 inlet and 2 out let...using 2 of them...one befor the blocks and ofcourse one after.

After BigBens post...I thought of put it like this:

res-->pump-->pump-->rad-->splitter-->block-->splitter-->back to res.

So what do you think about it...??

About the fans...I have 2 Enermax 120mm with VR on it...and a 80cfm max each...Should be enough..don´t ya think..??
And ofcourse I will use shrouds on both sides.

[airspirit]

If the barb width of your core is 5/8", that arrangement you described should work for you. The only other thing I don't understand, then, is why you're using two pumps rather than one more powerful pump? If you're worried about case room ... see what I mean? Otherwise, that should work ... just make sure that core can breathe, because you're dumping well over 300W into it!

[bigben2k]

I would go with the 2nd configuration, but Airspirit isn't wrong either.

The point about the GPU receiving hot water, IMO, is moot: if you get 60 gph (1 gpm), and you have 1/2 gallon of coolant in your rig (which is a low estimate), then the water recirculates through each point twice per minute. Now you tell me, how much more hotter do you think that the water will be? Here's a hint: the speed of the coolant through a 1/2 opening @ 1 gpm is 1.6 feet per second.

It isn't an easy issue, but the general rule of thumb is: rads in parallel, blocks in series. Of course the general rule didn't take into account two pumps, but we know that centrifugal pumps work best in series.

So my suggestion is to optimize for flow, not heat. You've got two low pressure pumps so, IMO, it would be best to concentrate on flow optimization.

res-->pump-->pump-->rad-->CPU block/pelt--->GPU block/pelt-->back to res.

[edit] splitter removed.

BTW, your fan configuration should do well.

[TerraMex]

Just a thing. I was reading BB2k posts and something didnt look right. So i went and got my old fluidics manual (its a heavy one).

the question is :

- it states clearly that having two equal pumps in parallel gives you a marginal head gain , but almost doubles the flow.

- also states that in series, you get almost double the head and a marginal flow gain.

Ok, they depend on the pumps and mouting, but...

Now if we shouldnt reduce the inlet of the pump, not to cause any restrictions... why should we set them im series where the second one is particulary dependent on the first's outlet ?

Doesnt make sense to me. Parallel would be a better choise, bigger flow, less pressure.

[bigben2k]

You're gonna make me look it up, aren't you?

Here it is

[[NH]Naughtyboy]

Originally posted by airspirit
The only other thing I don't understand, then, is why you're using two pumps rather than one more powerful pump? If you're worried about case room ... see what I mean?

Well the pump isn´t that much of an issiue...I´m planning to put them next to each other..my case is wide enough to host both pumps next to each other. The BIX:s was another story...didn´t want to put in top of my case due to air traps...didn´t have room in bottom of case...and sandwich them seemed ocward to me.

Originally posted by bigben2k So my suggestion is to optimize for flow, not heat. You've got two low pressure pumps so, IMO, it would be best to concentrate on flow optimization.

res-->pump-->pump-->rad-->CPU block/pelt--->GPU block/pelt-->splitter-->back to res. [/b]

Hmm..well...even though that config. helps me even more with the space and also I don´t have to make the splitters...that feels somehow a little wrong...I mean pump "hot" water in to the gpu block. U think I would get THAT high pressuredrops with a splitter.


Would it be easier/better to ditch thoose 2 1048 and go for a 1250 instead..???

[TerraMex]

>You're gonna make me look it up, aren't you?

Of course. .

Anyway, i've read the threads. Thanks.
I'm not that convinced but hey, that's my problem. I'll get around to test it in the near future (hope so).

[airspirit]

Look away from the eheims if you're pushing two loops. Try looking at a single Danner Pondmaster type 700GPH pump ($50-60 new on Ebay) and you'll have the flow/head you need for two split loops like you described to me.

res->pump->split->(A)&(B)->split->rad->res

Pelt units produce so much heat that you never want to put them in series. The issue with two eheims is that you will be using MORE power (heat to coolant) to do the same job ... and in series it will be at a slower flow rate than one powerful pump.

Series: doubles head
Parallel: doubles flow rate

By using a pump that is twice as powerful (and uses only 50% more power than one of those eheims), you will get the best of both worlds with less heat being inserted into the coolant via the pump (and in this situation, you want to limit that as much as possible because you are going to be heavily loading that rad!!!).

Many people don't like running their CPU in parallel with anything because they don't want to reduce flow to it. This is unique because of the sheer amount of heat that will be generated. You may be able to run them in parallel with the CPU running 1/2" and the GPU running 3/8" to give most of the cooling potential to your CPU while not starving your GPU.

[airspirit]

As far as order, you will want the rad before the pump to cool the water before it hits the pump as much as possible. Water that will get that warm is NOT good for the pump and can increase the chances of failure. Just a tip.

[airspirit]

To everyone:

The reason that running centrifugal pumps in series is preferable for reliability purposes is that if one goes out, the other is there to pick up the slack.

