Chylld's guide to watercooling

The alternative to direct air cooling

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Chylld's guide to watercooling

Post by chylld » Tue May 11, 2004 11:22 pm

Chylld's guide to watercooling

Watercooling is becoming ever more popular these days, no one doubts that. More and more people are coming to this forum wanting to know a bit about this and a bit about that and in general, a bit about everything to do with the phenomenon that is watercooling. I'm starting this thread in a hope that it can be used to the benefit of all users, new and old, so that we can better work towards our own (and others') quiet watercooling solutions.

I am not trying to pretend to be a watercooling god here, so I am more than welcome to suggestions so that this first post may be updated with the latest and most correct information.
Just keep in mind that we have a common goal here: silence.

Overview

Watercooling can be summarised as thus:

Advantages:
  • lower operating temperatures
  • higher overclocks
  • localised hotspot management
  • a potential for a quieter system
  • it just damn looks cool
Disadvantages:
  • cost
  • maintenance
  • risk
  • packaging (i.e. fitting everything into a case!)
The benefits of watercooling arise from the fact that water is much better at taking heat energy away from a heat source than air. This gives you lower temperatures, but what's the use of this if your computer is already stable? Because it also allows you to run the fans slower, hence your system will be quieter, as well as cooler. It also allows you to overclock your processor. Think of it this way: you could buy an athlon 3200+ for AUD330 and leave it at stock speed, or you could buy an athlon 2500+ for AUD140 and overclock it to over 3200+ speeds. The overclocking headroom given to you by watercooling has saved you AUD190 in this case. This is only a rough example with estimate numbers, but you get the idea.

Watercooling also allows you to focus on removing heat from a singular source: the radiator. Airflow inside a case can be tricky to manage, as the CPU is over there and the video card is over there, etc. With watercooling, you don't need to worry about how this fan is stealing airflow from this fan, or this is obstructing that, etc. All you have to do is cool the radiator.

Because of this, you can use 1 or 2 large 120mm low-airflow fans to cool your cpu, gpu and whatever else you have plugged in to your loop. Adding a new powerful graphics card? No worries, just plug it into the loop, and you won’t have to suffer from additional noise from additional fans. If desired, turn the radiator fan speed up a tad to cover the added heat source, but that’s it. A decent radiator has more than enough cooling power to gobble up the heat from both a hot CPU as well as a hot GPU.

The main downside of course, is the cost. Your average waterblock may not cost much more than a high-end heatsink, but when you add up the cost of the pump, radiator, tubing, clamps etc it starts to look pretty messy. However, when you keep in mind the savings you can make through overclocking (as illustrated above) and also the simple fact that a watercooling system will provide adequate cooling for many system configurations to come, it turns out to be a worthwhile investment.

Maintenance, on the other hand, is a nightmare. If you want to change your cpu block for example, you can't just unmount it and whack a new one in. Instead, get a towel, lay it down and hope that it catches most of the spill, because it's going to be messy and there's little you can do to avoid it. Hose shut-off clamps help a little in this regard, but there is always going to be some liquid spilling out so you'd better be prepared!

However, spillage isn't as bad a thing as some people make it out to be. As long as you're using a good brand radiator coolant, the electrical conductivity of the coolant will be reduced to the point that you can actually have a block leaking into your system without you even knowing it. I myself had my cpu block leak onto my bare Radeon 9800 card, and it had zero effect. YMMV. And if it does stop working, people have had success by throwing the affected components in the oven at 50C for 30 minutes to an hour or so, to get the coolant out of the tricky spots, and their equipment turned out as good as new. Again, YMMV. Always take the appropriate precautions:

