To combine the original topic and MikeC's potential project, I see the entire issue as the following:

You have several, small (in size) sources of heat in a computer. These components need to be kept at <65 deg C (for the sake of argument) under all loading conditions.

You have one recepticle for that heat - the air in the room. (I'm limiting this summary to that).

The different cooling methods simply offer different ways of moving that heat from the components to the air.

The actual transfer of heat from cooling system to air is usually from metal heatsink/radiator to moving air. Unless you have a v. large heatsink with widely spaced fins, this will be forced movement of the air by fans. More on natual convection later.

Broadly speaking, the bigger the area of metal and the bigger the delta-T between heatsink and air, the more heat will be transferred. And in most cases, we are running out of room to be able to have that area of metal close enough to the components to rely on conduction through a heatsink. So we move onto other ways to move the heat from the CPU/GPU/whatever to a bigger heatsink - here heatpipes and watercooling are actually very similar. Watercooling is more complex but can move the heat further. Both methods can distribute the heat across a large heatsink/rad more efficiently than straight conduction through the heatsink.

And then it's just a matter of balancing noise of air movement, size and distribution of heat across the heatsink or radiator. Clearly a big radiator which is mostly near room temp but has one hot spot isn't terribly effective for its size... And equally lots and lots of airflow in the wrong place doesn't help much...

Now the ultimate ideal (from an SPCR point of view) is to go completely fanless and rely on natural convection + radiation. For this you want a big, hot heatsink with widely spaced fins arranged vertically, possibly arranged as a chimney. Clearly you are never going to get this inside a computer case next to the motherboard - you have to move the heat to it using some method that's more efficient than conduction through metal... and that's going to either be heatpipes or watercooling to move the heat to where you want it.

So you see? Mike's project is actually much closer in concept to something like the Reserator1 than to traditional CPU/GPU heatsinks

yes, exactly - watercooling in and of itself isn't very effective because you have to get the heat to the water and then air-cool the water, whereas with air you basically have an infinite supply of it at 18-25C. the benefit is being able to use what amounts to a heatsink dozens of times larger than anything you could fit on a CPU, or even inside any normal PC case if you're brave.
the other interesting/important thing about wc is that it's sort of a network for cooling - only need to cool the rad and you can do that any way you please, instead of having to add more fans to different hot things in your PC.

don't forget though you are also adding another noise source, the pump.

You could very easily add multiple heater cores, but you may need a more powerful and possibly noisier pump.

The other plus heatpipes have over WC, besides silence, is reliability. Heatpipes don't have pumps to fail, tubing to leak, or water to grow science experiments in.

True, but sort of. That's where know-how comes in. You can have an extremely reliable, besides quiet, contentional WCsetup, it just takes know-how, while heatpipes are basically cut and dry, idiot-proof.

MikeC's project related,

Perhaps it would best to have the heatsinks slanted like a "V"? Then the air would be forced by the fins as the air ascended upwards.

This is precisely why WC will not go mainstream for a while yet; on the other hand, heatpipes already are mainstream -- Shuttle has been using them for >2 yrs now.

Re - the V-slant: No such complications are needed IMO. Anything that puts stuff at non-right angles spells work & trouble. In this case, I think the gain would be extremely marginal -- if in fact there are any.

Also, I think one HS is enough to cool both CPU & GPU. It is a huge area of fins -- about 1600 sq in in the fins alone, not counting the spaces in between or the back side. Maybe 2000 sqin total. Compare that to a Zalman 7000... which is ~3000 cmsq.

Hmmmm this thing is not THAT much bigger in cooling area: ~13,000 cmsq. Whatever. I still think with the 3 Nexus fans it'd be very quiet and capable of cooling extremely well.

In water cooling's defence however it is much easier to aquirce as opposed heatpipes and such massive heat sinks. also it is much simpler to put togeter.

Can you ben heatpipes easily? or will you need them kind of prebent.

also what about putting the heatsinks on some stands and have one fan angled at each heatsink and the other go straight doe the center?

