Watercooling Basics (for silence) - any help appriciated :)

The alternative to direct air cooling

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Sat Apr 24, 2004 9:18 am

As you say , the temperature differential will be negligible , but only if there is a large enough flow and that is exactly the point where I disagree.
I don't think even the 1250 will manage 3lpm through 4 blocks , even if they are of very low restriction.
As for the 1048 , people have got below 3lpm even with 2 moderate blocks.

But , I'm only speculating , every loop is different.

Actually , yes , the water cools the pump , that's more evident with a big pump , the 1048 isn't hot enough to be noticeable.
I have a 1250 , if I leave the PC off , but the pump on , the water temperature rises , because of the heat of the pump.
That's why manufacturers forbid running the pump without water (it will burn within a few minutes) and it is also one of the big disadvantages of getting a big pump (a 1250 will dump ~20Watts in the water).

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

Post by pdf27 » Sat Apr 24, 2004 1:35 pm

At 3 litres per minute, a loop that dumps 200W of heat (i.e. massively overclocked unless you're using Peltiers) will raise the water temperature by about 1 deg C over the course of the loop. This, to me, is negligible.
In terms of what will affect your cooling, you're going to have do be down to around 0.5 litres per minute (8 gph - i.e. practically strangling the pump) before you would even notice the difference.
Try working out the thermodynamics of this sort of thing - it's often quite interesting...

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

Post by Seal » Sat Apr 24, 2004 1:51 pm

toilet i really dont reccomend you to have 3 blocks in your loop! rethink that, every extra block will have severe effect on your cpu cooling effectiveness/efficiency. EG. think of getting zalman passive coolers for your gfx/hd if you can. Thats what i have done. not to mention its a lot less hassle + better IMO + cheaper.

also in uk watercoolings pricey, the only thing that pushes your cost over 300 is the waterblocks, my setup cost just about £180ish with 1 waterblock. mostly from that store and pclincs (which dosent exist anymore i dont think)

Seal

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Sat Apr 24, 2004 3:03 pm

I'll have to admit that when it comes to thermodynamics , I'm just a layman.
Most of my knowledge on watercooling comes through experience/testing.

The way I've seen it , higher waterflow improves the performance of all waterblocks , decreases temperature differences within the loop (even if that is a small gain) and is not too hard to get or silence.

Actually some mates of mine did a test between the 1048 and 1250 in a single block loop (with triple radiator).
The 1250 gave only 20-25% better waterflow (difference increased with increased restriction).
The least restrictive block (they tested with 4 different blocks , including cascade) , gave them a flow of 5.3lpm (1250) vs 4.4lpm (1048) , always including the rad.
That was enough to improve the cpu temperatures by ~3°C on all blocks (cpu ~90Watts).

I would expect the difference to increase , as more waterblocks are added and waterflow decreases (as is demonstrated on the graph in the previous page).
The same should be true as cpu (or any component's) heat increases.

To counter myself : I have seen someone using a Hydor L20 to pump through 4 (low restriction) blocks , at almost 1lpm , the pump overheated severely so he had to use it submerged (to keep it cool) and performance wasn't very good , but it worked.
I'd prefer to not strain the pump so much , since it's the only watercooling part that is prone to damage.

A big pump also has disadvantages.
1) bigger (actually the 1250 is HUGE)
2) hotter (more heat in the water , may be significant if the radiator isn't very good)
3) louder (although it can be silenced like any pump)
4) more expensive

All said , I'd still recommend a big pump , because 3-4 degrees gained are important to me.

And to restate the obvious , things would be much simpler if you just got 2 blocks (cpu & gpu).
The nb doesn't need one (I use one , but my nb is overclocked + overvolted) , neither do the HDDs.

toiletduckuk
Posts: 67
Joined: Wed Apr 21, 2004 3:53 pm
Location: Hull, UK

Post by toiletduckuk » Sat Apr 24, 2004 6:31 pm

Bleh - problem is, I think even slowing down my current NB fan with a fanmate caused overheat reboots. Replacing it with a Zalman isn't going to help its temps, especially if i introduce watercooling so airflow will drop right down... again, passive solution for the GPU seems a bit worrying too with a lot less airflow..

Are three blocks really so bad? It seems ridiculous to bother to set all this up and just have one block :(

Also, surely if my CPU heats up the water that goes to the GPU and NB, I just pump from GPU -> NB -> CPU? Then the hottest part is hit last...

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

Post by chylld » Sat Apr 24, 2004 9:48 pm

just go with an NB block then, you have a very valid reason to go for it. just try to get a low flow resistance one.
toiletduckuk wrote:Also, surely if my CPU heats up the water that goes to the GPU and NB, I just pump from GPU -> NB -> CPU? Then the hottest part is hit last...
order does NOT matter. the order should only be dictated by tubing convenience.

