You are here

Quiet MP Dual-CPU Workstation

January 7, 2003 by Leo Quan

Introduction by Mike Chin, Editor -- Leo, a graphic designer and visual artist, tells us about his trials and tribulations with a new topic for SPCR: The challenges of cooling a dual-Athlon, dual-hard drive workstation quietly. Leo applies PC silencing concepts freely, modifying them to his needs. He also creates new ways and techniques, and generally goes the whole 9 yards to achieve his ends. While his work is of particular interest to those with a noisy dual-processor machine, it is interesting to any quiet PC modder. You may reach Leo at info at powergyoza.com.

I've tried to quiet my dual-Athlon workstation ever since I first put it together. The very first setup would not be considered loud by overclocking or gaming enthusiasts, but it was driving me mad. It had Thermalright SK-6's w/ Y.S. Tech 60mm 26 CFM fans, 2 Antec case fans as back exhausts and 2 Sunon 30CFM fans as front intakes.

People may call the Y.S. Tech 26 CFM fans quiet compared to the Delta 38CFM and 50CFM screamers, but they were still bloody loud! Ditto for Sunon and Antec case fans. Out went those fans and in came the Panaflo L1A 80mm fans for the case and Panaflo H1A 60mm fans for the SK-6's. I could instantly hear the improvement, and I was even satisfied with the setup for a while.

But I wanted better. Better as in, "I want to hear the movie, not the whirr and whine of the computer." Better as in, "I wish I could sleep at night while my 'puter crunched away at RC5/[email protected]"

Editor's Note: Here's a preview of the final result of Leo's work. See if you can spot all the little tricks and techniques Leo artfully applied. Pop quiz later!

BASELINE SETUP

My dissatisfaction led me to the purchase of a pair of Thermalright AX-7's and Panaflo L1A 80mm fans. I figured that by removing the 60mm Panaflos H1A's (rated at 0.17A and 19.1CFM and quite noisy), I would eliminate most of the racket. The hard drives were an obvious problem too, but I couldn't afford the time and expense of getting new hard drives and reinstalling all my applications and data. So with the AX-7's in place, I had my baseline setup.

Editor's Note: A look at his PC before Leo caught the quiet bug. A blue nostalgic look?

Call my standards low, but it was a decently quiet machine. It was certainly quieter than my old K6-500 rig that had 4 hard drives! That old thing had a Quantum Viking II and was a horribly loud bugger! Still, I had many improvements that I could make to quiet my hot dual-Athlon baby. I certainly had cooling headroom with this setup. My Athlon MP 1600+'s were running quite cool, as were the hard drives. I could surely sacrifice some temperature headroom for a quieter setup. This was my "before" setup:

Antec SX-1040 case w/ Panaflo 24CFM fans (2 rear exhaust, 1 intake for hard drives)

Antec SmartPower 400W power supply

Tyan Tiger MP (2460)

2 AMD Athlon MP 1600+ @ stock voltage and MHz

Thermalright AX-7 w/ Panaflo L1A fans (12V, blowing)

2 GB RAM

2 Maxtor D740x Hard Drive (40 GB and 80 GB, ball-bearing models, rigidly mounted)

Floppy Drive

Lite-ON 24x10x40x CDRW

Pioneer 305S 10x SCSI DVD Reader

Matrox Millenium G550 Video Card

D-Link DFE-538TX NIC

Pinnacle DC30+ Video Capture Card

Tekram 395 SCSI Card

Samsung Syncmaster 700NF 17" monitor

A couple other SCSI peripherals

System Power

The total AC power draw of the system was measured with a Kill-a-Watt power meter.

  • At idle, it measured 210W.
  • At maximum, it was 250W

If one assumes 70% PSU AC/DC efficiency, this translates to

  • At idle, 147W DC power delivered to components.
  • At maximum, 175W DC power delivered to components.

Maximum power was the highest peak seen with CPU stress testing plus random opening/closing of CD drives with disks in them to force

Noise and Temperatures

Using an old analog Radio Shack sound level meter and the temp monitors from my motherboard, I registered the following readings:

Baseline Setup Sound level (dBA)
Front of case (0.5" distance from case at HDD height)
62
Hard drives (inside case between 2 hard drives)
70
Exhaust fans (outside case, near corner closest to fans)
69
CPU fans (between sides of 2 CPU fans)
69
Temperature (C)
CPU 0/CPU 1 (stressed with CPUBurn - BurnK7)
51 / 48
40GB/80GB HD (Dtemp)
38 / 34

Although the other readings were higher, I considered the 62dBA front of case measurement to be most important. The front of the case has the most direct path to my ears while I'm working or using my workstation as an HTPC. In addition to making my computer sound quiet to me, I also wanted to reduce this measurement as much as possible. I was sure I could do this as I had lots of cooling headroom that I could take away from CPUs and the hard drives.