ERM ( www.heatsink.com ) has some nice heatsinks...

"You could very easily add multiple heater cores, but you may need a more powerful and possibly noisier pump."

More heatercores will be less work for the pump, just put them in parallel.

For DIYers, WC are a viable option if expensive & complex. Aircooling w/ custom heatpipes is not really a DIY option -- except when we use HS that have heatpipes in them.

A custom heatpipe rig by a DIY requires a much greater commitment to customizing, etc, (like fmah's project) because the industry caters by necessity to large volume orders of the same custom parts -- ie, a heatpipe/hs cooling system integrated into a standard chassis by some one like say Dell. But once it's made, installation should be a cinch & the air/heatpipe solution is basically no maintenance & no wear & tear -- just have to make sure the fan is running. (like every other coolling except passive convection)

AFAIK, current heatpipes can and are bent after manufacturing in straight rods to get them to the right shape. One presume some care & maybe heat needs to be applied for the average DIY to do this safely.

Very easy to check whether any heatpipe damage has been incurred: Inert the rod into just boiled water and check how long it takes before it gets so hot that you must left go. Repeat after bending to check for damage. I am gessing the only thing you could really do is crack it -- and then you're hooped.

Watercooling is of course for people who want to sqeeze the most from there system without much noise. For example, people who get fed up with noise on every one of there systems but won't give up high end performance on say, there gaming rig.

I like to overclock my workstation because i need every mhz i can get, I used to use high end air cooling but I couldn't put up with the noise any longer. It was either loose some performance or some hearing. I realized watercooling was the only option. But on my living room HTPC, watercooling is not the answer. Watercooling has its place, but it's not for all low noise systems.

I also believe it is a bad idea to expect silence from watercooling on your first attempt. It takes experience and no matter how much research you do you'll make a few or a lot of mistakes.

I'm really looking forward to my next upgrade. I setup a watercooling rig six months ago so i could overclock hard and run a silent rig. It didn't work as i had hoped (performance wise its a beast) but i learnt a hell of a lot. I'm sure my next attempt will be extremely succesful.

If people want silence from watercooling i recommend using an external and very very large rad, running two low powered 120mm fans at 5 or lower volts. Large car rads do not require much air flow at all to destroy the performance of those thermochill 120.3 overpriced rads. But i'm talking about rads which can fit 8 120mm fans on and need to be hidden away under a desk.

Without doubt it is the pump that will drive you crazy. I still am yet to find a pump that doesn't piss me off, but like i said, i overclock and what alternative is there other than watercooling. Don't get me wrong, pumps don't make much noise, but its going to be the loudest part of a well done watercooling rig.

hm. how can a pump be hard to silence? you could encase it in a ball of silicone then bury it in a foot of concrete and it wouldn't make any difference to the pump; unless it's both huge and magnetic drive, it should have no problem being cooled entirely by the water going through it.

neologan said it

lol holy 2 year old thread revival batman!

And that was written in october 2004, several degrees before the big heatpipe towers arrived in style.



But the Pentium Extreme Edition seemed to have been the turning point!

In the SPCR review of the Thermalright Ultra 120, Devon Cooke says "And — let's be honest — the Ninja is already good enough to cool the vast majority of systems without increasing system noise." Also, in the concluding finish of the SPCR review of the Scythe Infinity, you can read "Fan @ 7V: At this level, the Nexus fan was inaudible from one meter."


Watercooling is quieter huh? I just built an AMD X2 computer with a Ninja, an ASUS EN7600GT Silent graphics, a Silverstone passive PSU, a Raptor in a mCubed Vertical Silence box and two 250 GB Samsung 5400 rpm HDDs as storage (hanging from silicone strings) in a sound isolated case. It spins a total of two Nexus 120, one Nexus 92 and one Papst 4412FGL, all under 4 volts with dust filters hence running four fans hooked up on a T-balancer (still trimming in). Rest assured it's absolutely super quiet, cross my heart and hope to die.
Just starting a water pump makes more noise unless it's in a dead box. Anything but passive rads are sure to lose out because of the fins that actually can make as much noise as the fan itself on low rpms.

