PDA

View Full Version : Parallel Loops - Same Radiator/Pump: opinions?



Arcygenical
07-25-2008, 06:33 PM
I asked this over at [H] but got less than a stimulating discussion going on. It seems we've lost all of our intelligent water coolers recently to outside endeavors. I have a feeling I'll be making XS my new home, and leaving [H] for the OT stuff ;).

So, I posted the following:

Well, I've decided I need to watercool my NB and my MOSFET assembly on my p5b deluxe motherboard... I just don't like the fan setup I've got right now, and I can eliminate another source of noise in my case by watercooling the motherboard. Additionally, the top of the NB heatsink (the stupid plate that the brilliant Asus engineers included... Negating any advection benefits of the HS) reaches a blazing 68c during gaming... I can only imagine the chip will hit 75c, maybe even 80c!

I do, however, have a small case, without enough room for a full second loop (mainly a radiator and pump just won't fit anywhere internally) so I need to split my current loop into two portions: A highflow loop for the CPU+Radiators+GPU and a low flow loop for the NB and the SB.

Here's a picture of my current setup.

http://i99.photobucket.com/albums/l308/Arcygenical/Flow.jpg

Water travels from the light blue (pump) into the CPU block, then to the GPU block (purple) then to the main triple radiator (red)... it then travels down the radiator (it's a single pass radiator) along the side of the case behind the drive cage, and using some elbows, into the second radiator in the bottom left corner.

I'll be splitting the tubing in two places... First, where the darker blue meets the larger green line, which will allow water to flow to the NB and MOSFET blocks, then back through the case floor, to connect with the light blue line at the very bottom...

This second loop will be using very restrictive components (6mm tubing and fittings, mainly) to force more water through the main loop. I will be putting a ball valve in series with the smaller loop, to adjust restriction as required. My question is as follows:

Since both the split points are BEFORE the pump, do you think this will work out for me. I seriously think it will, but I'm not entirely sure, and don't want to drop a few hundred dollars to find out it doesn't.

Cheers!

So... What do you guys think? :D

Conumdrum
07-25-2008, 08:51 PM
Ya, use the 120 for the NB/Mosfet cooling. Parellel makes all sorts of odd things happen.
BTw, very clean setup. Hate to see another pump in there, but ohhh well.

Martinm210
07-25-2008, 09:08 PM
It might work, but you may also have so much more restriction in the green loop vs the radiator that you get almost no flow at all through it. If I was going to try parallel on something like that, I'd leave the CPU alone and just split out between the GPU block and the chipset/mosfets then back to gether again before the radiators.

Or just go with series and forget about parallel. I think parallel is ok but you need to be looking at the pressure drop closely between the parallel loops and I wouldn't do it with the CPU block, I'd keep that part in series.

Jedda
07-26-2008, 01:11 AM
Get yourself some room near your pump by moving the HX620 to the right might help fit what ever you wind up doing.

http://img.photobucket.com/albums/v408/sulk/pe01a_02.jpg

Am adding this to my box atm. Should have more to say in my WIP on monday.

calvin
07-26-2008, 04:26 AM
In plumbing systems differential pipe sizing is used to manage flow in the legs of a distributed system. Smaller tubing on the parallel loop makes sense as the flow need is much less, not to mention the routing beneift in a smaller case such as yours.

Your plan appears to be a sound idea.

F1ZZY
07-26-2008, 06:08 AM
You already have 2 radiators; making 2 loops really only needs an additional pump and a DDC could fit in there somwhere.

I'd be tempted to put the CPU and gfx on the tripple rad and then all the other stuff dump onto the single rad using a DDC pump.

RickCain
07-26-2008, 06:23 AM
I would use the lower radiator for a second loop running 6mm or 3/8" and a small DB1 pump. You surely can fit a DB1 in your case. You don't need more than your top radiator to cool the CPU and GPU.

IanY
07-26-2008, 06:30 AM
Way too much for a lone D5 pump. It would be mediocre, at best. And give up on 6 mm tubing, please. Leave that for the Germans. Stick to 3/8 or 1/2 or 7/16.

MpG
07-26-2008, 06:45 AM
Have to say, love your case layout. Amazingly clean and tidy.

You considering grabbing a some Thermalright air sinks? Would save a hell of a lot of hassle, and they're proven at this point. But if you're determined to watercool the chipset, you're probably going to have to resign yourself to having a less-clean case. Whether you go two loops, a parallel set up, or series, things are simply going to become more crowded and messy inside the case.