Avoiding catastrophe:
  • always use hose clamps. worm-drive / worm-gear clamps are very reliable and quite cheap too.
  • leaktest the system for up to 24 hours before you turn the computer on. Note that you don't really have to do this outside the system, as long as it is off and unplugged from the wall, no ill effect will be done. Lay down some pieces of paper under various parts so that leaks can be detected easier. I recommend leaktesting inside the system, as it can be a total B#*@H to reinstall a whole watercooling system, all connected up, with only 2 hands.
  • use a radiator coolant additive, pretty much any radiator coolant product at your local store will suffice. Just mix as per the product's instructions. Also, if you like glowy things, pick a coolant that is UV reactive, like what I use (Nulon Ultra Cool) so you don't have to mess around with UV dye later on.
  • avoid mixing metals. ideally, you want everything to be either copper, brass or some form of plastic. while mixing aluminium together with copper in a watercooling system is not as bad as some people make it out to be, it creates a battery effect which speeds up an ugly process known as galvanic corrosion. radiator coolants help greatly in this regard, but still, avoid mixing metals where you can.
  • make sure your system is properly bled, i.e. air-free. with a reservoir, this job is done automatically. with a closed-loop system, you have to get it right the first time or you're in for a noisy pump and hampered system performance.

Common questions

OK, with that out of the way i'd like to have a section where answers to common questions can be read right off, so here goes.

Does the order in which my system is plugged together matter?

NO. There are too many people who believe in the myth that cold water goes into the waterblock, and comes out as hot water, where it is cooled by the radiator and comes out again as cold water. This is WRONG. The flow rate of any respectable watercooling system is such that the water never changes temperature significantly in the time it spends in any one item. As such, the temperature differential between the hottest and coolest parts of the loop is next to ZERO, if not no more than a single degree celsius.

In fact, the ONLY order rule that exists is that the reservoir (if you have one in your loop) must be immediately before the pump.

So, when you're plugging up your watercooling system, the only thing you need to keep in mind is the physical bending of the tubing, which is explained in the answer to the next question:

Is it OK to use 90 degree elbows in my loop?

Sometimes this is unavoidable. An example would be the tubing leading to/from the GPU waterblock, which can get tight especially in SFF or HTPC cases. However, if you can avoid it, you most definitely should! Generally speaking, watercooling performance depends on the flow rate of the coolant in the system. Elbows and really tight turns hurt flow rate more than you'd think. So when plugging your system together, pay attention to achieving the order which involves the cleanest, gentlest tube routing as this will afford you the greatest possible flow rate and thus performance.

One of the worst places to put a 90 degree elbow is right at the inlet of the pump. A water restriction right before the pump means the pump has to work harder to move the same amount of water, possibly causing it to run less efficiently, as well as at a higher noise level. The same goes with any other restrictions, like tight bends or lengths of smaller diameter tubing. Whatever the case – keep these as far away from the pump inlet as possible!

Do I need a reservoir? What are the alternatives?

You do not need a reservoir; however it makes life very easy during the fill-and-bleed process. Here are some alternatives (blue = pump, water flow = in from left):

Image

If a reservoir is too bulky to fit inside your case, or you just plain don't want one, another option is a t-line. This is ridiculously cheap to construct (a dollar, probably less) and essentially acts as a reservoir but is thinner, can be placed almost anywhere in the loop (ideally before the pump) and so doesn't intrude on case space as much as a reservoir. It is, however, harder to fill and bleed, as air bubbles are much more likely to fly straight past the t-line.

Another option is a closed-loop circuit. This has no t-line, no reservoir, none of that. Just the pump, radiator and blocks. How do you fill it? Well, you submerge the pump in a bucket of coolant (don't do this with non-submersible pumps!) and you turn the pump on. You then connect each component in series (starting from the pump's outlet), and give them a shake to get rid of the bubbles, until you get back to the pump and you make the last connection and that's that. While it can be a bit tricky (it's all done submerged!) it guarantees that there is no air in your loop (provided you've done it properly) and evaporation is not an issue at all.