Save something drastic happens I won't touch W/C again, that's for sure. If you have to have the latest bombastic GPU then maybe there are miniscule reasons for using a pump to move that heat a couple of centimeters, but really, come on... if a computer can't be heard while air-cooled then why the **** bother with something hard to install, hard to move, expensive, maintenance demanding and risky instead?

It depends. Many people have samples of Laing DDC that runs quieter than undervolted low-rpm fans. They tend to be 10W models with acrylic top - in fact I am yet to hear anyone with 10W+acrylic model claim his pump is loud (I might have just missed, thou).



Usually people who use water cooling either just like to experiment and tinker, or then they want to cool system that is not coolable silently with air, like powerful gaming rig.

Remember that nothing in WC demands you to use regular radiator. Water is just a way of moving and spreading heat. You could very well like install waterblock under the Ninja and then have it cool the water - and in this configuration you could position the Ninja so that it gets less restricted airflow and/or heat is not dissipated inside the motherboard chamber first.

In addition to that, unlike the fans, pump can be embedded deep inside the chassis to prevent vibration/noise transfer. It's enough for the water tubes to pass through, while with fans they generally need to have fairly open path from outside of the case, through them and then back to outside.

WC is just a way to move heat, like heatpipes, it just can carry a lot more heat and a lot farther, costs more, takes a bit more space and requires more maintenance. Not for everybody, but provides a lot more flexibility and performance for those of us who like to have it.

With enough metal thrown to passive radiator there should be no problems cooling any rig with just a single DDC pump being only source for noise. Naturally you'd need some careful thinking and possibly a lot of metal to do it, but it'd definitely be very quiet.

Joe's recent editorial on Water Cooling - Wheres the Magic? over at Procooling sums up the current state of the industry pretty well I think. My comments on it (and this thread, basically) are in the Procooling forum thread about the editorial.

I'd say there is plently of magic in water cooling with silence as goal. That's one ingredient the procooling guys don't seem to prioritize.

That's not to say W/C by itself isn't appealing anymore, there are so many more variables to air cooling that I don't think it will ever loose the appeal.

You are on the money here. Yes, you can build web surfing machine for Average Joe and make it quiet with air by selecting proper components. But the emphansis is in "proper components". Where water comes to play is when you do not want to set to the proven components, as you want your machine to be more powerful, yet quieter, smaller, cooler-looking etc.

I definitely agree that water cooling is not for everybody and you do not have to even consider it if you are willing to make some compromises (either noise-wise or by component selection). But there is still a lot of ground where there is no going with plain air, and that's the area I want to explore.

An example: these air-cooled heatsinks that can cool quietly almost any CPU are almost always huge. What if you need to transport your rig often? The torque they can produce might be too much for my nerves in such.

It's a niche area, but it exists.

How can it be smaller? That doesn't seem likely to me.

Also, there are simple ways to brace/secure the top of the HS to other parts of the chassis so that it's not just the motherboard taking all the stress. I did this for PC with a tall and heavy Scythe that was shipped all the way from Canada to rural Thailand. No problems.

IMO, the issue is no longer with cooling CPUs quietly. It's vidcards: It is when you have a high end dual vidcard gaming system and you want to still cool them quietly that WC makes any real sense. In fact, unless you are one heck of a modder, there's no way to cool something like a dual X1950 gaming machine quietly w/o going to a high end quiet WC rig. The concentration of heat is simply too high for effective removal from that area without resorting to fairly high airfllow.

Hmmm.... An idea for quiet air cooling of dual monster vidcards: Perhaps something like a 2~3 slot space between the two cards and an integrated tunnel HS structure between them, with a big fan pulling in outside air from the front and another fan blowing it out the back might work.