Going with your parallel idea, the pump NEEDS to be before the split, and the split NEEDS to rejoin before the pump. In other words, you can't position the pump in one of the parallel splits.

The idea could probably work, although I strongly doubt it would be the most effective method. For best results, the tubing/piping layout would have to strongly favour the CPU/GPU path. The layout should be kind of like a t-line, except that the offshoot would lead to the chipset pathway. You don't need very much water to keep a chispet cool, so a minimum of water would need to be diverted.

That said, I'd be much more inclined to try the dual-loop idea.

IanY
07-26-2008, 06:52 AM
Even a RD30 will heat up and blow up ! Differential restriction? 6 mm tubing?

DB1.. please.. my 3 year old pees further than a DB1 when he pulls down his diaper. You should see the head pressure of that pee.

The only DBs that I like to talk about are Aston Martins.. and even then lol

That case is a midtower, but its got space. You already have two radiators. All you need are a better DDC3.2 w/ XSPC top for your cpu, and then move the D5 for your chipset blocks. You can even fit more radiators in that case.

IanY
07-26-2008, 06:59 AM
At the rate we are heading, this would be XS, and then there's a need for a Lunatic PITA subforum hahaha

Martinm210
07-26-2008, 07:14 AM
How about a little cpx-1 on only the chipsets/mosfets in their own loop?, maybe even another 120 sized rad in the back for that loop and leave the current one like it is.

I've been using one for my rad testing since I fried my DDC2 and so far getting 1.5GPM on the test setup for a couple of radiators. They are pretty nice for low heat low restriction loops as they only consume less than 6 watts, so the heat dump is really nice and low. probably in the 3-5 watt range

Martinm210
07-26-2008, 07:16 AM
Here is another idea, circa 2004, this guy was getting sub-ambient temps, that TEC'ers can only dream of....

This might be your solution since you don't have the room for additional rads and pumps in your mid-tower case....


Here's my radiator, made of 50' of 3/4" copper pipe:
http://farm4.static.flickr.com/3055/2703233631_6c132fce77_o.jpg


The hole for the radiator has plastic so I can keep the ground wet to increase heat transfer.
http://farm4.static.flickr.com/3144/2704056036_cb37d0a86e_o.jpg


Iwaki 20RZ
http://farm4.static.flickr.com/3260/2703233757_7c4ec9412f_o.jpg


The water is 10-15C. To prevent condensation without insulating, I sealed up the case so no air comes in or goes out. The air is recirculated inside and cooled by the heatercore, which of course has the cold water going through it.
http://farm4.static.flickr.com/3109/2703233833_3a8334c0d1_o.jpg

Wow, that's intense..:up:

IanY
07-26-2008, 07:17 AM
If I did that, I would be hearing from my wife's lawyer :D

NaeKuh
07-26-2008, 07:59 AM
HOA would have a field day.

no i swear i was chasing a ground hog!!


or sorry my dog found my hidden weed stash and started digging all over the place. :rofl:

IanY
07-26-2008, 08:03 AM
Homeland Security will stop by and investigate lol There's fertilizer in the garden shed !!

Arcygenical
07-26-2008, 09:18 AM
Wow... Lots of replies! Much appreciated guys!

I'd just like to take the time to point out two loops, with a bunch of parallel components, that work fine:

Parallel Hex by mysterrae (Which I'm sure most of you remember) And Maxhardware (I think that's his username, I can't find the post right now...)'s new loop (dual pumps, tonnes of tubing, and the Koolance Ram blocks in parallel with other parts of the loop). Both supposedly work very well.

And I must take the time to clarify the main purpose of this machine: Silence. It's often used in a recording booth, and therefore must maintain a noise level of <25dba. Ergo, I'm not looking for extreme performance (Sorry!), but I am interested in fluid theory, which is understood by many of the members here.


Why split the loop, makes no sense....?????......:shrug:

Just run the rads in series and be done with it...:up:


The rads are already in series... I Want to split the loop off, and leave the rads in series. With my fans at around 5v, the computer is completely silent, removing the bottom radiator requires me to run my fans at 8v, which are noticeable on booth recordings.


It might work, but you may also have so much more restriction in the green loop vs the radiator that you get almost no flow at all through it. If I was going to try parallel on something like that, I'd leave the CPU alone and just split out between the GPU block and the chipset/mosfets then back to gether again before the radiators.