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

Post by pdf27 » Sun Apr 25, 2004 12:47 am

apocalypse80 wrote: Actually some mates of mine did a test between the 1048 and 1250 in a single block loop (with triple radiator).
That was enough to improve the cpu temperatures by ~3°C on all blocks (cpu ~90Watts).
What you're experiencing here is a heat transfer effect due to the block design. What is happening is that the higher flow rate is enough to partially blast away the stagnant layer of water which sits right next to the metal surface. This layer acts like an insulator - so removing it or making it thinner has a major effect on cooling performance. If they'd measured the temperature of the water in the loop it would probably have been higher with the bigger pump.
toiletduckuk wrote:Are three blocks really so bad? It seems ridiculous to bother to set all this up and just have one block
Also, surely if my CPU heats up the water that goes to the GPU and NB, I just pump from GPU -> NB -> CPU? Then the hottest part is hit last...
As mentioned above, it takes so much heat to make water hotter that in your setup you won't be able to notice any difference between the temperatures at the start and end of the loop - it will probably be under 1 degree C.
Three blocks will reduce the flow rate in your system compared to a single block, and increase the cost of the system as a whole. Unless you're going for huge overclocks, the flow rate doesn't really matter all that much. Water cooling will still beat air cooling temperature-wise for even very low flow rates.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Sun Apr 25, 2004 3:18 am

It's generally beneficial to connect the blocks in a hotter to colder series , since hotter components usually require better cooling.
In your case that would be CPU -> GPU -> NB.
However if you need significantly more tubing to achieve it , it'll negate any benefit.

Just go with the shortest route you can get.

As for the nb , I don't think you should worry , you aren't overclocking.
Probably the components that will be most irritated by the lack of airflow will be the motherboard's power regulators (common problem in Prescott systems).

pdf27 , I agree that it's turbulence that improved temps , however that is partly because of the higher flowrate.
They couldn't get any "real" difference in water temperature (the rad was too good) , but I agree , the bigger pump should increase it by a bit.
I actually have the pump directly before the cpu block , cause it allows shorter tubing + the added turbulence gains me 1 degree.

To everyone ; I might seem too much at times , but bear with me , I've been hunting degrees and mhz for way too long....
Last edited by apocalypse80 on Sun Apr 25, 2004 3:22 am, edited 1 time in total.

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

Post by chylld » Sun Apr 25, 2004 3:22 am

apocalypse80 wrote:It's generally beneficial to connect the blocks in a hotter to colder series , since hotter components usually require better cooling.
In your case that would be CPU -> GPU -> NB.
ffs. do you read ANYTHING that we say??? the temperature differential is NEGLIGIBLE, hence the ORDER of the components DOES NOT MATTER!.

get it through your thick skull already, i'm tired of putting up with this *#$t.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Sun Apr 25, 2004 3:23 am

Read my post's 5th row.......

And allow me to clear any misconseption that I may have created.

Higher waterflow will improve cooling , but not hugely , 2-3 degrees on all components is achievable.
But , you'll need a twice as powerful pump to get a 20-30% higher flow (aquarium pumps have low pressure) and the added heat may negate any benefits , if you radiator isn't powerful enough.

The order of components will affect the cooling of some of them by a little bit , but increasing tubing length to do it will decrease cooling efficiency of ALL components.
You should only use a route other than the shortest if you can get a significant benefit (like more turbulence through 1 block) , but you can only see that through testing.

I'm sorry for any inconvenience I might have caused.
It seems that the whole issue was caused because the , not so great , benefits are not so important to most , as they are to me.

snutten
Posts: 341
Joined: Mon Nov 25, 2002 1:27 pm
Location: Sweden

Post by snutten » Sun Apr 25, 2004 3:52 pm

toiletduckuk, it seems you need to sort out your priorities.
Get the big pump for optimal cooling performance.
Get the smaller pump for quieter performance. Don't worry about the number of water blocks. JUST SPLIT THE LINE, saving your pump, because who the f cares if your processors then run 3 degrees C hotter??? (Your HDDs is what limits your water temps anyway. Or are you going to overclock after all?)
The big worry is how to dump all that heat. Things like getting a passive heatsink for the NB relieves your radiators. If you are planning to cool all that stuff with a single 2x120mm heatercore then you can forget about a really quiet machine. The key here is how to dump the heat quietly. The water is just another heatpipe, transferring the heat to a place where it's easier to get rid of it.

You did say you wanted to watercool for SILENCE?