A Thermochill PA120.3 will cool ANYTHING with low-noise low-CFM fans.

The only problem is, the damn thing costs $125+ USD, due to the fact that it's extremely expensive to have them shipped from the manufacturer in the United Kingdom.

Having just WCed my new vidcard, I agree completely! While systems offer sufficient room for massive CPU heatsinks, the expansion slots are a different matter. The Condor is an early step to the solution, but I'd say GPU cooling is 2-3 years behind CPU HSFs.


That probably qualifies as "high end quiet WC."

A water cooled setup could be significantly silent if the water pump was enclosed in a miniture sound proof chamber within the case. One thing that's not possible with fans.

http://www.sharkacomputers.com/aqcofumejafp.html

Its main purpose is to minimize the EMI output of the pump, but I guess it can muffle sound too.

I did some research and just found out that I'm a distant descendant of Nostradamus. I now predict 2007 to be the year when good GPU coolers shall come fourth and also the year when the GPU manufacturers follow suit and present new cards with lower power dissipation. I saw it my tea this morning. You don't have to take my word for it. Go look in some goat intestines yourself.


I've been fiddling with w/c a lot and in my opinion it's a myth that a single rad can cool even a moderately powerful computer without making it a bit noisy. The folks that keep repeating this probably haven't had the chance to compare their system to a properly built air cooled computer. I have, on several occasions. Not even in the same league. I think I know the ropes in w/c but still ended up with two big rads and a passive array just to get my 4412s under 3 volts and still couldn't beat the air cooled alternatives I've built because of the damn rad fins and pump noise.

I again admit to w/c superiority if you feel needy for a big gaming card. Just so nobody, again, need to write about this exception. I also by first hand experience know that a cleverly conceded passive w/c rig can be mindboggingly quiet. Still only an academical point if the simpler air rig can be made seemingly silent. At least when not playing games but aren't they noisy as well even these days? Most w/c systems don't cool the hdds and the psu and the airflow needed to cool that is more than you need over the average ninja and passive gpu card anyway. Intelligent ducting often nulls adding fans.

Without dissing the potential need or virtues of good w/c designs, could we not agree to stop calling it a short-cut to silence? Stop recommending "noobs" to go for w/c with stuff like a pump placed on some rubber feet and a single rad? It just won't deliver. It's tough to get there. It bothers me that these so called tips continue to float around here at SPCR with nobody raising any objections.

Btw, you w/c paladins would be the ones pissing up the wind if this was discussed in say The Silent Front section instead of here where experienced water-coolers and overclockers still lurk around after dark

W/C delivers that's why there aren't objections. When aiming at silence there just isn't an elegant solution to GPU air cooling and that's where a water rig gets recommended. A single rad will cool non-overclocked very well and very quietly which is what most beginners may be looking for. And yes the pump isolation is certainly an option which would make it virtually silent, something that’s not possible in air cooling. These things are difficult to achieve but possible and most people entering W/C are warned and aware of the difficulty and expenses.

This is BS. Such a system would both cost and sound more than air cooling. I have tried this and I have tried the alternatives, more than 5 different air solutions and countless configs with my w/c, all compared side by side. Getting rid of all that wattage through one rad demands more airflow than a 5V Nexus (with fin turbulence). Air does not. Have you tried or are you just repeating what you've read?

I'm running a single rad now, cooling CPU, GPU and chipset averaging at 37 C and rad Scythe fan at about 7V. I can easily run it at 5V and have it augment as needed with the T-balancer, but the difference in noise is barely noticeable among the two, so I’m leaving it with an initial 7v. Rad is a Swiftech Quiet Power, and the pump is a DDC 10w with acrylic top. Like I said, the system is very quiet, with my WD S16 2500KS giving me the most problem now (idle motor noise not just seeks).