Or just go with series and forget about parallel. I think parallel is ok but you need to be looking at the pressure drop closely between the parallel loops and I wouldn't do it with the CPU block, I'd keep that part in series.

Good advice. Well, in that case, I'll probably be using 3/8" tubing and a ball valve to adjust my flow. Really, with the voltages I'm running through the chipset, a small, slow fan would suffice (but due to noise...), so I'm assuming I only need a very tiny bit of water to flow through.


You already have 2 radiators; making 2 loops really only needs an additional pump and a DDC could fit in there somwhere.

I'd be tempted to put the CPU and gfx on the tripple rad and then all the other stuff dump onto the single rad using a DDC pump.

Your post mirrors my original plan. Then I thought, dealing with the noise of a single DDC/MCP655 is enough for me, I could hardly imagine the harmony between them.

I've got around 5 DB-1's kicking around that I could use... But the triple radiator at the top (an Xflow, which hurts it's performance) just isn't quite powerful enough to dump the heat of a Q6600 and a 4870 at load, without putting my fans on higher than I'd like.



Maybe the db1 is accepted on [H], but it's not nearly Xtreme enough to even be mentioned here.............:down:

Bahaha. They're good in their own right, but I'd never use one personally!


How about a little cpx-1 on only the chipsets/mosfets in their own loop?, maybe even another 120 sized rad in the back for that loop and leave the current one like it is.

I've been using one for my rad testing since I fried my DDC2 and so far getting 1.5GPM on the test setup for a couple of radiators. They are pretty nice for low heat low restriction loops as they only consume less than 6 watts, so the heat dump is really nice and low. probably in the 3-5 watt range

And that's what I think it comes down to. I'm going to try my setup, throw in a flow meter and take some readings. If it doesn't work (like most of you tell me it won't, at least, up to my expectations) I'll throw a cpx or a db1 in series with that second loop, and feed it into the main loop. If that doesn't work, then I'll re purpose that front radiator (I'm switching it to a BIP mind you, rather than a BIX), or even experiment with stacking radiators.

I'm mainly doing this because of the "what if" factor... And was just looking for some discussion to help me find flaws in my theory (of which, I definitely found one).

I'll post back when I get my stuff in... Just waiting on the EK 4870 block to drain the current loop... And get rid of this:

http://i99.photobucket.com/albums/l308/Arcygenical/lolsetup.jpg

Yes, my 8800GTS block is currently being held up by zip ties, waiting to be replaced with the 4870 block ;)

Arcygenical
07-26-2008, 10:14 AM
Sorry you are now on Xtreme systems... We live in the Xtreme here....

Silence is what I'm after also, even more so than you. Are you willing to run a CPU only loop with (2) PA120.3's ..???.. I'm going to try it...

:D. Oh, I know. I enjoy the Xtreme just as much as any other guy! But, barring this fanatic predisposition towards performance here, posters tend to have a much higher knowledge (and interest level) of the theories and physics behind water loops. I know this system is far from extreme (except, possibly extreme silence), but I needed to draw from some pool of knowledge.

I'd love to run 2 PA120.3's to cool just the CPU, but my system has maximum size clearances for both ease of transport and general manageability (I often bring it with me when I'm mixing or recording, since It's set up perfectly to my tastes... and there are certain problems associated with this requirement). I also hate clutter, both cable or tubing... So, yeah.

Careful of diminishing returns though... I've never been in a room with a SPL of under 22dba... And 3 yate loons at 5v measure around 24dba from 1 foot. Run them at 4.1v and they don't surpass ambient level noise. I wonder how much heat a PA 120.3 would dump with 3 yates at 4.1v...

IanY
07-26-2008, 11:07 AM
I wonder how much heat a PA 120.3 would dump with 3 yates at 4.1v...


No offense, but a PA120.3 is a radiator. It doesn't produce heat. It removes heat.

I hope that was a typo. If not, here's a whole lot of learning ahead.

You want silence, and yet you prefer a Hardware Labs radiator. Hope you see the paradox.

IanY
07-26-2008, 12:45 PM
Returns are always diminishing.

I don't drive a 75 hp 1.0 liter 3 cylinder car, although its capable for driving at 65 mph on the freeway. Anything more powerful than that provides diminishing returns.

In fact, why water cool at all? Water cooling provides diminishing returns compared to stock air cooling. In fact, the marginal cost is huge.

Arcygenical
07-26-2008, 08:17 PM
No offense, but a PA120.3 is a radiator. It doesn't produce heat. It removes heat.