A more complex solution is the 3-tap system, as is used by Swiftech in their fill-and-bleed kits. The coloured spheres in the diagram above are valves, and they let the user shut off water flow at the turn of a knob. During normal operation, the yellow and red valves are closed, whereas the green one is open, allowing water to flow through from left to right. When filling the system, the green valve is closed, and the yellow and red valves are open. Hoses are connected to the yellow and red valves, with their other ends submerged in a bottle/tank of coolant. The pump then pulls water through the red valve so it goes around the system and out the yellow valve. Once the bubbles are out of the system, the red valve is shut off, then the yellow valve is shut off, and then the green valve is opened.

I've heard that a reservoir will give me better performance, is this true?

No, this is not true. In fact, the opposite is true. Let me explain why.

A lot of people think that the increased coolant volume a reservoir provides affords them lower temperatures, but this is false. The simple truth is that the system will take longer to reach equilibrium, i.e. for the water temperature to stabilise, because there is more coolant to heat up. Readings taken before this equilibrium is reached are misleading and often incorrect.

In fact, a closed-loop system gives better performance. Why? In experiments done by numerous watercooling enthusiasts (including Aussie legend Stew Forster, a.k.a. "Cathar") it was found that in a reservoir system, the pump was burdened by the task of having to accelerate still water. In a closed-loop system, the water entering the pump is already travelling at speed. Thus, the flow-rate in a closed-loop system is higher, enhancing the performance of the waterblocks and thus the overall system.

Which way should I have the fans setup, sucking through the radiator or blowing onto it?

Sucking is quieter, gives better performance, and is often more space-efficient than blowing, as is explained in the "Shrouds" section below. Sometimes, however, your case configuration forces you to use a 'blow' setup, but it's not that bad, so there's no need to cry over it.

The concept that sucking air through a radiator is more efficient takes some getting used to, especially since aircooling usually involves blowing air onto a heatsink. The critical thing to remember is, we’re not blowing air onto the object (as we do with heatsinks) but instead, we’re trying to pull air through it. With axial fans, pulling/sucking is the most efficient way to do so.

Why should I bother getting a better pump? How will higher flow rates benefit my system?

You always want a greater flow rate in your system. Higher flow rates create more turbulence, which leads to more efficient heat transfer. Higher flow rates also help to break through what is known as a ‘stagnant’ layer of water between the coolant flow and the surfaces of the waterblocks themselves. This is especially true for impingement blocks which rely on jet impingement effects to effectively transfer heat from the heat source.

However, this does NOT mean that you should just go and get the best pump out there. Pumps dump a significant amount of heat into the loop, and more powerful pumps dump more heat than less powerful pumps of the same type/brand. For open impeller centrifugal pumps like the Eheims, a doubling of pumping power (e.g. Eheim 1048 -> Eheim 1250) will only increase the real-world flow rate by about 25%, which in turn only improves the performance of the waterblocks by around 0.01 C/W (temperature rise (in Celsius) per power unit (Watt) produced by the CPU). On a CPU in the 100W region (i.e. a very hot overclocked CPU) the temperature improvement would be 0.01 x 100 = 1 degree Celsius. However, when you add the extra pump heat into the equation this benefit is greatly nullified, especially if your radiator setup isn’t particularly top-notch.


Component choice advice

Now onto what is undoubtedly a touchy subject - choosing components. Instead of suggesting particular components, I'm going to try to keep things as general as possible, keeping in mind the fact that silence is our primary focus.

Waterblocks

Unless you are aiming for major overclocks, you do not need a fancy waterblock. However, this does not mean that because you don't have the best pump, you can't get so-and-so high performance waterblock that performs best with high performance pumps. What a lot of people don't realise is that these high performance waterblocks outperform the average waterblocks even at low flow rates. So if cost is not an option, get the best block you can buy.

Radiators

Stay away from retail watercooling radiator offerings, such as the Black Ice series. These may perform okay-ish, but the reason I suggest you stay away from these is the price. These items are SEVERELY marked up and are out there to earn an easy buck (or a hundred) out of unknowing consumers. The fact is, you can buy a BETTER radiator for less than HALF the price at your local radiator repair place / wreckers.

