The theory behind fanless systems?

The forum for non-component-related silent pc discussions.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

The theory behind fanless systems?

Post by stmok » Sun Feb 24, 2008 10:56 pm

Inspired by the link provided in another thread...
http://www.geocities.jp/numano333/index.html

...I'm thinking about modifying an existing system (midtower) to be completely fanless. (use of large heatsinks and heatpipes)

The problem is, how do I calculate the necessary heatsink size for CPU, GPU, HDD, etc? What about the loss of efficiency caused by having a heatpipe between the CPU and large heatsink? Granted, it sounds like I'm turning this into a mechanical engineering problem. For heatsink specs; they give size, weight and Thermal Resistance (C/Watt for 80 °C rise).

Say the plan is to turn the case, cooling, etc (ie: the infrastructure) to be flexible and compatible with desktop components. ie: Today I may have dual-core. Tomorrow, maybe quad? What about a GF9600GT? Is that asking for way too much?

Essentially, I wish to work out the theory, so I can build it without too much trial and error OR experimentation.

I've seen Core 2 Duo's specifications on the Intel site and they offer a TDP. (Although, I'm curious as to what they define as the TDP...Is it max? Or typical use? What's the real value? Should I assume 200W and plan for that?)...What about GPUs?

So far, I'm looking at all sorts of combinations.
* A pre-built fanless case as alternative? (Zalman and mCubed...This gets expensive!)
* Laptop parts? (good for low power, expensive, limited availability compared to desktop parts.)
* Fixed shaped heatpipe vs flexible?
* Location of heatsink? Side panels? On top?
* Attaching copper block to heatsink? Screws with thermal goop? Epoxy with thermal poop? (The screws approach introduces thermal resistance).
* Heatsink attitude? Flat or vertical?

One other question is in regards to "above ambient temp". How do I know how big is this suppose to be? (in reference to CPUs, GPUs, etc).
To see what I mean, look at the Thermal Performance and Temperature Rise Above Ambient part in this link. (scroll down until you see the graph)
http://www.conradheatsinks.com/technical-details.html

Any help, opinions, and suggestions are welcome. :)

Bluefront
*Lifetime Patron*
Posts: 5316
Joined: Sat Jan 18, 2003 2:19 pm
Location: St Louis (county) Missouri USA

Post by Bluefront » Mon Feb 25, 2008 1:49 am

I'm not sure where you're going with this....but. Determining the necessary cooling components in advance of testing, will be difficult/impossible. The first reason being that for most home situations, the ambient temperature is constantly changing.......and without a fan you have no way to compensate.

Which leaves you with only a few options...the main one being to use the biggest/best heatsinks, combined with the least restrictive airflow. After that you keep your fingers crossed that things don't overheat.

Keep in mind that computers are relatively cheap these days, and it's much more cost-effective just to buy a newer/faster computer......rather than design such a thing as you suggest. Of course you could DIY, but not without much experimentation.

Right now a good, quiet setup can approach a totally fanless setup in the noise department, without a whole lot of trouble. But to take the next step and eliminate all fans......very difficult, and probably very expensive.

Atmosper
Posts: 103
Joined: Thu Aug 17, 2006 7:14 am
Location: Västerås, Sweden

Post by Atmosper » Mon Feb 25, 2008 2:03 am

If your thinking that noise-less is the same thing as fan-less I think your making it unnecessarily hard. A couple of very low speed fans are usually completely inaudible, certainly quieter than hard drives. Forced convection (fans) is many times more efficient than plain convection (no fans).

A fanless system will have to be very low powered and, in my opinion, overengineered not to overheat. A fanless solution with heatpipes isn't going to be very easy maintenance or flexible.

To conclude - going fanless is not certain to be quieter and is likely to be much more cumbersome but certainly poses an interesting engineering problem!

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Mon Feb 25, 2008 5:09 am

First ,you need to get a grasp of the science involved in this. For instance:
What about the loss of efficiency caused by having a heatpipe between the CPU and large heatsink
There is no loss of efficiency, there is just additional thermal resistance (and, with heat pipes, an upper thermal limit)
(The screws approach introduces thermal resistance).
:? There is thermal resistance all the way from the CPU to the air, your job is to calculate this and allow for it in the design.
I've seen Core 2 Duo's specifications on the Intel site and they offer a TDP...
Intel's TDP and TS are based on their own reliability tests. If you design the system to meet the specs ( i.e. remove TDP watts while under the TS temperature) the system will be reliable. You could run at a high temperature (upto a maximum specified in the spec sheet), but the life time of the CPU will be reduced and/or the system will become unstable.
One other question is in regards to "above ambient temp"...
All part of the fun :) You can guess-timate that the heat-sink temperature will be somewhat lower than the CPU's TS (the actual figure will depend on the length of the thermal transport system you're using) then calculate the design based on the guess (remember the heat-sink manufacturers figures are measured in 'free air', your design will not be in 'free air' so don't trust them too far :lol: ).