THE MODS

With the advice and encouragement of SilentPCReview's Mike Chin, I took on the challenge to quiet my computer. The project consisted of quieting all the major noise producers: the fans, the hard drives, the power supply and the case itself. ( Editor's Note: In fact, the project has engulfed Leo's life and trapped him in a never-ending cycle of mods and more mods. I'm really sorry, Leo! Someone please rescue this man!)

This article was originally written in a chronological fashion. Later, because of all the going back and forth I did with so many mods, that approach was more or less abandoned. It might be best regarded as a collection of related noise-reduction experiments on a dual-CPU PC.

1) Undervolt fans to 7V and remove intake fans

Undervolting the fan was a simple mod that only required swapping some power leads of the Molex power connectors. Instead of connecting the negative terminal of the fans to the ground line, I connected them to the 5V line. I have no variable voltage but I have 7 volts! Judging from Mike's high end cooler shoot-out, I figured that 7V was the lowest I could go anyway.

Case & HS fans @ 7V
dBA
ChangeB*
Front of case (0.5" from case at HDD height)
62
0
Hard drives (inside case between 2 hard drives)
70
0
Exhaust fans (outside case, corner closest to fans)
69
0
CPU fans (between sides of 2 CPU fans)
69
0
Temperature
C
ChangeB*
CPU 0/CPU 1 (with BurnK7)
64 / 56
+13 / +8
40GB/80GB HD (Dtemp)
NA
NA

*ChangeB = Change from Baseline

While my CPU temperatures jumped, I gained no improvement in noise levels. In fact, I could hear my power supply fan work harder, probably because the undervolted exhaust fans were allowing more heat to build up inside the case. I could hear the pitch of the fan noise slowly pick up after only 5 - 10 minutes of turning on my computer.

2) Replace the PSU fan with a quieter model

The 80mm fan that came with my 400W power supply was a 38.6 CFM Adda ball-bearing unit (AD0812HX-A70GL). It's specs are listed as 0.25A, 3010rpm and 34dBA. I tossed that fan out and replaced it with a quieter 18CFM NMB ball bearing unit (DC3110KL-04W-B19). It's rated at 0.06A and 18dBA.

I ran into some trouble with connecting the new fan to the header. The header is some sort of 2 pin mini connector, smaller than a Panaflo 2 pin connector, but thankfully still the same pitch. In my attempts to get the connector to fit to the header, I dislodged the balls of solder connecting it to the PCB! Nothing that some solder can't fix. In the end, I found that I could shave a Panaflo 2-pin connector enough to fit.

Quieter NMB PSU fan
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
58
-4
-4
Hard drives (inside case between 2 hard drives)
60
-10
-10
Exhaust fans (outside case, corner closest to fans)
64
-5
-5
CPU fans (between sides of 2 CPU fans)
63
-6
-6
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
66 / 61
+15 / +13
+2 / +5
40GB/80GB HD (Dtemp)
NA
NA
NA

*ChangeB = Change from Baseline; ChangeP = Change from Previous

This modification made a noticeable acoustic improvement. The noise levels all around the case went down. The most audible components now were the hard drives. My components were running a bit hotter still and the exhaust from the case and PSU fans were quite warm, but my computer was still stable. These results cannot be attributed entirely to the change in PSU fan. Before I took these measurements I also inverted the HS fans so that they'd be sucking and made a exhaust duct so that the hot exhaust would be vented out directly.

3) Decouple Mount the Fans

Rather than buy elastomer fan grommets, I cut up 2-3 inch pieces of soft foam and used them as grommets as shown in the photo below. The foam pieces get shoved through both the panel mounting hole as well as the fan mounting hole. Friction keeps them in place.

The little strips of tracing paper hanging off the fan opening help me confirm with a quick glance that the fans are working: the light paper strips flutter in the breeze -- or at least stand up a bit even when the fan is running slow so I can't hear it easily.

The drawing below details how the foam grommets are installed. Not pretty but that's not a real concern for me. It is very effective at reducing mid/low frequency noise caused by fan vibrations.

Here is a photo of the inside, showing how the red colored electric tape was used to close the gap between chassis and fan to minimize airflow diffusion. I don't believe the tape short circuits the mechanical decoupling of the foam grommets, but I have not seriously tested this.