A lot of people simply overlook the fact that once you watercool your computer, you essentially have the same system in your case as you do in your car. Coolant is pumped through a hot thing (engine, waterblock) and is cooled by a radiator with fans blowing onto it. Now, a car engine radiator is obviously too big for a casual watercooling enthusiast, so what you want is the car's heatercore. This is essentially a smaller radiator that is mounted in the car, just behind the dash, and is plumbed in parallel with the car's big radiator. When you turn your car's heater on, it turns on the fans that blow air through this heatercore and into your car's cabin. (A neat trick is, if your engine is overheating, simply turn your heater on at full blast and your engine will thank you.)

Also, most heatercores have denser fins than commercial watercooling radiators, and so offer better performance for the same core area.

OK, so you’ve decided on a heatercore, you have a few radiator repair places jotted down, it’s time to head out and meet some quite interesting people :) Now the heatercores that come from cars are often suited to tubing larger than you want, for example the popular Camry sv21 (a.k.a. “Big Arseâ€
Last edited by chylld on Mon Oct 29, 2007 12:30 am, edited 11 times in total.

caxis0
Posts: 55
Joined: Sun Apr 25, 2004 4:20 am
Location: Little Rock, AR, USA
Contact:

Post by caxis0 » Wed May 12, 2004 1:29 am

Great post man :) It reaffirms all that I have come to believe over my past couple of months of research. I just wish you'd have posted it then and saved me all the trouble!

Sklug
Posts: 34
Joined: Tue May 11, 2004 8:53 pm

Post by Sklug » Wed May 12, 2004 8:19 am

Very helpful post.

One thing that could be added is about heater cores. You say that they're cheap, etc. But is there a particular one that's easiest to use for watercooling? And do they come with all the proper connections to attach to 1/2" tubing? Or do you have to have them modified somehow. Someone should write a quick guide on how to build a shroud as well. These are both steps that have been left rather mysterious in my research.

In any case, great start for a guide!

afrost
Posts: 141
Joined: Thu May 01, 2003 9:42 am

Post by afrost » Wed May 12, 2004 12:34 pm

Dangerden has some nice heatercores that come clean and have fittings on them already. They also have some plexi shrouds for them.

Goat_guy
Posts: 24
Joined: Wed Apr 28, 2004 4:23 pm
Location: Rochester NY
Contact:

Post by Goat_guy » Wed May 12, 2004 1:26 pm

How about adding a part on adapters for different tubing sizes? Such as making a 3\8 Heater Core work with 1\2 tubing...

Brian

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Wed May 12, 2004 2:30 pm

all added, look under the radiators and shrouds sections.

shathal
Posts: 1083
Joined: Wed Apr 14, 2004 11:36 am
Location: Reading, UK

Post by shathal » Wed May 12, 2004 3:07 pm

NICE work Chylld - grand stuff.

Might even convince me to get "wet", if the noise of my 3.4 GHz Prescott is too much. Hmmm - now, to wait for someone to confirm that the Zaalman Reservator CAN take on a 3.4 GHz CPU and (preferrably) an ATI 800XT Pro // NVIDIA 6800 Ultra in ONE :D.

HammerSandwich
*Lifetime Patron*
Posts: 1288
Joined: Sat Oct 25, 2003 3:21 pm
Location: 15143, USA
Contact:

Re: Chylld's guide to watercooling

Post by HammerSandwich » Wed May 12, 2004 6:30 pm

Good work, chylld.
chylld wrote:Disadvantages:
  • cost
  • maintenance
  • risk
Add packaging.
chylld wrote:This gives you lower temperatures, but what's the use of this if your computer is already stable? Because it allows you to overclock your processor.
And/or slow your fans. In fact, watercooling systems can significantly reduce the number of fans in a system and provide the same advantages as ducting an air-cooled HSF.
chylld wrote:Adding a new powerful graphics card? No worries, just plug it into the loop, no noise increase whatsoever.
Eeewww, yeah, um, I'm going have to go ahead and sort of disagree with you there. Adding anything to a loop increases the demands on the aircooling side of the system (i.e. radiator and fan). Either you increase airflow or the water temp goes up. When water temp increases, so does CPU temp. With a good setup, this isn't a showstopper, but no one should approach this blindly.
chylld wrote:Also, most heatercores have denser fins than commercial watercooling radiators, and so offer better performance for the same core area.
Are denser fins worse for noise? I can see arguments both ways but know of no testing.
chylld wrote:Barbs can be bought from your local plumbing supply shop for a few bucks a piece. When selecting one, just try to match the thread diameter of the barb to the diameter of the stock tubing on the heatercore, and the tail end of the barb to whichever tubing you intend to use.
You want brass barbs!