I was going to answer this in full but, I've filled two A4 sheets with explaination so far and am still a way from finishing so, for now, I'll just point you to the worked example of my silentPC project

But, as others have said, the fans are only one part of the noise equation.
Last edited by BillTodd on Mon Feb 25, 2008 5:44 am, edited 1 time in total.

ZircularLogic
Posts: 28
Joined: Sat Feb 16, 2008 3:00 pm
Location: Southeastern USA

Re: The theory behind fanless systems?

Post by ZircularLogic » Mon Feb 25, 2008 5:20 am

stmok wrote: Granted, it sounds like I'm turning this into a mechanical engineering problem. For heatsink specs; they give size, weight and Thermal Resistance (C/Watt for 80 °C rise).
Weeeellll...I'm a mechanical engineer, and I know a mechanical engineering problem when I see one. ;) To me, there are two ways to go at this.

1) The "fun" way: solve the engineering problem and see if you're right! Heat transfer really is a simple science. Heat flows down hill, heat flows through some materials better than others, and heat crosses some boundaries better than others. Everything else is nuance. (OK, "everything else" is a lot.) An indroductory level text book would contain many of the answers to your questions (like heat sink location and orientation, whether or not your screws really matter), and give you the tools to make conservative assumptions about the many questions that can't be answered directly. If your execution matches your analysis, go get yourself a job with a PC manufacturer! And be prepared for iterations.

2) The "easy" way: get yourself a cheap/old laptop to take apart. (They only seem to have one fan, anyway.) Get rid of that fan by kludging some kind of BIG heatsink onto it. Mount the parts in a large case with a really big outlet on top and a really big inlet near the bottom. Connect, stress test, monitor. See what happens. And keep in mind that "easy" is in quotes.

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Mon Feb 25, 2008 5:45 am

Heat flows down hill
:lol: :lol:

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Mon Feb 25, 2008 1:27 pm

Thank you both (BillTodd and ZircularLogic) for the info and links.

In your experience, is there a good intro text you can recommend?

walkingjohn
Posts: 29
Joined: Sun Mar 18, 2007 9:51 am
Location: Ontario, Canada

Post by walkingjohn » Mon Feb 25, 2008 1:34 pm

This is a popular question, so here's a starting point, a fanless computer recipe; modify as needed:

Caveat: I assume you're doing this for fun, not because it's the best way to make a quiet computer (it isn't, unless maybe with an SSD).

Biostar TF7025 mobo, Athlon X2 3600+ (or BE2300)
Ninja on cpu, HR-05 on chipset
Undervolt using rmclock
Stick an emergency fan in the cpu header, use bios to control
Replace case top and sides with mesh
STNF-30 fanless psu, mounted away from mobo

On my machine, the fan never comes on.

Placing HDs, optical drive and PSU is the only tricky bit--keep away from other hot bits.

pipperoni
Posts: 218
Joined: Sun Oct 24, 2004 9:10 pm
Location: Toronto
Contact:

Post by pipperoni » Mon Feb 25, 2008 1:39 pm

I had a go at a completely fanless system. It is a lot of work to say the least and is very challenging to get enough surface area for any serious cooling. Here is a picture of my project, for a heatsink about 16" square with fins about 1" tall, I could only effectively cool a 100W system. The heatsink alone weighed in at something like 25lbs.

As I lose interest in projects I have completed half way I never did finish any professional looking case.

The modding work required also started to get a bit silly. At one point I desoldered all the heatsinks from my powersupply, removed all the big MOSFETs, isolated them and reattached them to different heatsinking.

I hope this gives you some idea of scale, good luck! :D

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Tue Feb 26, 2008 12:24 pm

Hmmm...OK.

16" square => 40.64 cm^2
1" tall => 2.54cm
25lbs => About 11.36kg
100W cooling capacity.

The heatsinks I'm looking at are:
http://www.conradheatsinks.com/products ... .html#MF35

4.8cm tall
35cm length
15.1cm width
Weight: 2.46kg
Thermal Resistance (C/Watt for 80 °C rise) => 0.21

I guess I need to punch in some numbers on a spreadsheet to get a feel of heatsink and CPU.

Core 2 Quad Q9450
TDP: 95W
Thermal Spec: 71.4 deg/C

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Tue Feb 26, 2008 4:42 pm

ok, so, you've 95 watts to get rid of, at less than 71'c at the source.

Assume 30'c ambient inside the case, you need to dissipate 94w with less than 41'c drop

41/94 = 0.43 K/w absolute maximum between the cpu and the air. this has to include the thermal resistance of each interface, the heat-sink itself and anything else in the thermal transport system (i.e. heatpipes, contact blocks etc.)

