Quiet DIY OC'd Pentium D 830 System, Part One...

and Part Two

My God....I thought I went over-board with computer quieting. Apparently all the fan speed changes were done with resistors.....I wonder why a manual fan speed controller was not used instead, maybe replaced by resistors when the final necessary fan speeds were determined? Or perhaps left in use for fan speed changes on hotter days. Or better yet.....some type of automatic fan speed control used instead?

Also, would it have been possible/easier to replace the NB heatsink with a better/higher performance model. The one in the photos (stock I suppose) looks pretty pathetic for a high performance system.

Lots of fans, complicated convoluted airflow, looks like a very difficult project. Hope the end result was worth the extreme effort.

Too many fans to my liking, I'm sure it would have been possible to reduce the number of fans in half.

wow... i have to give props to the author just for shear amount of work involved in achieving his goal. i'm sure most of us have gone through a severe 'tweaking' period when first learning how to silence, but i know mine was not even close to what he went through.

on the other hand, i can't help but think how much easier the path would have been had he better followed SPCR fundamentals in component choices: he has some VERY hot parts in there. it is impressive to see that even very hot components can be made suitably quiet, but i think the real lesson here is that doing so is generally more trouble than it's worth. cooler parts exist that can provide comparable performance with a lot less work. of course, if you're a die-hard tweaker, then a project like this is probably a pretty fun challenge .

No offense but why even try with an Intel 8xx CPU? Why not a Yonah and make life easy?

On the other hand, it would have been a very short article if he'd gone with an X2 3800+ instead: "bought CPU and motherboard. Installed in case with Ninja heatsink and quiet 120mm fan. The End."

However I have to commend the author on his ducting, rarely have I seen such neat and effective ducting outside of one of Bluefront's rigs.

wow - impressive article and silencing effort! I can't help but think there must be a better way to route airflow in the top chamber - I just can't quite think what it might be. At the mo' hot air from the graphics card and NB over/ around the CPU HS and duct to get to the top fan, whereas the ideal fan to exhaust that air would be the rear case fan This looks like an ideal situation for a GPU cooler, like an AC Silencer, that exhausts hot air directly out of the case.

Shame the XP-120 didn't work out - perhaps an o/clocked 830D simply produces too much heat for 2nd hand air to have much cooling potential.

Personally, I'd remove the foam on the exhaust end of the CPU duct - it will add a considerable amount of impedence to airflow through the HS (resulting in higher fanspeed being needed) and shouldn't make a huge difference to sound levels, as the HS fan is far inside the case, on the other side of the HS. I noticed a big difference (~10C) when I added a foam inlet filter to the side air duct on my case, though it does help with noise in my case as the fan's quite alot closer to the case side and doesn't have a HS between it and the outside.

Is there any way to include the NB (w/ a better HS as Bluefront suggested?) and VRM HS into a bottom to top duct, exhausting via the top vent? Or ducting the CPU HS directly to the front of the case?

Certainly shows how quite you can get even a small space heater like this though - it's definately shown it's possible to quiet down even an o/clocked dualcore Prescott. I just hope you have deep pockets for the electricity bill!

This project brings to mind the ..."Rube Goldberg machines"
— devices that are exceedingly complex and perform very simple tasks in a very indirect and convoluted way.

As someone already posted above, some basic research into these components would lead most people to a simpler solution using a cooler cpu and m/b that perform just as well without all the fuss.

I just purchased a p180 to move my computer into. I have a p5nd2-sli deluxe with a pentium D 830 and i was planning on getting the thermalright xp-120. I didn't want a tower heatsink because I tried that once and the voltage regulators got to hot and caused the system to crash. Should I not get the XP-120?

Not intending to start anything here, but makes me love my x2 4400+ (38c idle/45c load), system in sig a little more

Many people at SPCR would not choose to build a system with these components due to the high wattage of the CPU. However, Chris did state that he made a *conscious* decision to go with a hot processor. It is clear that he had his own reasons for doing this. If we accept that quietly cooling this hot processor is a requirement for Chris, then Chris has done an excellent job of meeting this requirement.

I especially like the CPU cooling. Installing a duct from the CPU cooler to the rear air vent and then installing a fan to PUSH air through the CPU cooler and out the back of the case is a wonderful idea:

* The duct maximizes airflow through the heatsink

* Installing the fan in the middle of the case should result in lower noise than a fan mounted on the edge of the case, as would be the case in a PULL configuration. This should be especially true in the thick, well-damped walls of the P180 case.


A couple concerns / thoughts:

* I would be a little concerned by the leverage (cantilever force) exerted on the motherboard by the combined weight of the CPU cooler + fan + duct. As long as the machine is not moved, this should be a non-issue.