Also visible in the above photo: Two small pieces of foam were also used as decoupling washers on the screws used to mount the 80mm Panaflos on the CPU HS, one between HS and fan and another between screw head and fan.

4) Provide Cooler Air for the PSU

The vent slots on the intake side of the PSU seem too bulky. I felt they could be more open to allow better air intake, so a pair of needle noise pliers were used to twist them 90 degrees, substantially increasing the air gap between each metal slot.

Then I decided to provide a path for fresh air to come in from the top front of the case directly to the PSU intake vents. I did this with the hope that the case inside would stay cooler which would help the CPUs, the hard drive and other components such as the voltage regulators. An AMD system builder guide states that the front vents of a PSU can rob the HSFs of cool air, thereby preventing them from cooling to their full capacity. A tunnel was made with foam and stiff paper. This is visible in the photo above as well as the one below. It is marked #3 in the photo below.

5) Case Door Mods

The new PSU intake duct required a larger intake vent in the front of the case. So...

Although the Antec case has some ventilation holes that appear to be designed in around the edges of the front behind the hinged door, the intake air tunnel seemed a bit silly without a larger opening. So the top of the plastic door was cut open as shown in #1 above. A lip that reduced air intake on the bottom of the door, #2, was also cut away.

Later, I created a rubber sound-damping gasket, marked #4, that presses up against the optical drives to reduce their noise when accessed. Note the recessed notches cut into the rubber piece to prevent the door open buttons from being pressed when the case door is closed.

6) Quiet the Hard Drives (and other mods)

I next underwent a long process of experimentation to quiet my hard drives. I tried several materials and used multiple configurations, but only my final attempt gave me the best performance acoustically and thermally. Only the noteworthy versions are reviewed here. There were a few other... aborts.

VERSION 1: camera foam, 1" thick, "taco" w/sidesinks. Good acoustic performance, but poor thermal characteristics and case fitting issues

I built a foam structure in the shape of a taco. Each hard drive was covered in foam on the top, bottom and front. Side heatsinks were made from doubled U-channel aluminum pieces and screwed tightly to the sides of each hard drive. This is the cooling concept presented in the SPCR article Effective Passive Hard Drive Cooling.

Sorry for the blurry photo. It really is 2 drives covered in foam and equipped with side heatsinks.

V1: Camera foam HDD sandwich
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
52
-10
-6
Hard drives (inside case between 2 hard drives)
55
-15
-5
Exhaust fans (outside case, corner closest to fans)
56
-13
-8
CPU fans (between sides of 2 CPU fans)
61
-8
-2
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
64 / 59
+13 / +11
-2 / -2
40GB/80GB HD (Dtemp)
~56C
+18~22
NA

*ChangeB = Change from Baseline; ChangeP = Change from Previous

This mod was a minor success. Even though the drives were not in perfect suspension, idle noise dropped to the point where I could almost sleep at night, but temperatures shot up. Way up. So far up that I stayed awake at night worrying about my hard drives. How ironic! Maxtor states max. operational temp is 55C. Seek noise was terrible with this setup. While the upper hard drive was okay, the lower hard drive made a huge racket.

VERSION 2: Vented 1" rebond foam pita pocket w/sidesinks & case hacking. Acoustic deadening better than V1, but even worse thermal performance

In this version, I constructed a foam enclosure that resembles a pita pocket using the densest rebond foam available at my nearby shop. This is much like the stuff used for carpet underlay. Not only were the tops, bottoms and fronts covered in foam, but also the sides. I tried to ensure adequate airflow by creating convection vents for the sidesinks. To fit the drives into my case, I had to remove the lower drive cage tray and suspend the pita pocket in a vertical orientation.

Before I took these measurements, I also added foam to the feet of the case to decouple it from the underlying surface.

V2: 1" rebond foam pita pocket w/sidesinks
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
-12
-2
Hard drives (inside case between 2 hard drives)
-20
-5
Exhaust fans (outside case, corner closest to fans)
-19
-6
CPU fans (between sides of 2 CPU fans)
-19
-9
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
68 / 61
+17 / +13
+4 / +2
40GB/80GB HD (Dtemp)
~66C
+28~32
+10

*ChangeB = Change from Baseline; ChangeP = Change from Previous

Although the measurements state otherwise, this setup actually sounded a wee bit louder than V1. There seemed to be more high-pitch whine coming through. I would have considered this a pretty good attempt, except that my hard drive temps were way, way over spec. Closing off the sides of the hard drives must have really choked the airflow. I stayed with this setup for a nerve-racking 2 days before I abandoned it.