Also be aware that the spec for heat-sinks assumes the whole interface surface is heated. In the real world, the heat source is usually small compared to the heat-sink, so you have to allow for the conductivity of the heat-sink material and de-rate it appropriately (this is why the above 16" square sink caused the guy so many problems - it would need the heat to be spread over the surface with heatpipes or other high conductivity devices).

And another thing; Heat-sinks need to be de-rated if running at a lower temperature you will not be running at 80'c.

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Tue Feb 26, 2008 10:41 pm

I see. Hmmm...Yeah, I see they have a correction factor, "K" with the graph.

I did some number crunching with selected CPUs.

Code: Select all

				        	            Q9450      QX9775    E8500      A64 2.7Ghz   T9500														
Heatsink Perf.			      TDP	95W		  150W		65W		  65W		    35W
R@80deg/C = 0.21		   Thermal	71.4deg/C  63deg/C   72.4deg/C  68deg/C 	  100deg/C
																			
Ambient (Deg/C)	K	   R (C/W)	Req. R	  Req. R    Req. R     Req. R	    Req. R		
20	             1.5	 0.315	  0.541		0.287	  0.806	   0.738		  2.286		
25	             1.4	 0.294	  0.488		0.253	  0.729	   0.662		  2.143		
30	             1.33	0.279	  0.436		0.220	  0.652	   0.585		  2.000		
35	             1.29	0.271	  0.383		0.187	  0.575	   0.508		  1.857		
40	             1.24	0.260	  0.331		0.153	  0.498	   0.431		  1.714		
45	             1.2	 0.252	  0.278		0.120	  0.422	   0.354		  1.571		
50	             1.16	0.244	  0.225		0.087	  0.345	   0.277		  1.429		
55	             1.12	0.235	  0.173		0.053	  0.268	   0.200		  1.286		
60	             1.1	 0.231	  0.120		0.020	  0.191	   0.123		  1.143		
65	             1.07	0.225	  0.067	  -0.013	  0.114	   0.046		  1.000		
70	             1.04	0.218	  0.015	  -0.047	  0.037	  -0.031		  0.857
This is direct Heatsink to CPU comparison.

Its a theoretical comparison of the thermal resistance a heatsink (no fan) is capable of handling (R (C/W)).
VS
The resistance required for the CPU, (Req. R).

As commented by BillTodd, it does NOT account for copper blocks, heatpipes, etc sitting in between. As well, it assumes a heatsink is uniformly heated. (In real life, this isn't true).

If I'm reading this correctly, the Req. R needs to be bigger than the heatsink's R (C/W) for this to be feasible (in theory).

If that's the case...

Q9450
=> Heatsink can handle it up to 40 to 45 deg/C above ambient.

QX9775
=> Don't even bother with a single heatsink...Maybe have 4x heatsinks and heatpipes?

E8500
=> Heatsink can handle it up to 50 to 55 deg/C above ambient.

A64 2.7Ghz
=> Heatsink can handle it up to 45 to 50 deg/C above ambient.

T9500
=> Should be totally fine. It is a mobile CPU, isn't it? :) But its costly. :(

I guess I need to read more...That Wakefield Engineering document looks interesting.

As mentioned, the heatsink is NOT uniformly heated as seen by many DIY designs. (And pipperoni's design appears as an example of the effect of this...Limited to 100W).

What about multiple heatpipes and multiple blocks in equal spacing, arranged in a crucifix configuration on the back of the heatsink?

I read somewhere one should do that, to try to distribute the heat to the heatsink in a more even manner.

Say you have a copper block on the CPU, and that has 5 heatpipes (say 6mm ones). On the other ends, have one in the centre of the large heatsink and the other 4 in a North, East, South, West arrangement.

What do you think?

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Wed Feb 27, 2008 8:06 am

=> Don't even bother with a single heatsink...Maybe have 4x heatsinks and heatpipes?
You can use a single heatsink, if you choose one optimised for point source use:

e.g. http://www.aavidthermalloy.com/cgi-bin/ ... Reff=0.411

[This is only an example, there are profiles that will give you down to 0.3 K/W in a (roughly) 5" cube.]

What about multiple heatpipes and multiple blocks in equal spacing, ...
Can work well (albeit at a price) - see Zalman's TNN case

Watch-out for the Thermal Res of the heat-pipe (6mm sintered pipe: say 0.15 - 0.2 K/W), it may be worse than the TR of the heat-sink;

You can calculate the thermal resistance fom the materials conductivity, the cross sectional area and the length of the heat path.