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

Post by HammerSandwich » Mon Apr 26, 2004 3:16 pm

pdf27 wrote:If they'd measured the temperature of the water in the loop it would probably have been higher with the bigger pump.
Isn't that because of the bigger pump's higher wattage? When a CPU is making, say, 100W, doesn't it put 100W into the loop regardless of the block, pump, etc.?

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

Post by chylld » Mon Apr 26, 2004 3:28 pm

HammerSandwich wrote:
pdf27 wrote:If they'd measured the temperature of the water in the loop it would probably have been higher with the bigger pump.
Isn't that because of the bigger pump's higher wattage? When a CPU is making, say, 100W, doesn't it put 100W into the loop regardless of the block, pump, etc.?
not necessarily. this is a complicated situation.

the fact that the pump is of a higher wattage will mean that yes, the pump will dump more heat into the water than a pump of similar design with lower wattage. but what pdf27 is referring to here is the cooling ability of water. with low flow rates, the surface tension is not counteracted as much, thus the water does not take as much heat away from the waterblocks. with a more powerful pump, the water breaks down the surface tension more, and thus takes more heat away from the waterblocks.

one thing to note, is that you don't have to get a more powerful pump to help break the surface tension. you can get radiator coolant from the auto section of most superstores (intended for cars) that increases the 'wetting' ability of water, and reduces the effects of surface tension.

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

Post by HammerSandwich » Mon Apr 26, 2004 5:05 pm

chylld wrote:with low flow rates, the surface tension is not counteracted as much, thus the water does not take as much heat away from the waterblocks. with a more powerful pump, the water breaks down the surface tension more, and thus takes more heat away from the waterblocks.
I believe this is incorrect, chylld. A better block allows the heat to transfer more efficiently and, therefore, at lower temperature differentials. However, even with a thick boundary layer, the same wattage will transfer once the CPU gets sufficiently hotter than the water. A better block keeps the CPU's temperature lower without affecting how much total heat is transferred or, I think, the bulk water temperature. This should be no different than in aircooling, where an SLK900 transfers a given amount of heat better than a retail HSF.
chylld wrote:one thing to note, is that you don't have to get a more powerful pump to help break the surface tension. you can get radiator coolant from the auto section of most superstores (intended for cars) that increases the 'wetting' ability of water, and reduces the effects of surface tension.
And the savvy racers reduce surface tension with a couple drops of dishwashing soap instead of Water Wetter.

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

Post by chylld » Mon Apr 26, 2004 5:08 pm

HammerSandwich wrote:
chylld wrote:with low flow rates, the surface tension is not counteracted as much, thus the water does not take as much heat away from the waterblocks. with a more powerful pump, the water breaks down the surface tension more, and thus takes more heat away from the waterblocks.
I believe this is incorrect, chylld. A better block allows the heat to transfer more efficiently and, therefore, at lower temperature differentials. However, even with a thick boundary layer, the same wattage will transfer once the CPU gets sufficiently hotter than the water. A better block keeps the CPU's temperature lower without affecting how much total heat is transferred or, I think, the bulk water temperature. This should be no different than in aircooling, where an SLK900 transfers a given amount of heat better than a retail HSF.
well pointed out. i guess i'm living proof that you do actually get dumber when you go to university :)

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

Post by HammerSandwich » Mon Apr 26, 2004 5:30 pm

chylld wrote:i guess i'm living proof that you do actually get dumber when you go to university :)
Thank God I dropped out before it was too late! :D

1911user
Posts: 109
Joined: Sat Mar 13, 2004 12:08 pm
Location: Oklahoma

Post by 1911user » Tue Apr 27, 2004 4:33 pm

My opinion FWIW,

I would recommend looking at the swiftech MCP-600 pump instead of the ehiem. It is compact, powered by the computer PSU, and only uses 9 watts of power. It also has a pressure head of about 10 feet! That is more than any of the other aquarium pumps being discussed except the Danner Mag3 which hasn't been mentioned, yet. Pressure head is the key rating for a computer waterpump, gallons-per-hour doesn't mean much.

The whitewater cpu block is good (and you don't have to be an overclocker to benefit from it). The swiftech MCW-5000 or -5002 or the new MCW-6000 are also good. If you want to cool the northbridge, get the lowest flow restriction block which the swiftech MCW-20 (they make good stuff) . For the GPU, the swiftech MCW-50 is a good choice as is the Maze4 gpu block.

For tubing, 3/8ID 1/2OD is much easier to install and will quick-connect to most swiftech blocks. www.swiftnets.com if you haven't been to the swiftech website. If you could keep everything internal to the case, I'd say try 3/8ID. If you have to go external, then I'd use 1/2ID tubing for lower flow resistance. External cooling boxes use much more tubing than internal and tubing flow resistance becomes a bigger factor.