I hope that was a typo. If not, here's a whole lot of learning ahead.

You want silence, and yet you prefer a Hardware Labs radiator. Hope you see the paradox.

What's with the hostility here? :down:

Heat "dump" is the term commonly used to denote the thermal wattage moved from the water to air, when air is passed through the fins of a radiator. Heat "load" would be a term commonly used to denote the production of heat in a component. Just as a lightbulb is considered a "load" on an electrical circuit.

Where did I say I prefered a HWlabs radiator? I had no *choice* in the matter due to my setup. I selected the BIPXflow for 2 reasons, 1 it's thin, and 2, it's single pass. I required a thin radiator due to space constraints and the opposing barbs helped a ton with tube routing.

No offense, but I'm quite learned in the area of watercooling. I've been doing it for three years, two years professionally, and have a vast array of experience under my belt. I've never done a parallel loop before, as I've never needed to, and I'll admit, I know relatively little in this area outside of theories (which, for the most part, appear wrong in practice).


When OP want to step up with the Uber Xtreme one day, he'll hopefully have a better understanding of what he's spewing.

What was I thinking, buying a case that's made for (4) PA120.3's...???...

Buying (4) MCP355's, (3) XSPC red tops, (1) XSPC top, (4) PA120.3's, (4) PA Shrouds, (14) YL D12-12SM fans, (2) Aquaero's and (8) Mcubed Fan Amps (to control my pumps/fans for Xtreme Silence) ..??..

Who would have thought, adding a second PA120.3 to my Dual MCP355, CPU only loop, would introduce even more heat into the loop...???...

I now know using a single D5, (4) PA120.3's, with a quasi-parallel loop, would've been the better choice...

Where did I say *anything* similar to the :banana::banana::banana::banana: you just came up with? Where did I ever say your setup was a poor choice. Where did I ever say my setup was more "extreme" than yours? In fact, I never said that, I never even commented on your setup, your choices, or your loop design.


You state that you've got "(5) db-1's just kicking around" and then in the same response state that "I'd never personally use one" ??? Maybe it's just me, but did you just totally contadict yourself in the same answer to someone else's statement..??..

Yes. I've got a lot of *stock* kicking around, because I build and mod systems for a *living*. I produce around 4 watercooled builds a month for students on my campus, often with full paint jobs and modifications. I would never use a DB1 myself, but I've got them in stock. Imagine that.


OMG, your going to further restrict your loop with a "flow meter"... But one that does a complete 180 and accepts 1/4" ID hose....

Yep. You got it slugger. I guess you missed the meaning of "Take some readings" a phrase commonly used to denote, "adding it, then removing it". I never said it was permanent, nor did I say it was staying in the loop.

Then if your theory regarding the cpx/db-1 in quasi-series, in your quasi-parellel loop, doesn't work, that your going to RE PURPOSE (WHAT???) your front radiator. The one your switching to a BIP, not a BIX (like it matters) or stack radiators..??.. Again, WHAT??? you complain about the lack of space in your mid-tower case, but you want to stack radiators ???????

Re-purpose means to change the purpose of something. The front radiator is currently used to help cool the main loop. Reporposing it would mean to remove it from the main loop, and put the MOSFET and NB blocks in series with it, alone, and a db-1. I'll use a db-1 because these other components do not require high flow, nor do they require a pump with lots of head pressure. I'd never use a db-1 for a main loop, but for something with around a 30w heat load (NB and MOSFETs) I could make an exception.




Your posting about your revelation into the theory of watercooling for the "what if" factor???? I do truly believe more than "one" flaw has reared it's ugly head, with this epiphany of yours.....

According to you...

Dude, get that EK 4870 block overnighted, I'm on pins and needles to see this outcome...

As soon as PPC's has it in stock I'll have it. Importing from EK directly would cost me around 160$ CAD before shipping

You want to have a theoretical discussion on watercooling, but you secure your blocks to your GPU's with zipties?????

So? What's wrong with using zip ties to hold the block up while it's hanging off the card? It's not like it's actually ON the GPU. Ever zip tied a fan to a HSF or some sort of addin card while running a quick benchmark? Hows this any different?



In fact, why water cool at all? Water cooling provides diminishing returns compared to stock air cooling. In fact, the marginal cost is huge.

Wow. You totally missed my meaning there. I was speaking about acoustic diminishing returns. You said you were "more after silence than I was" (paraphrased) and I inferred that, below a certain noise point, you will see no actual difference. Unless you run your system in a recording studio or a DJ booth, under ~28dba will be a useless return.