Aluminium has a conductivity of about (varies with alloy) 200 W/(m.K) (i.e. a cube [of any size] will pass 200W with a drop of 1K [1'C] = 0.005 K/W)

I read somewhere one should do that, to try to distribute the heat to the heatsink in a more even manner.
Exactly. Try to think of the path the heat energy takes, from source to the air (the air is the real heat sink), everything along that path will have a thermal resistance, so even when the heat enters the metal heat sink it has to travel along the metal until it gets to the surface, where it meets the air and gets (*slowly) removed.


* air is a really bad conductor and has a very low heat capacity, only molecules close to the surface of the heat-sink will be heated, so to remove any sensible amount of heat, fresh cool air molecules need to replace the heated ones. The only force moving the air in convection is gravity (because warm air is less dense, it gets pushed up by the heavier ones above) so it's vital to keep the air flow path as unrestricted as possible.

Plekto
Posts: 398
Joined: Tue Feb 19, 2008 2:08 pm
Location: Los Angeles

Post by Plekto » Wed Feb 27, 2008 3:47 pm

Still, it is possible to build a system with only one fan and it not be an engineering nightmare. I highly recommend this approach, as even 10fcm of forced air is orders of magnitude better than nothing at all.

As for noise, typical ambient noise in a house is between 20 and 25db at its quietest, and maybe 15 in a rural area where it's "dead silent". So a 20db system will be "silent".(a well known trick Apple uses for instance. One fan and lots and lots of air holes. One fan can easily stay under this threshold and offer excellent cooling if you do your homework.

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Wed Feb 27, 2008 4:29 pm

I can only speak from the experience of my 'silent' pc, but once you've heard (or rather not heard) a really quiet machine it's hard to compromise and add even a single fan;-

While I was experimenting with the p4 version of my machine, I tried adding an 80mm fan ticking over at <5v . It was clearly audible over the tiny noise the machine makes (with the hdd cased, it's only just possible to feel the vibration, it's at too low a frequency to hear it without pressing ones ear against the box).

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Thu Feb 28, 2008 3:53 am

Correct. The reason I want to consider fanless, is because I'm sick of fan noise altogether.

I like the idea of fanless when I first encountered it. It was a modified EPIA Mini-ITX setup with a custom fanless case on a loan from a friend.

Granted, VIA's processors don't produce as much heat as an AMD or Intel desktop solution, but I found how pleasant it was not to hear a thing from a PC.

So here I am, trying to figure out the theory in designing and making my own one. (after researching and finding pre-built ones are really expensive or are limited in component hardware flexibility/compatibility...Like in the case with the Zalman TNN line).

I'll admit, I've had limited experience in thermodynamics and heat transfer, but I'm willing to learn the theory and how to accommodate the factors of real world situations into the calculation. (This is where experienced people are most helpful).

The goal is to be able to get any modern day, mainstream class hardware and turn it into a silent platform. I'll accept for now, that there is HDD noise, but that will be replaced by SSD in due time. The primary noise is from fans. By spending time in figuring out some good concepts in a logical manner, I hope to provide some sort of foundation for other silentpcreview forum folks with their fanless projects. (should they choose this path).

walkingjohn
Posts: 29
Joined: Sun Mar 18, 2007 9:51 am
Location: Ontario, Canada

Post by walkingjohn » Thu Feb 28, 2008 7:00 am

You're way ahead of me as far as the thermodynamics stuff goes, so it would be interesting if you figured that out in general.

But it might help to reframe the problem a bit: there's no way you need to dissipate the TDP of any parts fanless, because no actual system will draw that kind of power for significant periods.

Instead, if you have an emergency fan or two aimed at your heatsinks to take care of worst-case guarantees, you can design for typical maximum dissipation, which is way lower (mind you, I don't know how to find out what it is for a given chip without experimenting--maybe AMD's new ratings would do).

Factor in a bit of undervolting (I've yet to hear here of a chip that won't undervolt considerably with perfect stability) and the problem gets much easier.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Thu Feb 28, 2008 10:44 am

The primary noise is from fans.
this is debatable; okay, modern 3.5" and 2.5" HDD have come a long way in terms of quietness, but if you use SPCR recommended fans (nexus/yate loon/scythe slipstream) and don't have ferociously hot parts it's hard to imagine the fans being the main or dominant noise signature. it's very hard to get a fanned PSU which is dead silent, hence why I have an ST30NF, but for the other stuff unless you're very lucky the hard drive will set the noise floor.
The goal is to be able to get any modern day, mainstream class hardware and turn it into a silent platform.
as long as gaming is not involved, this is perfectly possible. some judicious undervolting and a 500/600rpm scythe slipstream should see you there. this is not meant to put you off fanless, it's a fun project, but even on SPCR (hardcore noise freaks) fanless systems are the exception rather than the norm, and there is a reason why that is the case.