VERSION 3: Backtrack to bare suspension, sidesinks, many other mods. Somewhat quieter than before quieting the hard drives

I was hoping to see the difference between a bare suspension and a sandwiched one, but I had an impatient streak and performed a whole bunch of modifications before I made any measurements. I did the following:

a) removed the lower exhaust case fan and replaced with filter

b) inverted the HS fans back to blow air onto the heatsinks

c) rerouted SCSI and floppy cables behind the motherboard for better case airflow

d) moved PCI cards down to make more space around video card

V3: Not-so-baseline bare suspension
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
54
-8
+4
Hard drives (inside case between 2 hard drives)
60
-10
+10
Exhaust fans (outside case, corner closest to fans)
50
-19
0
CPU fans (between sides of 2 cpu fans)
54
-15
+4
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
69 / 65
+18 / +17
+1 / +4
40GB/80GB HD (Dtemp)
45 / 49
+7 / +15
-17~21

*ChangeB = Change from Baseline; ChangeP = Change from Previous

The bare suspension revealed no surprises. Noise levels increased, but the heat decreased.

VERSION 4: 1" rebond foam club sandwich and PSU ducting mod. Acoustic performance comparable to V1 with thermal performance that's just barely acceptable

Here, I modified the pita pocket from V2 by detaching the side pieces of foam from the pocket and instead attaching them to the sides of the case. Between the hard drive was a 1/2" piece of EVA foam.

I obtained the following results:

V4: 1" rebond foam club sandwich + PSU ducting
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
50
-12
-4
Hard drives (inside case between 2 hard drives)
50
-20
-10
Exhaust fans (outside case, corner closest to fans)
50
-19
0
CPU fans (between sides of 2 cpu fans)
53
-15
-1
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
66 / 58
+14 / +10
-4 / -7
40GB/80GB HD (Dtemp)
55 / 55
+7 / +15
+10 / +6

*ChangeB = Change from Baseline; ChangeP = Change from Previous

I can report that my rig sounds about as quiet as it did in version 1. But HDD temps were too high.

PAUSE: Materials Analysis

In all the previous versions of HDD silencers, I kept running into 2 issues:

1) sound leaks through wherever there is no sound barrier, and

2) sound leaks though the sound barrier itself.

Issue 1) could be easily observed with a close ear-listening test. I decided to test issue 2) using my alarm clock radio as a noise source. Just by simply placing the different materials over top the speaker, I could get an idea of how well each blocked the sound.

In addition to the materials used in the previous attempts, I also included a high-performance acoustic ceiling tile (as mentioned by Melchior on the SPCR Forums) and block print carving rubber (inspired by the ideas on http://www.geocities.jp/numano3).

This is what I found when I measured from 5" above my radio speaker. It was tuned in to an unused AM radio band that contained a lot of broadband noise, essentially a type of pink or white noise.

Material
Sound level (dBA)
Noise Reduction (dBA)
None (reference)
74
-
1" camera foam
70
-4
1" rebond foam
66
-8
1/2" EVA closed-cell foam
64
-10
3/4" acoustic ceiling tile (NRC: 0.85)
60
-14
3/8" block print carving rubber
55
-19

The block print carving rubber for artists was by far the best. So, I decided to build V5 from the rubber. It can be found at good art supply stores. It is the same material I used to make the rubber noise reduction gasket for the door that closes over the CD drives.

VERSION 5: 3/8" Rubber Enclosure, foam feet decouplers. Best acoustic performance so far, just-bearable thermal performance

One rubber enclosure was built for each of the two drives. Each box measures 4.75 x 1.75 x 7 inches. The pieces were cut up with a utility knife and super-glued together. A cover allows the IDE cables and power wires to be fitted snugly for minimal noise escape. The boxes are sized so that the drive fits tightly, ensuring maximum heat transfer by direct conduction. Before the drives were inserted, I applied thermal grease liberally on the top, sides and front surfaces of the drive. The boxes were placed one on top of the other at the bottom front of the case on a bed of foam to decouple them from the case. The cost of each box was approximately US$10.