Finally, I'm not sure where to put this note, but it's an important one. Whatever you do, ensure that the pump's inlet is not obstructed! If you need 90s in the loop, put them after the pump. If you need smaller-diameter tubing somewhere, put it after the pump. Etc. Restricting the pump's intake can add noise and seriously hurts the pump's, well, pumping.

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Wed May 12, 2004 7:37 pm

thanks HammerSandwich, original post updated under 90 degree elbow and radiator sections.

Sklug
Posts: 34
Joined: Tue May 11, 2004 8:53 pm

Post by Sklug » Wed May 12, 2004 8:47 pm

You recommend an Eheim 1048 pump, which seems to be the overwhelmingly trusted choice. Is it true that it only works with 3/8"ID fittings? If you want to use 1/2" do you have to use some kind of adapter?

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Wed May 12, 2004 8:51 pm

the eheim 1048 pump takes 3/8" BSPM fittings for both inlet and outlet. the barbs that it comes with have 1/2" tails on the other end and so are suitable for use with 1/2" tubing right away.

optionally you can buy brass barbs for the pump, you want 3/8" BSPM x 1/2" T fittings. for the eheim 1250, the outlet is the same but the inlet requires 1/2" BSPM threaded barbs.

Sklug
Posts: 34
Joined: Tue May 11, 2004 8:53 pm

Post by Sklug » Thu May 13, 2004 7:54 am

I think it would be good to add elaboration on how to build a T-line and how to use that to fill and bleed the system. Sounds like a better way to go than a reservoir to me. Also, what do you think about devices like this?

http://store.pcpowerzone.com/swfbfikit.html

Looks like it introduces a lot of angles in the plumbing, but it could conceivably give you a "sealed" system that's easier to fill/bleed.

jk1
Posts: 15
Joined: Tue Jan 13, 2004 10:20 am

Post by jk1 » Thu May 13, 2004 10:05 am

Radiators

Stay away from retail watercooling radiator offerings, such as the Black Ice series. These may perform okay-ish, but the reason I suggest you stay away from these is the price. These items are SEVERELY marked up and are out there to earn an easy buck (or a hundred) out of unknowing consumers. The fact is, you can buy a BETTER radiator for less than HALF the price at your local radiator repair place / wreckers.
I've read this several places, but always have the same question: what if I want an a radiator that fits where a 120mm fan would? This is one advantage of the BIX in my mind. Plus, I think that BIX radiators look better than any heatercore that you can buy. So if you want a good-looking setup without making mods to your case and price is no consideration, I think BIX radiators aren't that bad of a deal...

HaloJones
Posts: 109
Joined: Mon Nov 10, 2003 9:09 am
Location: London, England

Post by HaloJones » Thu May 13, 2004 12:10 pm

I think a BIX is a good radiator but I agree it is over-priced. Second-hand they are worth a look.

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Thu May 13, 2004 2:32 pm

jk1 wrote:what if I want an a radiator that fits where a 120mm fan would? This is one advantage of the BIX in my mind. Plus, I think that BIX radiators look better than any heatercore that you can buy.
that's what i thought. when I bought a BIX i expected it to fit nicely at the back of my BQE, however it didn't - it bulged out at the sides too much. I had to modify it to get it to fit, and that wasn't fun at all.

a quick spray paint job on an automotive heatercore works wonders.