Some of you guys really need to calm down. Now I remember why I don't post here often... There are a lot of well educated individuals that reside on these forums, but for the most part, you're all too trigger happy. Even if you don't understand something, you immediately assume the person whose posting is crazy.

Jedda
07-27-2008, 02:41 AM
Have my cpu on a PA120.3 imho, adding a second is really into diminishing returns.
Any fan at all and the one rad is within a degree of ambient.

RickCain
07-27-2008, 09:12 AM
What's with the hostility here? :down:

Welcome to Xtremesystems liquid cooling forum!


A db1..??.. :shrug:

This is Xtreme Systems my friend, not [S]oft Forums..errrrrrr...[H]forums....

Maybe the db1 is accepted on [H], but it's not nearly Xtreme enough to even be mentioned here.............:down:

Spoken by a guy who pumps an AC duct into the side of his case? :rofl:

ripken204
07-27-2008, 09:29 AM
RRR - how about 2 DB1's ? surely you cant make fun of Petra(Alex) for that.. they are very queit

IanY
07-27-2008, 10:01 AM
If you like to use lingo like "heat dump," then at least understand that it refers to the level of heat transfered from a component to a water loop (like a pump, like a video card, like a cpu). Radiators are measured using thermal dissipation.

If you would like to call MaxxxRacer and the rest of the moderators and founders of XtremeSystems wrong (or something stronger), then be my guest.

http://www.xtremesystems.org/forums/showpost.php?p=532353&postcount=1

There wasn't a drip of hostility in my post. If you like to grace it with a hostile reply, also be my guest. I wouldn't be participating in hostile discussions, and those will be handled accordingly.

Any half brained nitwit can watercool. Its not brain surgery. My 12 year old son has also been doing it for three years now.

I'm way too old to discuss how long I've been doing this nonsense. My age would only be embarrassing.

I will take my leave now and I bid you good day.

Welcome and have fun here.

Movieman
07-27-2008, 10:22 AM
Well it was a quiet Sunday with my coffee till I got a reported post here..
Friends first guys ok?
Thanks..
No, no mod hat, just said as a friend.

Arcygenical
07-27-2008, 07:07 PM
There wasn't a drip of hostility in my post. If you like to grace it with a hostile reply, also be my guest. I wouldn't be participating in hostile discussions, and those will be handled accordingly.


Yep, I wasn't really referring to your post directly, I just quoted it since it seemed overly condescending, and was the first example of so in the thread.


+1 ......... The only hostility was coming from the original poster.....

I'm out too........... enjoy your quasi-serial / quasi-parallel loop run by db-1's in series....

I'm only replying to this because I know "I'm out too" means you'll continue reading this thread... Probably as your first choice for awhile.

There was no hostility in any of my posts. I believe the very first example of hostility in this thread was the post:


When OP want to step up with the Uber Xtreme one day, he'll hopefully have a better understanding of what he's spewing.

What was I thinking, buying a case that's made for (4) PA120.3's...???...

By yourself :p.

I'm going to get this :banana::banana::banana::banana:show locked. No point continuing this crap any longer.

shachar2
07-28-2008, 02:45 AM
Spoken by a guy who pumps an AC duct into the side of his case? :rofl:

it's the best solution I've ever seen for pc cooling so far.
when designing a room/house you should think about putting your ac near your computer and doing exactly that.

that is until Tiny refrigerators (http://news.uns.purdue.edu/x/2008a/080619GarimellaMinicool.html) will come into the market in a decade or so

jonny_ftm
07-28-2008, 03:33 AM
Silence is what I'm after also, even more so than you. Are you willing to run a CPU only loop with (2) PA120.3's ..???.. I'm going to try it...

A 4+3 loop and a 3+2 second loop here :rolleyes:

By the way, just at the OP:

Nice build, but I'll also agree to what the nicer and less nicer people told you:

I'll avoid parallel loops on your setup. All in series would be even better. Two pumps / two loops is the best

_G_
07-28-2008, 04:12 AM
Back on topic, i think something like this is what you're looking for.
http://www.performance-pcs.com/catalog/index.php?main_page=product_info&cPath=341&products_id=20815
Put one between the pump and cpu for you're supply and another right before the first rad for you're return. you might be able to even plum you're return line into the outlet side of you're new gpu block.


ps: selling you're old gpu block? :D