An Optimal Way To Choose Components?

You know what I think we need? General Tips For Purchasing a Custom Waterooling System. For an optimal design one must consider the following:

A) The less heat the pump puts into the water the less heat has to be removed by the heatercore.

B) But at the sametime flowrate has an impact on cooling performance and usually higher flow/pressure pumps dump more heat into the water.

C) Some systems react to flowrate changes differently.. i.e. a White Water block has a larger performance increase than a Swifty for a given increase in flowrate.

So basically it boils down to this:
# How does one go about creating a high perfomance cooling solution?
# How do you balance all of these variables into something that is efficent?

============================================

Critique my planned system:

Iwaki MD-20RLZT
2 x Chevette Heater Cores in Parallel (shrouded in Push/Pull config.)
D-Tek WW Poly Top
D-Tek Chipset Cooler in Parallel with GPU Cooler*

* Still looking for GPU Cooler without Aluminum

Here are things I am positive about :
- I do not want any aluminum in my system
- I want the WW CPU Block (Cascade would be better... but I cant get one)
- I want two heater cores.

Now would this system benefit from a different pump? I am not really factoring in the price/performance issue because I want my water cooling system to be somewhat future proof and extendable. The pumps I am considering are:

* MD - 30 RLZT
* MD - 20 RLZT
* Johnson CM30P7-1

But how does one know without actually buying all the pieces, calculating flowrate and C/W for the heatercore if the system is efficent?

I'd remove the parellel setup... run the GPU and NB in series. Too many flow problems with serial setups. Your pump is more than powerful enough for the whole she-bang.

You will need to have a very large case for a pair of Chevette HCs. I've often thought that it would be neat to have a pair of smaller (chevette sized) heatercores on their sides, one on top of another. That would be about as high as a typical Caprice HC and only slightly wider. It would make for pretty nice tube-runs and allow for an easy parellel configuration. Throw in a pair of fans and some shallow-shrouds and the space-problem isn't *that* bad.

Knowing the potential effeciency of a particular planned setup is simply a matter of experience. Typically speaking, if you combine a well-reviewed water-block with a highly-regarded powerful pump, throw in a good radiator and provide ample airflow... you'll have a very well performing system. There are always small things you can do to improve *any* system, it just becomes a matter of taste. Sure, I could probably drop an additional 2° by implementing blahblah to my loop... is it worth the space and cost? Shrug.

That is a good point. I guess it doesn't need to be optimal as well as it performs well.

"Too many flow problems with serial setups." Do you mean too many flow problems with parallel setups?

Here is a question for Cathar: Why are your Heater Cores in series? Everything I have read at this forum supports parallel heater cores yielding better results.

Yeah, that's what I meant =p. Reword it ^_^. I doubt that there is very much difference between heater cores in parallel vs series.

Does the Maze4 GPU cooler have aluminum top? I don't want to open mine since it has that voids warranty if you open it.

96

ninety6 said:

Does the Maze4 GPU cooler have aluminum top? I don't want to open mine since it has that voids warranty if you open it.

96

Click to expand...

Yes, I believe it does.

I have been reading-up on water cooling a little over a year now (researching and stuff) and I was going to do a similar post today or tomorrow since I'm about to buy my system soon.
My system will be very similar to MrMOSFET that's why I'm posting here. I will be following this thread closely and will be helping out if I can.
I already have my case which is a Lian-Li PC71 so space will be of little concern to me and I also tryed to get a Cascade but don't know what will happen to my previous order (ordered but not paid) so I'll be getting WW, Iwaki MD-20RLZT, 2 Pro Core Radiators ,no chipset block and a swify for my aiw 9800 (it's the only one that will fit)
MrMOSFET are you planning on using a reservoir? and why?
and I had read that a serial setup for my radiators would be better intead of a parallel.
Also I think that the MD-20RLZT is a very well balanced pump between heat and efficiency the MD-30 RLZT would be too much for me since the 20 is more than capable and I don't whant to worry about radiator pressure.

I think the problem with serial is that it hurts your flow rate. I'm fairly sure that Cathar's pump can handle a lot more than most people's pump could.

Don't be a sheep.

Serial vs. parrallel depends mostly on the radiator used and minorly on the water block. For instance, two FEDCO 2-342 core in SERIES have a total flow resistance slightly less than a SINGLE 2-261 (Chevette core, Dtek Pro Core) at any given flow rate.