Here are the results:

V4: 3/8" rubber enclosure
dBA
ChangeB
ChangeP
Front of case (0.5" from case at HDD height)
>-12
?
Hard drives (inside case between 2 hard drives)
>-20
?
Exhaust fans (outside case, corner closest to fans)
>-19
?
CPU fans (between sides of 2 cpu fans)
>-10
>-3
Temperature
C
ChangeB
ChangeP
CPU 0/CPU 1 (with BurnK7)
66 / 61
+15 / +13
+1 / +3
40GB/80GB HD (Dtemp)
55 / 55
+17 / +21
+10 / +6

*ChangeB = Change from Baseline; ChangeP = Change from Previous

The dBA readings really are not very useful; they fell below the scale of the sound level meter, the bottom limit of which is 50 dBA. There is no question this is the best HDD silencer so far. The 20 dBA measured reduction from the hard drives seem consistent with the noise reduction from the alarm clock radio test. The rubber seems to conduct some heat away from the hard drives too because they perform comparably to the sidesinked, foam-sandwiched hard drives.

The noise of the HDD was greatly reduced, and high pitch whine was mostly eliminated. I could live with this.

With this rubber box mod, the PC became the quietest noise producer in my space - quieter than the forced-air heating, quieter than the refridgerator in the next room. Although much hotter than before - between 13C - 21C hotter - the components were still within maximum temperature specs. But the borderline high temperature of the Maxtor hard drives worried me.

Back to the drawing board.

VERSION 5A: Add sidesink and front fan to V5.

Again, the sidesinks are based on the SPCR article Effective Passive Hard Drive Cooling. Several small holes were drilled through the rubber to expose the mounting holes on one side of each of the drives. Metal grommets and bolts were used to tightly secure the aluminum u-channel pieces, with heatsink grease at the interface. Between the hard drive and the inside of the rubber enclosures, I also put a generous amount of heatsink grease. A 6V Panaflo at the front of the case is blowing onto it.

  • Old Temps: peak at 60C / 60C (for deach drive, vertical position)
  • New Temps: peak at 39C / 42C

Although the drives are much cooler, perfectly safe, in fact, the noise went up substantially. There was considerably more whine than before. Perhaps the aluminum channels were transferring the vibrations out or the small gaps were allowing the high pitched sound to escape more. I needed to quiet them down more again.So...

VERSION 5B: Remove sidesinks from drives, leave Panaflo fan

The side HS were removed and the holes in the rubber boxes plugged up tightly with foam. The noise level returned to what I had achieved with VERSION 5, the rubber boxes alone, because the extra Panaflo fan at 6V is pretty much inaduible.

  • My new max HDD temps are now 44C and 45C!

This is a good stop on the HDD silencing road for me, at least for now. (Editor's Note: With the amount of time, money(?) and energy put into HDD silencing, we know Leo could have got quiet Barracuda drives and transferred the data without any problems. But he wouldn't have had nearly as much fun, and we would not have a new quieting solution for loud hard drives.)

*

Yet another Mod: Better Fan Control with Diodes

The 6V supplied to the HDD fan above comes from a a circuit I built using diodes and a 6-way rotary switch (see schematic). The circuit is similar to the Diobus at http://www.cpemma.co.uk/sdiodes.html, but rather than soldering diodes directly to the rotary switch, I created mine on a circuit board. This allowed me to more easily solder extra headers to connect my heatsink and case fans.

I used diodes becaused they are a good way to control voltage. The voltage drop of a single 1N4001 diode is fairly constant at about 0.75 volts and only increases slightly with high current draw (eg: 1 amp). This allowed me to control the voltage in fairly small and predictable increments. Diodes also do not require heatsinks because they give off little heat (1 watt for a 1 amp load).

I wired 4 diodes in front of the rotary switch to give me 9 volts at the first step. I added a diode at each step to give me voltages all the way down to 5.25 volts. My fans are normally at 6 volts, but my machine is stable down to 5.25 volts.

*

Conclusions and Future Mods?

Living with a quieter computer has been very rewarding. It's now subjectively the quietest noise producer in my space - quieter than the forced-air heating, quieter than the refridgerator in the next room. It's even beginning to act like a half-decent HTPC. Although much hotter than before -- between 13C - 21C hotter -- the components are still safely below maximum specified temperatures.

Because of the progress I've achieved with my computer, some new components are becoming the greatest noise producers, while other previously quieted components are becoming "noisy" again. Like other SPCR forum members, I'm getting annoyed by the buzzing of my monitor. Also, my newly undervolted fans are now audible. And the PSU fan, even though it is a quiet fan running at 10.5V, calls for still more noise reduction.

I know it is a subjective thing but it is very real: my hearing sensitivity seems to go up with every step closer to silence. Sigh... A single CPU system or one with quieter drives would have been easier to deal with, but after getting used to the power of my dual MPs, there is no going back.

* * * * *

Discuss this article in the SPCR Forums.

Google

www SPCR