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Thu May 13, 2004 3:09 pm

Sklug wrote:I think it would be good to add elaboration on how to build a T-line and how to use that to fill and bleed the system. Sounds like a better way to go than a reservoir to me. Also, what do you think about devices like this? (...)
Updated the main post with a diagram and info about the 3-tap system as used by swiftech.

That one that you linked to actually isn't too restrictive since the water flows around a big loop and back again. However, in a weird move, swiftech have released a new version of their fill-and-bleed kit that makes the water go through 2 90-degree bends. This possibly makes it easier to fill and bleed but it adds unnecessary restriction to the loop.

Personally i don't trust valves to keep my system watertight, but if it works for you then all the better :)

shathal
Posts: 1083
Joined: Wed Apr 14, 2004 11:36 am
Location: Reading, UK

Post by shathal » Fri May 14, 2004 1:46 am

Hmm - so here's a question I have.

What if "the worst" happens (OK, so "the worst" is the CPU frying, but bear with me) and you *DO* get algae in your setup? How do you get RID of it? *CAN* you get rid of it?

I don't quite understand how algae can GROW in there, there's a very limited quantity of water, thus "nourishment" for the algae, and they're not going to be seeing a whole world of light, which they need if memory serves (photosynthesis, and all that). :).

So - two questions really.

1 - (more importantly) - how can you get RID of the critters (if at all, I imagine it to be QUITE painful), if you get them?

2 - Just out of interest - how do they "live" and survive in there? I doubt most water-cooling rigs make a good environment to habitate, but then life has a way of finding a way... :)

pdf27
*Lifetime Patron*
Posts: 264
Joined: Wed Nov 12, 2003 12:59 pm
Contact:

Post by pdf27 » Fri May 14, 2004 2:14 am


chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Fri May 14, 2004 2:51 am

i've heard of people flushing their system with alcohols, but i'm not so sure if this is a good idea because some plastics don't take too well to alcohol.

if i were to deal with an algae problem i'd probably flush the system with distilled water, or possibly even soda water (which is a common thing used to clean the internals of radiators) and then topping it up with fresh coolant which includes an anti-algae additive.

shathal
Posts: 1083
Joined: Wed Apr 14, 2004 11:36 am
Location: Reading, UK

Post by shathal » Fri May 14, 2004 3:55 am

Great find PDF27 - so it *IS* possible to get rid of it.

So - other than algae and galvanic corrosion - are there other mortal enemies (other than obviously - leakage) of getting wet? Or is that "it"?

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Fri May 14, 2004 4:40 am

well if you look at the disadvantages i listed above in the first post, the 2 other big ones are cost and packaging. i.e., it's gonna cost you a heckuva lot more than aircooling (duh) and it can be a challenge to fit everything neatly into a case.

luminous
Patron of SPCR
Posts: 717
Joined: Sat Oct 04, 2003 6:31 am
Location: UK

Post by luminous » Sat May 15, 2004 2:20 am

I'm very new to watercooling - ie I haven't even built a single watercooled PC yet (or played with one either).

Firstly - excellent work, the watercooling has needed a sticky for a while, and you were the one to sort that out :) I've a few points/ questions so here goes:

Tubing :
You mention ID and OD. Total newbies (ie me a couple of weeks ago) will not know that they stand for Internal Diam and External Diam
Also you mention there are different sizes for different purposes. It may be useful to mention what an "average" setup needs - 1/2" ID ?

Pumps:
Pressue head and flow rate. Errmmm, err panic :) I don't undertand the difference in these values. Afaik a pump with a large pressure head should have a high flow rate. I know that you are going to say this is wrong, cos it is. Do you have a link that points to a good explanation that explains this? (add the link to the sticky if you do)

Fan direction:
All air coolers are used to the fact that blowing air onto a HSF is the most efficient way. The concept of sucking air may not be well understood in terms of watercooling. I'm guessing that the reason is due to the size of the shroud you need with the different methods.