* Is the top case fan is necessary and to what extent is the airflow to the top case fan is blocked by the CPU heatsink / duct. What happens when the top case fan is turned off?

ooh I hear you about the p5nd2... gave myself a blister from touching the anodized red heatsink a few months back. That hurt like nothing I've felt in a long time, massive heat off that thing.

I recall the best temps obtained in the SPCR review with that same heatsink were rear fan blowing in, and top fan blowing up.
Ducting used with that config would be an interesting test.

I have a phantom 350. The bottom gets much hotter than the top under load and the fins are bigger there too. I would put the cables on top.

Really good to read about others builds, and an extensive article.
But when he went through the trouble on applying new TIM to the GPU, why not have a shot at the NB HS? Or perhaps even replace it? And the other one too?

And why the complaint on the number of fans? It may work for him.

Very nice reading.

I wonder if the XP-120 would have performed better had it ducted cool air through a side vent like the CAG on the SLK3000B. Would have been not a pretty mod to the case however.

I think the same, that top fan is just in front of one flap of his CPU duct. Very inefficient and lots of turbulences that end as noise.
Besides, as I understand it, he runs the computer with the case open. So why not remove that fan altogether, since it's useless because the heat can go out of the case already.

Regarding that point, I find it strange to buy a P180 and to end up running it with the side door open. When he realized that heat was building up inside, and that it was getting too hot for the mobo, he chose to open the case.
I would have chosen to duct the top fan to the CPU heatsink to prevent that heat buildup. The CPU fan would have become useless, and with proper ducting/air flow management, even the fan in front of the northbridge/passive graphic card may have been removed.
He chose to have lots of fan running very slowly, the result is that he couldn't remove the heat and had to open the case. Not an ideal solution IMHO. But with an overclocked, overvolted double-core Intel CPU he sure did a great work to end with something he can call 'quiet'.

The stack of coins he put on the top fan mesh is very puzzling to me.

Does that dampens the case opening, or the motor of the fan through the foam ?
Or is it just that it reduces air turbulence ?

I always thought it would be a good idea to put something there to reduce air turbulence.
Air flow is going around that place, but just behind the motor you have a sort of negative pressure that causes turbulences and then noise.
It's like the end of a car : to improve aerodynamics, you don't end sharply but you have a profiled trunk and even a fin (aileron ?) on some cars. Why not put something there on fans to remove the negative pressure problem ? Like on that drawing (yes, yes, I know I'm the master of MS Paint)


The air flow is slower near the motor because it depends on the distance from the centre of their rotating movement, but still, on 120mm fans that might be sufficient to contribute to the total airflow induced noise. Plane propeller heads are like that too.
Even if it doesn't reduce a lot the airflow-induced noise, it will improve aerodynamics, thus efficiency and thus allow to rotate slowerly for the same airflow, resulting in less motor-induced noise.

I think the aerodynamic effects you are talking about, on the scale of an average 80/120mm case fan, are too small in relation to the other large sources of operational noise, such as poorly-lubricated bearings and noisy switching electronics, but your principles are sound.

So many questions and comments!
I'll respond to a bunch of them together in this post.


It's a hobby. It doesn't have to make much sense...

I wanted inline resistors in the end, just as sold with the AcoustiFans, small and tidy. At the very end, only one fan wasn't 5V, so the method doesn't matter. (What made you think I didn't experiment using a controller?) As for hotter days, I avoid running CPUBurn and Prime95...

The only automatic controllers I know about are the NMT-2 and NMT-3. I did consider them, but they don't have adjustable temperature ranges, and would always be 5V in my system.

The NB heat sink is custom with specialized attachment hoops. Rather than try to match these, a simple duct seemed better, and it certainly works well.


Of course. But bear in mind my primary goal was performance. Making it quiet was more for entertainment. Remember "A League of Their Own"? "It's supposed to be hard. It's the hard that makes it great."


Um, maybe because it didn't exist at the time?


Indeed.
Thanks for the comments on the ducts. That was one of the more fun and interesting parts of the project. Always remember: measure twice, cut once!


It depends how fast you want to run the fan on the XP-120. At 5V (700 RPM) the CPU was adequately cooled, but the air leaving the HS was very hot. Higher speeds would give better heat transfer and cooler air, but with more noise. There is no doubt that the Ninja cools the CPU better with absolutely minimal air flow, but then the VRM is on its own...