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Thu Feb 28, 2008 12:24 pm

The primary noise is from fans.


this is debatable; okay, modern 3.5" and 2.5" HDD have come a long way in terms of quietness,
It's always possible to encase HDD's, use solidstate or remote drives to remove that source.

There is another reason to avoid fans in quiet systems...

In accoustics or electronics, noise is not just unwanted sound/signal. Noise is defined in mathematical terms as random and non-coherent, it always adds positively when mixed with other sound, and is perceived as peaks rather than average levels (for this reason, noise should be measured with a quasi-peak meter, rather than averaged). What I will call just 'Sound', is non-random, coherent and will both additively and destructively mix.

The sound (as opposed to the noise) a fan makes (particularly at low rpm) is dominated by the pur of the motor and the 'flap' as the blades pass the mounting frame. Both these sounds are coherent and are audible, even with a masking noise source, because they tend to modulate the ambient noise.

In a quiet environment, i.e. typically where a silent pc would be used, the effect of noise-floor modulation is often more disturbing than a louder, smoother, ambient noise.

Plekto
Posts: 398
Joined: Tue Feb 19, 2008 2:08 pm
Location: Los Angeles

Post by Plekto » Thu Feb 28, 2008 6:02 pm

Correct, but consider a passive setup:

Power supply - can be fanless. Check. Though you're stuck at 300W max, effectively. Another option is an external passive power supply. This at least allows you to have the heat vent straight up, like most home electronics. You will need ~2 ft of clear airspace above it, though, to keep it from overheating.

CPU - unless you're using something like an underclocked e8200(2.0Ghz would be fine, actually, even for most gaming and use maybe 15-20W), you need something the size of a Ninja to deal with it - and that's with something that uses no more than ~40W under full load. The Ninja is a good cooler, but with the fan removed and no airflow, it can only handle a tiny amount of heat before it gets too hot after an hour or two.

That leaves the obvious problems - the drives, which even the coolest ones operate at ~120F, and of course, the video card. With losses and such due to inefficiency, you're looking at about 100W minimum to get rid of. I suppose you could use solid state drives, though.(probably a good idea in any case as the HD is always one of the noisiest components.

No DVD drive, either - those are even more noisy when running. Well, maybe one to load in stuff. Definitely no movies from it, though.

Built-in video of course. Note - this won't possibly be able to run HD content, no how, no way. This actually is the big problem right now, in fact - it gets hot and without a fan, there's no way to direct the air. Even a 25W bulb in a small enclosed space will get shockingly hot. As in well over 100 degrees hot. Heat goes in, heat isn't taken out. Heat builds up until the entire PC case becomes a giant radiator. And that's only 25W. 150W from a typical system, or even 100W is enough to heat your case to where you can burn yourself touching it.

Short of a glorified PDA, it's just impossible to make a purely passive computer that can run anything modern. Not without insane engineering and sizes that make it all counterproductive.

Now consider one 120mm fan moving 5 or 10cfm mounted where the power supply was and an external power supply. No heat issues. That little bit of moving air is all you needed. 15-20db from one source- well under the noise of the rest of the house.

P.S. it's the multiple noises that creates interference and annoys us. One quiet drone on the other hand - our brains will filter it out in a few weeks to where you never notice it(like it does with clocks ticking and so on)

Two fans, though... it can't filter that out and it becomes "noise".

So, IME, one fan and one fan only is effectively silent after a few days as long as it's at or near the ambient noise.

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Thu Feb 28, 2008 8:15 pm

I have very positive experiences with active cooling. I am running two fans and still they are quieter than my soft mounted 5400rpm drive. After addressing it my system became inaudible. It was impossible to perceive any noise from where I sit under the most quiet conditions I could master.

On the other hand, the electronics in my ups are audible, so are many pc components.

I still think that a fanless computer is an interesting project but I wouldn`t expect many practical advantages. It`s more like something I would do to learn and practice my DIY skills.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Thu Feb 28, 2008 9:02 pm

With losses and such due to inefficiency, you're looking at about 100W minimum to get rid of.
it's quite easy to make a PC that idles at 30W and doesn't peak above 60W:

viewtopic.php?t=46650

this is without resorting to mobile CPUs or hard-to-find industrial motherboards.
Short of a glorified PDA, it's just impossible to make a purely passive computer that can run anything modern. Not without insane engineering and sizes that make it all counterproductive.
ok, it won't run Crysis at sick framerates, but it is possible to make a passive computer that can do email, word processing & other office apps, play MP3s/DVDs/MPEGs, surf the net etc (ie what 90%+ of people use their PC for) within a passive envelope. an undervolted Conroe-L Celeron for example can be run passive.

surfntom
Posts: 12
Joined: Fri Nov 16, 2007 11:34 pm

Post by surfntom » Thu Feb 28, 2008 10:23 pm

very interesting thread first and foremost. made me register to post my thoughts. I am a graduating mechanical engineer with an interest/concentration in heat transfer....

here are my 2 cents

first of all you should look for a very efficient cpu as you might expect.