I've experimented with serial/parrallel setups in search of my water-cooling nirvana . I've had better results with my Cascade and two 2-342 in series than I did with the same cores in parrallel. I feel this is because of the Cascade being the bulk of the total flow resistance making the arrangement of the 2-342 cores inconsequential as far as flow goes.

Would I put two 2-274 (Caprice) cores in series with a Cascade? Heck no. A single 2-274 has four times the flow resistance of a single 2-342 core. I would deffinately put 2-274's in parrallel.

IMO, if you have the room to run two 2-261 cores in series or parrallel, don't. You'd be better off with a single 2-342 (preffered) or a 2-274 because you're adding unnessesary restrictions via extra tubing, connections, and fittings to split the flow between two cores.

What are these 2-342's?

MrMOSFET said:

* Still looking for GPU Cooler without Aluminum

Click to expand...

seem to be a couple of possibles here :
Cyclone Fusion SL V1.02 and Fusion HL V1.02 From Silverprop - machined brass top with integrated barbs.

or one of these dangerden's may fit (depending what card it's for)

AngryAlpaca said:

What are these 2-342's?

Click to expand...

Same size as a 2-274 (Caprice), but single pass. In/out are on opposite corners of the core.

Buttload-O-info on radiators in this thread @ procooling.

MrMOSFET said:

Here is a question for Cathar: Why are your Heater Cores in series? Everything I have read at this forum supports parallel heater cores yielding better results.

Click to expand...

Why? Let's explain it with math shall we?

I'll start by saying here and now that I don't have flow/performance information for the cores that I use. I will, however, demonstrate the example using the Serck information presented by BillA, for which the cores that I use can be probably assumed to be "close enough" for our needs.

First, here's the Serck performance graph:



First thing to notice here is how the performance improves as flow rate is increased. Bill's data is presented as BTU/hr (just labelled as BTU's though) for which 1 BTU/hr = 0.293W. Also I'll be using the yellow line as that more accurately portrays the level of air-flow that I get through my rads with my blower set at low speeds like I have it. Bill's figures are calaculated for a 10C delta between the water and the air.

Now the cores I use have a PD (pressure-drop) of 0.7mH2O at 10LPM.

Need to calculate the predicted flow rates with my pump, and the cores in series/parallel. I won't bore you all with the math, but it works out to about 10.4LPM with the rads in parallel, and about 9.8LPM with the rads in series.

In parallel, each radiator is receiving 5.2LPM (or 1.38GPM). In series, each radiator is receiving 9.8LPM.

I have my CPU very highly overclocked/over-volted, and under load it dumps around 120W of heat into the loop (my estimate), and the pump around 55W of heat into the loop for a total of 175W of heat.

Look at the graph, we see about 1080BTU/hr at 1.38GPM (yellow line), or about 31.6W/C. Assuming the flow is split evenly, and so is the heat load, the radiator pair are dissipating 63.2W/C. The water will reach 175/63.2 = 2.77C above ambient for in-parallel operation.

In series, the lines don't extend that far, but let's assume about 1250BTU/hr, or 36.6W/C. Now the water volume flowing through the radiator has a thermal capacity of 9.8 / 60 * 4180 = 683 W/C. This means that the first radiator will always take away 36.6/683 = 5.36% off the water temperature that enters it, this the second radiator is just 94.64% as efficient as the first in the series. This gives us a total W/C of 36.6 + 36.6 x 0.9464 = 71.2 W/C

This means that in series, the water temperature rises to 175/71.2 = 2.45C

i.e. the water temperature is about 0.32C cooler in series than in parallel.

So what's the block's performance difference between 9.8LPM and 10.4LPM? Probably around 0.001C/W. The CPU is dissipating 120W, so the CPU will be warmer by 0.12C due to the slightly lower flow rates, but the water is 0.32C cooler, giving a net gain of 0.2C.

Pretty small values, I know, but it demonstrates the point that having the radiators in series can be better than in parallel provided:

1) The radiators are of a fairly low pressure drop in comparison to the waterblock and do not significantly affect the flow rate when in series
2) The radiators do perform significantly better at higher flow rates than at lower flow rates

Both of the above are what UberBlue alluded to above, and both of those apply to my setup.

Ah. Those large single pass ones. Most of the people I know, on EOCF, that use a pair of Chevette ones upgraded, rather than using one of those. When I asked what they were, I was referring to the common name.