Radiators......what else can be said except use a heatercore because it's cheaper and almost the perfect design for watercooling PCs. Also, use the biggest rad you can fit. More radiator size is an easy way to improve the system cooling without much expense. Some people use 2 rads. Maybe you could fit 2 smaller ones in the case, BUT make sure they can each handle a 120mm fan. Use a good fan shroud for each rad, they really help focus airflow throught the rad. The little 80mm rads won't come close to handling the heat load you'll have; 4 or 5 of them might be enough :D The 80mm rads main advantage is they are small, cute, and much easier to fit into a case without mods or hard work.

As far as running nearly fanless in the case, that is a big job to do right. You'll have to cool the mosfets, southbridge, northbridge, cpu, gpu, psu, hard drive(s), and anything else that gets even warm with normal case airflow. That is a very custom water system and IMHO you aren't ready for that yet. It can be done but is alot of work, usually involving multiple custom waterblocks. This website shows one solution for a fanless PC. http://www.zfz.com/
Last edited by 1911user on Wed Apr 28, 2004 2:22 pm, edited 2 times in total.

toiletduckuk
Posts: 67
Joined: Wed Apr 21, 2004 3:53 pm
Location: Hull, UK

Post by toiletduckuk » Wed Apr 28, 2004 2:05 pm

...wtf is a mosfet when it's at home :)

Looks like my PC isn't going to be completely fanless, I'm running a Zalman with L1A mod now. Is this enough airflow alone to move some air in the case and not melt any other made-up-part-names? ;)

Cheers

Dave

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

Post by chylld » Wed Apr 28, 2004 3:15 pm

i would recommend a case fan, but you'll be able to run it slower that normal.

note that you won't really need the case fan if your radiator is inside your case, since the airflow from the radiator fans will be sufficient to move air around inside (probably)

toiletduckuk
Posts: 67
Joined: Wed Apr 21, 2004 3:53 pm
Location: Hull, UK

Post by toiletduckuk » Wed Apr 28, 2004 3:21 pm

I'm going for a radbox, as I can't fit the dual heatercore I want in my case easily at all :)

Any hints and tips on building one of those would be appriciated though - maybe you should write a step by step guide on yours *nudge* :D

Cheers

Dave

(ps help my out my friend guys, who may type like hes illiterate but isn't really that stupid :D - http://forums.silentpcreview.com/viewtopic.php?t=12271)

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

Post by chylld » Wed Apr 28, 2004 3:49 pm

hehe yeh fitting those things in your case is a nightmare.

if you have maya i'd be happy to send you the source file for the design i did, but it'd be of limited use to you since the parts would be of different sizes

all you have to do though is find out the exact dimensions of all your parts and then build a box around it, there's nothing to it really! the rest is woodwork.

a few mistakes that i made with mine though that you'll want to keep in mind when designing yours:

1) pump vibration not accounted for, the pump being bolted to the box causes the whole box to resonate
2) didn't finish sanding it down properly before painting
3) potentially unsafe - mains electronics lie directly below the hoses, one leak....
4) handle - chosen for looks, found out it's only rated for 4kg. the box weighs around 8kg and the handle's having a cry
5) aesthetics - wasn't even considered, main objective with this box was functionality

1911user
Posts: 109
Joined: Sat Mar 13, 2004 12:08 pm
Location: Oklahoma

Post by 1911user » Wed Apr 28, 2004 7:23 pm

chylld wrote:i would recommend a case fan, but you'll be able to run it slower that normal.

note that you won't really need the case fan if your radiator is inside your case, since the airflow from the radiator fans will be sufficient to move air around inside (probably)
With no case fans (except the psu fan), I'd try to seal a shroud to the rad from the bottom front (air intake) of the case. That would help pull fresh air into the case and rads generally work better with cooler air :D

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

Post by pdf27 » Thu Apr 29, 2004 2:11 am

toiletduckuk wrote:...wtf is a mosfet when it's at home :)
Metal Oxide Semiconductor Field Effect Transistor!

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

Post by luminous » Thu May 06, 2004 5:28 am

I nearly wrote what pdf27 wrote.....

Except that I though someone would come along and ask what they did, and what they looked like. So...ermmm...what do they do, and what do they look like when they are at home ? :)

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

Post by pdf27 » Thu May 06, 2004 3:42 pm

Three legged metal cylinder. One leg is the signal input, the current flows through the other two legs. Think of it as a relay basically - if the signal input is greater (or lower, depending on type) than the critical value then current will be able to flow between the other two legs.
Good grief, I can actually remember first year Linear Circuits!

Post Reply