You can calculate the size and number of fins your heatsinks need but these are all approximations as all engineering is. I would expect 10-20% deviation from the real result. (could be above or under). I would refer to a book on heat transfer to look at fin equations to develop a basic understanding of how to solve these problems. Another option is to model a heatsink in FEA or a CFD program to calculate its thermal resistance to air.

Like many have said, a fan goes a long way and if you can have a silent fan then you should because you can get much larger amounts of heat transfer.

I would love to begin helping you calculate all of this stuff but it is too time consuming to start.

The book i used at schoo is "introduction to heat transfer" written by Incropera, DeWitt, Bergman, Lavine.

A nice silent option to look into is an ionic fan which can be found here ( http://www.inventgeek.com/Projects/IonC ... Page1.aspx ). This article has been mentioned here before. I tried to make my own but it did not succeed, i will retry my build once i have the time.

If you are going to try and build a system with no fans. You should seek a power supply with the highest efficiency and active power factor correction and orient it so that air easily able to flow in and out. In terms of heatsink orientation, the fins going in a vertical direction is at least 30% better than horizontally if not greater at higher fin densities. When trying to promote passive convection, one should increase the fin spacing to allow the buoyancy of air to take effect. If the fins are too close the boundary layers of air flowing over the fin will merge and result in poor heat transfer. Lapping your heatsinks and using minimal TIM would be highly advised. I use a very small amount, barely visible on each surface. If you are using any sort of graphics card i would highly suggest rotating the atx board 90 degrees so that the graphics card is vertical instead of having all the hot air pile up next to the heatsink. As long as your fins are vertical, and you allow a passage for cool air at the bottom and hot air at the top you should be fine with large heatsinks like the ninja and those from the accelero line. Pick a mobo with a cool NB and SB if possible. Your temps will be much higher than most users (low 30C for C2D but you might see anywhere from 40-60).

You mentioned the use of heatpipes and how that would be calculated. i did some tests with heatpipes that yielded much better results than the review by a a user on spcr. I was able to transmit 120 watts with 4 heat pipes with <1C in drop across the heatpipe. My guess is that the reviewer had a very poor surface contact.

The point of all this is that you can calculate all you want but the bottom line comes down to whether or not you are equipped to make or order a custom heatsink. Off the shelf products are much cheaper and in that case you might as well buy the best possible and not fret the calculations. I was once in your position but then just decided to buy a reasonable cooler and underclock the fan.

another option you could do is install a stereo in your computer so that it is always playing. This will drown out any fan noise ;).

Sorry for the long post. please email me or pm me if you want some more concrete answers. i didnt proofread this so forgive any typos in supposed to be doing homework =/

good luck with your build

Tom

BillTodd
Posts: 132
Joined: Wed Apr 05, 2006 4:10 am
Location: Colchester (nuked in 1984)
Contact:

Post by BillTodd » Fri Feb 29, 2008 5:56 am

A nice silent option to look into is an ionic fan which can be found here ( http://www.inventgeek.com/Projects/IonC ... Page1.aspx ). This article has been mentioned here before. I tried to make my own but it did not succeed, i will retry my build once i have the time.
What a good idea!

I'd completely forgotten about ionic fans - I'll have to try one on the sPC

walkingjohn
Posts: 29
Joined: Sun Mar 18, 2007 9:51 am
Location: Ontario, Canada

Post by walkingjohn » Fri Feb 29, 2008 9:08 am

Have you looked up the "chimney effect"? I recall there being a couple of threads here.

In effect, a tall enough vertical duct acts as a very slow fan.

Probably not practical on its own, but may contribute to heatsink efficiency.

Plekto
Posts: 398
Joined: Tue Feb 19, 2008 2:08 pm
Location: Los Angeles

Post by Plekto » Fri Feb 29, 2008 12:26 pm

jaganath wrote:
With losses and such due to inefficiency, you're looking at about 100W minimum to get rid of.
it's quite easy to make a PC that idles at 30W and doesn't peak above 60W:

viewtopic.php?t=46650
Since this uses onboard video, it's essentially a glorified PDA or a PS2 in a PC box. It won't do Hi-def video content. You must add a real video card to the equation, and this is why it's always in the 70-100W range. 150W is more like it, actually, for a proper high-end PC.

You also have to measure the total draw from the wall. If the PSU is losing 30W to heat, that's going to add to the problem.
ok, it won't run Crysis at sick framerates, but it is possible to make a passive computer that can do email, word processing & other office apps, play MP3s/DVDs/MPEGs, surf the net etc (ie what 90%+ of people use their PC for) within a passive envelope. an undervolted Conroe-L Celeron for example can be run passive.
Such a system with its on-board video and slow processor won't hardly play DVDs, let alone actually play games. You might as well go out and buy an eeePC.