Edit: So, Cathar, it seems to me that what you're saying is that it depends on the resistance of your radiator, and the resistance of your waterblock? Small values they may be, but we're dealing with some sensitive equipment, and we want to tweak to the max. If I'm understanding this, what you're saying is, with a high resistance waterblock and low resistance radiators, you'll have better luck (In general) in series, but with a low resistance waterblock and a pair of high resistance radiators, you'll have better luck in parallel? Also, is this because of the fact that heater core style radiators (What is the other kind? Tubing winding through the fins?) do better with higher velocities, that would spread the water out more?

Edit #2: Reread uberblue's info, and I would like to find out where to get information on resistance of heater cores.

AngryAlpaca said:

Edit #2: Reread uberblue's info, and I would like to find out where to get information on resistance of heater cores.

Click to expand...

Buttload-O-info on radiators in this thread @ procooling.

The above link is the only reliable testing of any sort that I have seen on that style of heater cores. At the very end of that thread there is a heatercore flow resistance calculator specificly for that style of core. I have a feeling that if reliable thermal testing is done on that style of core, then the spreadsheet could be redone with thermal data in a similar manner. I wouldn't bet the farm on it, but it's better than guessing.

Unfortunately, as I understand it, radiators are a PITA to properly test. That's why you hardly ever see reliable testing done on them. When BillA was still testing, he charged around $400 a pop. and that was for a bulk discount of 10 different cores at once.

redken said:

using a reservoir? and why?

Click to expand...

No.. I think I am going to go with T-Line because it is cheaper, takes up less space, and it is possible that you can get a better flowrate with T-Line, seeing a res. can introduce resistance into the circuit if done incorrectly.

Also I have read a number of posts where peoples res. starts to leak... not something I want to happen

So, a reservoir CAN restrict flow, and it CAN leak, so you aren't using one? I'm not saying to get one, I'm just saying that's pretty flawed. A tee line can do both those things. My mouse is screwed. Hmm... It seems that radiator performance is pretty much as good as it gets at 6 LPM, for those he tested. It was hard to follow the data, as I didn't know what colours indicated what, or whose name was associated with what radiator.

AngryAlpaca said:

So, Cathar, it seems to me that what you're saying is that it depends on the resistance of your radiator, and the resistance of your waterblock?

Click to expand...

Yes - it depends on the various flow resistances.

Note that I do classify both the White Water and the Cascade as moderate resistance blocks. It's just that most blocks that have come before them are quite low resistance.

By my definitions:

Low resistance: <0.5mH2O @ 5LPM
Moderate resistance: 0.5-2.0mH2O @ 5LPM
High resistance: >2.0mH2O @ 5LPM

For example, the Innovatek X-Flow has a pressure drop of ~3.3mH2O at 5LPM, which is about 2.5x that of the Cascade. Now that is what I coin a "high resistance" block, being one that demands a strong pump to even achieve 5LPM.

AngryAlpaca said:

If I'm understanding this, what you're saying is, with a high resistance waterblock and low resistance radiators, you'll have better luck (In general) in series, but with a low resistance waterblock and a pair of high resistance radiators, you'll have better luck in parallel?

Click to expand...

Yes. That's pretty much it. The radiator does also need to exhibit improving cooling performance from higher flow rates as well. Most do.

AngryAlpaca said:

Also, is this because of the fact that heater core style radiators (What is the other kind? Tubing winding through the fins?) do better with higher velocities, that would spread the water out more?

Click to expand...

The "other kind" is the winding tubing style. They, too, do improve with higher flow rates, but seem to have a "low flow" peak as well, as it they perform well at two points on the graph, rather than simply "better with more".

AngryAlpaca said:

So, a reservoir CAN restrict flow, and it CAN leak, so you aren't using one? I'm not saying to get one, I'm just saying that's pretty flawed. A tee line can do both those things.

Click to expand...

Yes, that is true, perhaps I should have used "is more likely to" rather than "can".

I would be willing to bet that reservoirs tend to leak more often than T-Lines. A proper made res. is probably every bit as safe and efficent as a T-Line (and a whole lot [edit]easier to [/edit]bleed!). I just dont like the idea of extra joints and adhesives that could possibly fail.

Not to mention that a res. does take up more space and add to the overall cost of the system with (arguably) very little benefit.

I on the other hand will be using one of d-tek's Bay-Rez since I think they are well made and will add to the "look" I'm planning for, unless somebody really advises otherwise.
On the other hand I find very interesting what AngryAlpaca and Cathar were discussing, as I said I've been reading up on Water Cooling for a little over a year now (on and off) and there is always something new to learn; very appealing subject I must say.
Will 2 Pro Core Radiators and a White Water in parallel be a "good" way to set these up?(the rads being high resistance) The White Water being considered a moderate block confused me a little.