If your system crashes, it's quite likely due to voltage droop (the VRM efficiency and output voltage drop when it's hot). Check this with CPU-Z and two copies of CPUBurn: the Vcore should always be above 1.20V for correct operation of an 830 D. If you see low Vcore, bump the CPU voltage in the BIOS.


Having the back fan blow in and the top fan blow out cools the CPU amazingly well (at least 10C cooler), but the rest of the system cooks...

The air flow across the top of the Phantom mounted in a P180 is way better than across the bottom, so naturally the top heat sink is cooler due to better heat transfer. I stand by my recommendation to put the cables underneath.


Thanks.
Actually I did put Arctic Silver 5 on the NB; I forgot to mention it in the article. When I did that, the HS temperature went up at least 5C (from hot to very hot); clearly the stock compound wasn't transferring heat very well.


Thanks.
When I tried the XP-120, the case was open, so the air entering the HS was only about 25C. The air exiting was over 40C. With the case closed this would have been even hotter.

Don't get me wrong, the XP-120 is a good HS, and it did cool the CPU. It just wasn't able to cool both the CPU and the VRM in my system without a cyclone blowing through it.

Thanks.
The air coming off the GPU and NB is barely warm in the final configuration; the big problem is that the air coming off the VRM (right at the top of the MB with the Ninja towering over it) is really hot. So I'm not so much concerned about air routing as I am about air flow across the VRM, and there just doesn't seem to be a good way to get any up in that corner.


I believe that to be the case. The XP-120 has small fins, so the air leaving them is pretty hot at reasonable fan speeds.


The foam I used there is only 5mm thick. It probably doesn't make much difference one way or the other; it just occurred to me that all those shiny flat surfaces should be damped. I'm running the fan at its lowest possible speed (5V).


I guess I could have built a unified duct for the CPU and NB. Never occurred to me. I thought long and hard about how to deflect some air down to the VRM and struck out. It's just in a bad spot on the board for cooling with anything other than the (incredibly loud) Intel stock cooler or an XP-120 with slightly quieter but still major air flow. As for ducting the CPU to the front, that would be Really Hard; too many corners and obstacles.


I consider the $0.75 a day it costs to run folding in the background to be a charitable donation.

Thanks for assuming I'm not an idiot!

[nice comments about CPU ducting omitted]


The foam strip added to the back wall of the case helps support the duct in addition to sealing off the air flow. That said, I wouldn't move this system more than a few feet without packing it full of bubble wrap, as MikeC describes in his "Torrid Thailand" article.


It is very necessary, not for the CPU, GPU or NB, but for the VRM. The hottest parts of the VRM are very close to the top case fan, and the air it exhausts on the edge closest to the MB is extremely hot. If it weren't so darn loud, I'd run this fan faster. The presence of the CPU duct should actually increase the air velocity past the VRM due to the Venturi effect. Or not! If I turn this fan off, the MB temperature soars and the system becomes unstable.

Some of the members have been unnecessarily harsh on the author of a very fine DIY piece.

It may be easier to build an X2 than an Intel 2 core desktop system, but that is not a reason to not do one. There are lots of ways to make an X2 system into a complicated project:

1. High ambient temperatures expected, as in the Thai PC.

2. Overclocking an X2 beyond 2500 mhz. 2400 is easy, 2500 requires a bit more care, but 2600 (and up) is difficult.

3. Using a small case can be "fun" if seeing blood is your idea of fun. My X2 system is in an Antec SLK2650B and the small size was a bit of a complication. Fortunately, Band-Aids are cheap and my hands are tough.

Building a one-off PC is a hobby thing, no mater what processor is used. If the result is good, it does not mater if the path to get there was out of the way.

True in theory, but this fan is running very slowly. I have to get within a few inches to hear it separately from the overall system noise.


Because the VRM heat builds up in the top back corner and cooks everything.


It is indeed strange. As explained in the article, opening the case let me slow down the fans, speed up the clocks, and at the same time reduce the noise! Your ducting ideas are based on the assumption that CPU heat is the problem; this is intuitive but not borne out by experiment. The real culprit seems to be the VRM.


Neither really. With the fan soft-mounted, its motor & hub vibrate and put out a 450-Hz tone that I found very annoying. The combination of the foam and the coins got rid of that tone.


Interesting ideas, do you think they are applicable to fans and air flow this slow?

Thanks diver.
This project provided less frustration and at least as much fun as getting my ego trashed on the golf course.

you'd be surprised at how much even a thin piece of foam can have on airflow - your CPU HS fan may be at minimum already, but the extra airflow may allow extra o/clock at the same temp, or lower CPU temps. If you can, just try it, that's the only way to find out

I wonder if there's a taller HS you could use for the VRM module that would project more into the CPU duct airflow? Something like the new Thermalright NB cooler would be ideal, if a bit expensive! Or, would bending down some of the lower fins on the Ninja (a la new AC Freezer) help direct some airflow over the VRM HS? It's a tricky one, but that VRM module must be handling a huge amount of power.