Closed loop flow temps v reservoir flow rates:
You mention that the order of the water blocks in a system can be attached in any order as the temp of the water is basically constant throughout the system. I've read a few topics on this and agree with you. You then mention that the addition of a reservoir can hurt your flow rate.

My point is this; if the flow rates are good enough for the entire loop to have an almost constant temperature you don't really care what your flow rate is. The various thermal interfaces that you have in the system depend on the difference in temps that they see. If the theory that there is neglible difference in the water temp throughout the system holds true, then you may as well always use a reservoir.

Like I say I am a newbie, to me it just seems that the "constant water temp argument" (which I think has been proved) seems to be at odds with the "removing the reservoir will improve performance argument". Perhaps one of you would be able to clear this up for me??

Please don't take any of this as a critiscm. I've learned loads from the people in this forum, and also this article. I'm just pointing out the areas where I just don't quite "get it" :)

pdf27
*Lifetime Patron*
Posts: 264
Joined: Wed Nov 12, 2003 12:59 pm
Contact:

Post by pdf27 » Sat May 15, 2004 5:43 am

I'm totally theoretical as well, but this may help...

Also you mention there are different sizes for different purposes. It may be useful to mention what an "average" setup needs - 1/2" ID ?
1/2"ID is about as big as tubing realistically gets. If you're after overclocking or cooling a very hot PC then you'll need it. Otherwise you may be able to get away with smaller tubing - basically you're making a trade-off between how well it cools and the difficulty of fitting everything inside the case.

Pressue head and flow rate. Errmmm, err panic I don't undertand the difference in these values. Afaik a pump with a large pressure head should have a high flow rate.
Yep. There are two effects happening here. Firstly, a pump will have a pressure/flow curve (e.g. the P-Q curves here: http://www.procooling.com/articles/html ... tus__1.php ). Secondly, the watercooling setup itself will have a similar curve - after all, if you want to push the water around faster, you need to push harder. Where the two curves meet will be the operating point of the system. Hence, a pump with a higher max head will generally be able to push more water around a given system.

The concept of sucking air may not be well understood in terms of watercooling. I'm guessing that the reason is due to the size of the shroud you need with the different methods.
What's basically happening is that you're moving air through from side to side of the radiator - and due to the water inside the whole thing is at a constant temperature. When expanding, and air flow seperates easily - so a fan blowing onto a radiator will really just be blowing a jet of air over a small area of the radiator and there will be stagnant air everywhere else. When sucking, there is no flow seperation so the fan is taking air from over the whole radiator, not just a small jet in the middle.

My point is this; if the flow rates are good enough for the entire loop to have an almost constant temperature you don't really care what your flow rate is. The various thermal interfaces that you have in the system depend on the difference in temps that they see. If the theory that there is neglible difference in the water temp throughout the system holds true, then you may as well always use a reservoir.
You're missing an important point about watercooling here. The interface between the water and the block itself is far from perfect. What you actually get inside a block at low flow rate is a boundary layer of stagnant water over the surface of the metal inside the waterblock. This acts as an insulator between the bulk water flow and the metal of the block. Higher flow rates/turbulence within the block act to break up this layer and so reduce the chip temperature without any change in the water temperature.

peteamer
*Lifetime Patron*
Posts: 1740
Joined: Sun Dec 21, 2003 11:24 am
Location: 'Sunny' Cornwall U.K.

Post by peteamer » Sat May 15, 2004 5:48 am

Very nicely done chylld, thats completely sweet. 8) 8)


Pete

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sat May 15, 2004 6:20 am

luminous wrote:Pumps:
Pressue head and flow rate. Errmmm, err panic :) I don't undertand the difference in these values. Afaik a pump with a large pressure head should have a high flow rate. I know that you are going to say this is wrong, cos it is. Do you have a link that points to a good explanation that explains this? (add the link to the sticky if you do)
Think of the difference between a fan and an air compressor. An air compressor uses valves and a cylinder to create what is called positive displacement. Using positive displacement an air compressor using a strong motor can create huge pressures - into the tens and hundreds of thousands of pounds per square inch - all without creating much flow. Conversly, a fan can create huge flow rates - 300 miles per hour - but by comparison, not much pressure.