But let's say it's 60W. Get a metal PC case. Put a power supply in it to simulate proper openings/airflow. No need to plug the PSU in. Just set it in the case. Now wire in a 60W light bulb/socket inside the case. Turn it on and come back in an hour.

NASA has fits with their satellites because of this sort of thing. A tiny bit of heat that is in an enclosed space will eventually heat the entire container/box/etc to the point where the entire thing is one big heat sink.

By comparison, one large fan will allow for a full-blown system. And if it's quieter than the background noise from your A/C system, refrigerator, cars outside... it's "silent". This is one instance where our poor hearing as humans is actually an advantage.

Edit - an ionic fan would also work here as well, though most of the noise you'll hear from a PC at this point, no matter what fan you use isn't the fan but the moving air. Nothing you can do about that, really, since if air isn't moving at all, your PC will die.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Fri Feb 29, 2008 3:25 pm

150W is more like it, actually, for a proper high-end PC.
it's very obvious that high-end PCs can't be made fanless unless one adopts the Zalman TNN/Hush approach. that's not what I'm talking about.
Since this uses onboard video, it's essentially a glorified PDA or a PS2 in a PC box. It won't do Hi-def video content.
You seem to put an undue emphasis on High Definition capability. The OP never mentioned it as a requirement, 99% of people get by without it.
Such a system with its on-board video and slow processor won't hardly play DVDs
I must be imagining it then, because my undervolted Celery 430 with IGP plays DVDs flawlessly.
But let's say it's 60W. Get a metal PC case. Put a power supply in it to simulate proper openings/airflow. No need to plug the PSU in. Just set it in the case. Now wire in a 60W light bulb/socket inside the case. Turn it on and come back in an hour.
refer to my earlier point about standard ATX towers not at all being designed for fanless operation, in fact quite the opposite. of course in a standard case the heat has nowhere to go, but all one needs to do is buy something like an NSK2480 and cut convection holes in the top. a little bit of forward planning of the airflow paths can go a long way.
By comparison, one large fan will allow for a full-blown system.
of course. no one is suggesting a fanless PC is the cheapest, most practical or most sensible solution, but after seeing the X millionth P180/2 with all standard SPCR parts, the urge to do something different is virtually irresistible. :wink:

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Sat Mar 01, 2008 3:15 am

surfntom wrote:I am a graduating mechanical engineer with an interest/concentration in heat transfer....
Cool.
surfntom wrote:here are my 2 cents
OK, shoot.
surfntom wrote:first of all you should look for a very efficient cpu as you might expect.
I was thinking mid-range dual core...So the lowest speed of the E8xxx series. Of course, it depends on the fanless cooling solution. If its really good, then bump the possibility to a low end quad-core?
surfntom wrote:You can calculate the size and number of fins your heatsinks need but these are all approximations as all engineering is. I would expect 10-20% deviation from the real result. (could be above or under). I would refer to a book on heat transfer to look at fin equations to develop a basic understanding of how to solve these problems. Another option is to model a heatsink in FEA or a CFD program to calculate its thermal resistance to air.
OK...For a FEA or CFD application, I'm assuming you mean something like Ansys MultiPhysics?
surfntom wrote:Like many have said, a fan goes a long way and if you can have a silent fan then you should because you can get much larger amounts of heat transfer.
I would prefer to have minimal or no moving parts if possible...
surfntom wrote:I would love to begin helping you calculate all of this stuff but it is too time consuming to start.