If you are running directly parallel, the parallel loop the other ran in will become a bypass, and most of the flow will backfeed through the dead pump's loop, following the path of least resistance. You'll end up with burnt components every time unless you put in elaborate back-pressure valves to prevent it, but that involves more hassle than it is worth considering that it is easier to just put in one higher-rated pump.

[TerraMex]

I find your lack of faith in out mechanical skills... disturbing.

The point is that many dont want redundancy, but bigger flow. I still belive that parallel gives a bigger flow that series.

Besides, a simple thing to do would be a flow meter from both pumps, or add a parallel circuit to the pump power wires. If one goes, it triggers a relay that closes a valve. Ok, so electrical valves are huge and expensive, but ... you only need to interrupt the flow of the dead pump in the outlet.

It can be done with a manual valve , a switch and a small 12v motor a one (yes, one) gear. Let's just say we had to improvise. It was a pneumatic system but the operating principle should be the same.

[SysCrusher]

Originally posted by airspirit
To everyone:

The reason that running centrifugal pumps in series is preferable for reliability purposes is that if one goes out, the other is there to pick up the slack.

If you are running directly parallel, the parallel loop the other ran in will become a bypass, and most of the flow will backfeed through the dead pump's loop, following the path of least resistance. You'll end up with burnt components every time unless you put in elaborate back-pressure valves to prevent it, but that involves more hassle than it is worth considering that it is easier to just put in one higher-rated pump.

You will get more pressure in series. Not much maybe.

[airspirit]

My point wasn't to belittle anyone's mechanical skills, but to point out that in this pump class, it'd be cheaper and easier most of the time to just use a stronger pump. There are drawbacks to both configurations, and that kind of setup is only good if you want a) redundancy, or b) cost effectiveness for high pressure applications. We're not talking about $500 pumps ....

[TerraMex]

Agreed, i was just giving a few ideias.

[saton472]

im just starting to get into water cooling but heres the way I think you should have it set up
I suggest this because if one of the pumps fail, then you would only have a chance of loseing that one component not both of them. If i'v got something wrong tell me please

[saton472]

oops i forgot to attach the file
If you don't see it download it and then open it.

[[NH]Naughtyboy]

Well it looks interesting...!!!!

What do you say Pro/Gurus...could the figure above be somethinge....or does that also require a lot of pump capacity..???

[scopEDog]

I was thinking more along the lines of:

pump->rad->cpu/pelt->rad->pump->gpu/pelt->res

If you still wanted to use 2 pumps (and still had the 2 bix's). With the single rad/single pump I would go for:

res->pump->splitter->(gpu/pelt-cpu/pelt)->splitter->rad

Are you looking for redundancy? I am looking to use 2 pumps myself, just I'm not sure how flow would be affected if one were to die. I'm interested to see what setup you finally decide to use, good luck.

*Just my humble opinion

[bigben2k]

I had to edit my above post: there's no need for a splitter, at all.

Here it is again:
res-->pump-->pump-->rad-->CPU block/pelt--->GPU block/pelt-->back to res

I think that everyone is exagerating the "hot water" part: the temperature of the coolant doesn't change a whole lot, at various points throughout a loop. If you see more than 2 deg C difference, I'd be surprised.

Correct me if I'm wrong, but the reason for two pumps has nothing to do with redundancy: it just happened that way, right?

[airspirit]

IMHO, the idea saton472 proposed is a bad one, reliability-wise. Again, if one pump goes out, you have a situation where the dead pump loop becomes a large bypass (least resistance). The only way to make it work is if there is only one loop intersection, not a loop in a loop. This would cause failure of both cooled components rapidly.

As far as putting the blocks in series or in parallel, we are not just looking at the effect of the blocks on the coolant, but the radiator as well. Granted, this will not make a massive change, but will be a minor change for the better.

Since there is one rad that is going to running a high heat load, odds are the water will be warmer than we are used to relative to ambient. Any way that we can reduce the temperature of the coolant to the blocks would be good. In this case it would mean parallel block loops. Again, we are probably talking on the order of 3-4 degrees C, but every bit helps, especially when you are going to the lengths of using pelts to lower temps. Why waste part of your advantage? Especially considering that it will add no extra complexity to the design besides two "T" splitters.

[bigben2k]

Good call, Airspirit: if one pump dies, some of the coolant is going to recirculate within the CPU and GPU loop, bypassing the rad.

And the rad will need to be dissipating a lot of heat, or the coolant temp will be higher.

I see your point about running the blocks in parallel: you'd supply the blocks with fresher water, rather than give them a higher flow rate: given that the blocks are Swiftech, and that they should have a relatively low pressure drop, and that their performance is not considerably improved with more flow (an assumption), then you'd be right. But is my assumption correct?

[airspirit]

In this case, I believe that the temperature difference in the coolant (3-4C) will give a better gain than the flow gain from series (probably 1C in those blocks). It also allows him to stick to higher tube ID (5/8" to match the HC) throughout the rest of the system, further reducing head loss (and increasing flow rate!). Parallel is win-win in this case.

Also keep in mind that in series you are also going to eat up more pump head than parallel due to the higher flow rate per block.