The same is true regarding different water pump designs.

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Sat May 15, 2004 7:30 am

  • moved 'fan direction' to common questions section and elaborated on it a bit
  • added a new section on downsides of more powerful pumps and flow rate stuffs (last question in 'common questions' section)

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Sat May 15, 2004 7:33 am

luminous, i think you may be getting a bit confused about why i said a reservoir hurts performance.

essentially, imagine a system running in a closed loop at a steady performance level. now suddenly add a reservoir, and you can see how the flow rate will drop. this drop will reduce the performance of the waterblocks, just like any other restriction would.

Gooserider
Posts: 587
Joined: Fri Aug 01, 2003 10:45 pm
Location: North Billerica, MA, USA
Contact:

Post by Gooserider » Sat May 15, 2004 7:29 pm

Very nice job, however I would make a few additional points...

1. You don't ALWAYS want brass barbs - they can cause problems of stripping and cracking when used with plastic parts, especially if putting a male brass part into a female plastic part. A nylon or plastic barb can be a bit less stressfull. My personal rule of thumb is to use similar sorts of materials on threaded connections if possible (i.e. plastic to plastic, brass to brass)

2. ALWAYS use some form of thread sealant on threaded connections that use tapered threads (I.E the 'NPT' (National Pipe Thread) threads used in the US) Either teflon tape properly applied, or 'plumbers dope' - some folks use both.

3. Brass to brass connections can have a fairly heavy torque applied to them safely, but be VERY gentle when tightening plastic connections, especially if the female connection is plastic. Use plenty of sealant, then go no more than 1/8 to 1/2 turn past finger tight. Otherwise you risk cracking the plastic part, either now or at some point in the future when the stressed plastic ages a bit, and getting either a leak or an expensive repair. This is a particular risk with pump bodies and plastic water block tops.

4. There are several formulations of Tygon tubing - what most people want for W/C applications is formulation "R-3603" usually listed as clear laboratory tubing (not vacum tubing)

5. One approach to dealing with the conflict between the high flow volume of 1/2" tubing vs. the easier installation of smaller tubing is to do a series / parallel type setup, with all the flow going into the CPU block via 1/2" ID tube, then splitting to cool the GPU on one branch and the NB, HDD's and other devices on the other with 3/8" ID tubes. (2 x 3/8" ID tubes have about the same flow resistance as 1 x 1/2" ID tube) this works particularly well with blocks that have dual outlets such as the WW and clones.

6. If possible, a single pass heater core is a better choice than a dual pass core, as the single pass core will only have about 1/4th the flow resistance as a dual pass core the same size. In the US at least, the most popular single pass core is the FedCo 2-342 or equivalent.

May think of more to add later, but this may help.

Gooserider

chylld
Posts: 1413
Joined: Fri Oct 10, 2003 4:45 pm
Location: Sydney, Australia
Contact:

Post by chylld » Sat May 15, 2004 11:46 pm

heya goose,

added tygon tubing spec to the original post.

i acknowledge the advice on the barbs, what you say is very true, but i think it's a bit out of scope for a simple guide. same with singlepass/dualpass heatercores, there's more than enough information to digest there already.

i didn't add the flow splitting idea because it's a bit iffy. i like how you didn't try to enforce it, you said it was "one approach", but the problem with it is that it complicates the flow too much. the gpu block is almost always more restrictive than the nb block and so the nb block will end up getting more flow than the gpu block, which is the opposite of what it should be.

perhaps it might be worth starting another thread to deal with the finer aspects of watercooling?

Seal
Posts: 522
Joined: Sat Oct 25, 2003 4:39 am
Location: Uk

Post by Seal » Sun May 16, 2004 3:58 pm

Great article chylld! Only had time to give it a quick read but I think you have most bases covered there. Will read it again later

Post Reply