There are a lot of factors to account for, and again, it's not easy, but it's doable. For one, a lot of W/C rigs seem to be done externally which are prone to more noise than internal ones. Rad placement and design is also important, dependent on these factors is the voltage at which you will run the rad fan. Normally the less fins-per-inch the better. Shrouds are amazingly important to eliminate fan dead air noise. Sucking air through rads has been recommended as better than pushing, but in my recent set up pushing has turned out to be quieter. I can only speak for my experience, and I know this has worked very well for me.

I’m curious what w/c set up do you have?

I have re-built my w/c machine countless times. I turned so sick of managing everything that now, after four years of enthusiastic experimenting, I've bought a new computer, completely air-cooled.

The old computer is an AMD 2600+, Radeon 9800, Seagate Super Silencer 400 with Nexus fan, Raptor and two Samsung spinpoint V (5400 rpm) for storage. Blocks and one rad Innovatek, one Black Ice, Eheim 1048, 12/16 mm aquarium tubes. This is squeezed into a G-Tower alu case. It sports one 120 mm blow-out fan in the back and one on top. Every inch of the case is sound isolated with the type of heavy sheets used in cars and machine isolation.

The water line was split using two T-joints. The "straight" line fed cpu block and a rad.
Second loop fed gpu block and rad 2. Almost as much water as the cpu.
A small volume of water passed an outer line, which made many turns but still had to be clamped to adjust flow. This outer loop ran along aluminium profiles of the type used in party tent constructs. The alu profiles were glued, with thermal glue, to big copper plates glued on (the inside of) top, back and bottom of the alu case, heating it up as the water got warmer. It actually worked very fine, lowering temps!
Splitting the water line made a huge improvement in flow. Strangely enough everybody "knows" splitting the line hurts flow, but it's the other way around with a pump like my 1048 because it can manage a very big flow but has poor head pressure.

The Raptor is in a suspended hdd box, but used to be in a home made water cooled sound isolation box. Made for a quiet hdd but water temps had to be kept so low. I tried install a DDC to create a separate hdd cooling line, namely the outer loop along the case. Then the water in the main loop to cpu and gpu could be allowed to go over the ~39 C I needed to keep the Raptor under 44 C. But the DDC noise was almost impossible to live with. I cast my DDC in cell foam and a block of cement and it still sounded more than the rest of the stuff together, ugh. Back to single pump.

The case is dust filtered and together with the rads this made the back pressure too great for the fans to work efficiently at the low speeds I use. The rad fans are sucking air out through the rads at the back and top. I had to use intake fans as well. The rad fans are Nexus 120 @ ~3V. I added a Nexus 92 mm in front and a 4412 bottom intake blowing on the hdds. A paper duct from the back rad made sure a miniscule portion of air is passing the north bridge. No fan goes over 4 V in normal use. That means under 400 rpm. Can go to 5-600 rpm if stressed.

The case feels warm to the touch after half an hour or so. (All w/c systems seem to perform fantastically well the first 30 minutes or an hour.)

The Seasonic sits in its own chamber upstairs and is fed cool air from side vents. Everything is controlled by a T-balancer.

Even though the rad fans are soft mounted with shrouds and everything they still whoosh a little because of the fins. The pump is the most trouble. It makes a noise so hard to dampen. The vibrations of a 1048 are easier even though it shakes like a powerdrill. I suspended it in silicon cords and used soft tubing at inlet and outlet. It can touch nothing solid or everything starts to hum.

Fancy project and lots of tinkering but still not as quiet as my new air rig. Bottom line is that if you can avoid w/c, you should.


Long boring answer, but hey, you asked!

Edited to solve some obvious language problems.

That indeed sounds like tons of tinkering, it's a pitty it didn't turn out as well as you wanted. What would you say is the toughest noise to get rid of? I do realize the whooshing is always a problem and at 3 or 4v Nexus your rads were probably running in "passive" mode since there was probably little to none air flow. That makes me think, you could have tried a completely passive rad, maybe custom made. That's something I might play with on my rig since I've tamed the DDC, the rad whooshing would be the only noise source apart from the HDs.