I didn't know enough about the P180 to know whether or not my front CPU duct was too silly. Still, if you're bored one winters day...

I run my computer without the side panel too. When the side panel is on, the motherboard temperature gets so hot that the cpu fan on my pentium D 830 spins too fast and it gets too noisy. It is quieter with the case off. Hopefully when I switch to my new p180 I will be able to run it with the side panel on. All I have to do now is decide between the Ninja and the XP-120(or SI-120). I'm worried that if i get the Ninja the VRMs wont get enough cooling (which happened before with a different tower heatsink and made the system crash), but I am also worried that the thermalright xp-120 will blow hotter air donw onto the VRM. I am probably going to run my fan at 12 volts and i don't overclock so it should be a little cooler his. Maybe the SI-120 would do blow cooler air onto the VRM since it doesnt have the fins onthe bottoms that would heat the air up.

cmthomson, thank you for creating such a thorough and carefully documented write-up on your experiences. I think it's very valuable for showing exactly how the whole process can go, and what kind of challenges you can face, sometimes unexpectedly.

From the ducting layout pictures in the article, I'm not sure whether there would be any Venturi created. To get a Venturi effect the airflow channel has to narrow and then widen out again; your CPU duct looks like its cross-section doesn't vary.



Yes, as a leisure pursuit it's surely better than getting moronically drunk and shooting up the local liquor store. However, while I admire the creative effort that has gone into this machine, there is no way I would recommend this approach to a newbie looking for a simple build to make a quiet computer. The simple fact is, the less heat a system generates, the easier it is to exhaust that heat quietly.



You can always create a deflector out of cardboard or foamboard to deflect air downwards onto the VRM's.

Just so Chris doesn't feel too picked upon , I realise how satisfying it can be to bring a build to a succesful conclusion, and how important it is to have both performance AND quiet. That is the holy grail of quiet computing.

Oh, you want my recommendation to a newbie?

DON'T DO IT! Buy a Core Duo iMac instead.

Alternatively, build your system around a Yonah or Sossaman -- you'll give up 64-bit, but also shed 100W. These didn't exist when I started my project. If you start after Q3 06, use a Conroe.

First of all I appreciate your work, it is a real challenge to quiet such a hot system.

But my question is related to a HDD topic mentioned in your article:

"Because I was optimizing the system for multiprogram performance, I considered the NCQ configuration to be better than the more popular bandwidth-optimizing RAID configuration, which disables NCQ"

It is true that enabling a raid configuration with NCQ capable hard drives, will disable the NCQ feature?

Hi Chris,



I agree with you on the iMac -- but you seem to be completely ignoring the Athlon 64? I just put together an Athlon 64 X2 4200+ system:

Gigabyte GA-K8N Pro SLi
Scythe Ninja -- "passive"
Evercase 4252 w/ SeaSonic S12 430
Scythe DF 120mm exhaust at 5v ~830RPM
Scythe DF 92mm intake at 5v ~1100RPM?
2GB (4X 512MB) PC3200
Seagate (7200.8?) 120GB SATA
AOpen nVidia 6600 128MB
Gigabyte X800 LT 256MB
WinXP Pro 64

Other than undervolting the fans -- which was just plugging them in using the 5V adapter that SeaSonic supplied, this is just a straight assembly job. It took me all of a few hours to put together. I used the BIOS to undervolt the CPU to 1.26v, and I had to bump the RAM voltage up by 0.1v, since one of my pairs of RAM required this.

It is very quiet (the Scythe fans are very nice!) and the highest CPU temperature I have seen is 54C, with both cores running Folding@Home. And that was when the heating system here at work had the room temp up well over 80F!

The loudest thing is the HD seeks, and even these are not really noticeable when it is under the desk. Fan noise is just audible even when you are next to the machine. The other computer under my desk (the one supplied for me to work on), is a Dell Dimension 3000 2.8gHz Prescott -- and it is far louder than my system. The Dell's case fan is easily audible even with it under the desk.

The bottom line is, at the moment, the Athlon 64 X2 is much easier to make quiet and cool -- and it probably is at least on par with, if not out-performing anything Intel has to offer.

I'll try to post it in the gallery; but there is no way I'll be able to rival your effort in writing up this article! You are to be congratulated for that, and for your dedication to such a hot-running system. But, I'll take an Athlon 64 any day; especially if I have to build it!