No kidding. :)
surfntom wrote:The book i used at school is "introduction to heat transfer" written by Incropera, DeWitt, Bergman, Lavine.
I looked around, looks like these folks updated the title...Its now called "Fundamentals of Heat and Mass Transfer". (Same authors)
surfntom wrote:A nice silent option to look into is an ionic fan which can be found here ( http://www.inventgeek.com/Projects/IonC ... Page1.aspx ). This article has been mentioned here before. I tried to make my own but it did not succeed, i will retry my build once i have the time.
Hello...What do we have here? Ionic cooler? How do these work?
surfntom wrote:If you are going to try and build a system with no fans. You should seek a power supply with the highest efficiency and active power factor correction and orient it so that air easily able to flow in and out. In terms of heatsink orientation, the fins going in a vertical direction is at least 30% better than horizontally if not greater at higher fin densities. When trying to promote passive convection, one should increase the fin spacing to allow the buoyancy of air to take effect. If the fins are too close the boundary layers of air flowing over the fin will merge and result in poor heat transfer. Lapping your heatsinks and using minimal TIM would be highly advised. I use a very small amount, barely visible on each surface. If you are using any sort of graphics card i would highly suggest rotating the atx board 90 degrees so that the graphics card is vertical instead of having all the hot air pile up next to the heatsink. As long as your fins are vertical, and you allow a passage for cool air at the bottom and hot air at the top you should be fine with large heatsinks like the ninja and those from the accelero line. Pick a mobo with a cool NB and SB if possible. Your temps will be much higher than most users (low 30C for C2D but you might see anywhere from 40-60).
Yeah, I know about this part.
surfntom wrote:You mentioned the use of heatpipes and how that would be calculated. i did some tests with heatpipes that yielded much better results than the review by a a user on spcr. I was able to transmit 120 watts with 4 heat pipes with <1C in drop across the heatpipe. My guess is that the reviewer had a very poor surface contact.
Thanks. Hmmm...mCubed has pre-made copper blocks and heatpipes. I think we can save some time and effort here...
=> http://www.mcubed-tech.com/english/index.htm
surfntom wrote:The point of all this is that you can calculate all you want but the bottom line comes down to whether or not you are equipped to make or order a custom heatsink. Off the shelf products are much cheaper and in that case you might as well buy the best possible and not fret the calculations. I was once in your position but then just decided to buy a reasonable cooler and underclock the fan.
I suppose. Off-the-shelf is cost-effective.
surfntom wrote:another option you could do is install a stereo in your computer so that it is always playing. This will drown out any fan noise ;).
How am I supposed to sleep then? :)
surfntom wrote:Sorry for the long post. please email me or pm me if you want some more concrete answers. I didn't proof read this so forgive any typos in supposed to be doing homework =/

good luck with your build

Tom
That's OK, and thanks.

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Sat Mar 01, 2008 3:53 am

Regarding that ionic cooler approach...After reading a bit more, I think I'll avoid that approach until its matured and proven.

stmok
Posts: 26
Joined: Fri Jul 09, 2004 2:35 am
Location: Sydney, Australia

Post by stmok » Sun Mar 02, 2008 9:18 pm

I think this will help out anyone looking to build a fanless system...

Thinking Small: New Intel Architecture Champions Sub 1 Watt x86
http://www.dailytech.com/article.aspx?newsid=10930

Its called the Intel Atom processor.

There's currently two versions being mentioned.
"Silverthorne" => Mobile Internet Devices
"Diamondville" => Low cost desktop and notebooks (eg: EeePC)

They're an in-order execution CPU (like VIA's C3/C7), but offer performance within the realm of the original Pentium-M ("Banias") series. Its very likely to be a tad slower. Diamondville will come in dual-core form and marketed as Celeron 3xx series.

The one we should be looking at, is the dual-core one. Said to have a TDP of 8W. :D

I suspect they will arrive in the form of Mini-ITX motherboards, replacing the current D201GLY and D201GLY2 solutions.


Now, back to the topic.

I think I should define my problem more clearly, as my original post was not specific enough.

The system

My problem is really to do with a 24/7 PC used for surfin the web, email, P2P, viewing videos, etc. Just typical desktop stuff.

The OS I'm using is Debian Linux. (Which means I can trim the fat and keep things nice and light as possible).

The PC I currently use now has:

* Sempron 2800+ (Socket 754, AMD64 capable)
* Video card is a GF7300GT (AGP-based, Inno3D brand)
* NF3 250Gb chipset motherboard (Abit brand)
* Hard disk is a Seagate 7200.10 of 320GB capacity.
* Sound is onboard one.
* DVD-burner is a Sony one.
* Case is a cheapo AOpen KF45
* Power supply is an Antec TruePower 430W

The cooling arrangement?

* Video card is cooled by default cooler.

* Hard disk is kept cool by using a Zalman ZM-2HC2 heatpipe solution. As well, I have one of those 5.25" HDD coolers that consist of a dust filter intake and three 40mm fans. I removed the 40mm fans, and it is more like a intake vent.

* CPU is cooled by Zalman CNPS7000B-AlCu
(Fan speed is controlled by Zalman FAN MATE 2 on the back)

* Obviously, PSU has its own fan.

* The system cooling is provided by cutting a hole on the side panel near the CPU and video card area. The fan used is a 120mm Panaflo (low speed) and this is attached to a Zalman FAN MATE 2.

The 120mm fan is pulling air out of the case. (exhaust fan).

As a result, the negative pressure causes the air to suck into the case. By expressly using dust filters in the bottom front of the case and the 5.25 drive bay (dust filtered vent), it will naturally draw cool air from outside those areas. (I have used masking tape to seal holes and such for optimal cooling performance).

Sources of noise?
* 120mm case fan

* Default video card cooler (You can see how its laid out in the link)
http://www.inno3d.com/products/graphic_ ... 7300gt.htm

* CPU cooler

* PSU

Post Reply