Review: STEALTH XP2000+ PC by ARM Systems

Complete|Mobile Systems
Viewing page 3 of 3 pages. Previous 1 2 3


Power Supply

The Seasonic 350W PSU is a higher power upgrade option. Our review of the Seasonic 300W PSU shows it starts quiet and stays that way even at the maximum power draw of our test platform (90W). Seasonic's specifications indicate that the 300W and 350W APFC models to be the same in noise performance. With the Panaflo fan swap, the Seasonic 350 is about as quiet as any fanned PSU available, several decibels quieter than the stock model: virtually inaudible even at high system stress. ARM Systems says the noise level of their standard PSU option, the Zalman 300, is very similar because it is also modified with a Panaflo fan swap.

Hard Drive and Mounting

Most readers of SilentPCReview already know that the Seagate Barracuda IV single-platter model used in this system is the quietest HDD money can buy today, with the possible exception of the newer Barracuda V. The main issue with a drive this quiet is the noise created by transfer of mechanical vibration into the case. (For more details on this problem, see the discussion in Recommended Hard Drives.) This issue is tackled neatly by the soft purpose-designed grommets used to mount the drive.

The grommets are combined with drive bays that have mounting holes specifically designed to accommodate the grommets. One of the two detachable U-shaped drive bays in the case is shown here. (This one comes from the bottom front of the case.) The width of the bay is slightly larger than 3.5 inches. The standard small mounting holes are inset to match the drive width. The larger notched holes, on the other hand, are offset slightly, making a bit more room to accommodate the additional room needed for the grommet. Simple and effective.

To assess the effectiveness of the grommet mounting, I used the system as delivered for a while, then reinstalled the Barracuda IV in the same drive bay conventionally with screws, without the grommets. The noise difference was noticeable only during seek/write, but that difference was significant indeed. Without the grommets, the vibration of the drive during seek/write could be easily felt with a hand on the top of the case. It was also transmitted to the floor and could be heard as a kind of thrumming sound.

Measured with a Heath sound level meter one inch from the top front edge of the case, the difference between grommet and no-grommet mounting was only ~3 dBA at best. This measured difference is smaller than what I heard. Subjectively, the improvement with the grommets is such that once you've heard it, you would not accept the noise level without the grommets. On a carpeted floor the Barracuda IV drive noise just about disappears. On a hardwood floor, it is audible as a kind of fluttering sound. The vibration from drive seek on the top of the case is also much reduced.

Case and Heatsink Fans

The Panaflo FBA08A12L1A is our own reference fan. At 12V, it is quiet compared to most 80mm fans. With the Zalman Fanmate on each of the four Panaflos outside the PSU, they can be turned down to a virtual whisper; at the minimum of 5V, each one probably makes no more than 12 dBA @ 1 meter. All four of them together would add up to around 18 dBA. Add the one in the PSU and you might have 20 dBA in total. At the highest 11V setting, based on the manufacturer's 21 dBA rating @ 12V, all five fans together would be 26~27 dBA. This is much quieter than the fans noise of most PCs.

CD-RW Drive

The Plextor PX-W1610A is a quiet unit. Slow by the >40X speeds typical of current CD writers, but plenty fast enough for general use. It performed well for both reads and writes. In general, the noise level of CD drives is directly proportional to its speed. CDs are intrinsically imbalanced. At high speed, they always wobble. That fact combined with the air turbulence of the spinning disk makes them almost impossible to quiet except by slowing them down. Are you surprised that limiting the maximum speed of reads and writes is a feature now found in CD Burn software such as Nero? (Speed them up and make them noisy so you can slow them down to make them quiet?)


Anechoic Chamber Measurements(!!)

There was a brief opportunity to take noise measurement of the system in the anechoic chamber at the University of British Columbia, due to my involvement in a fan research project. The conditions were not ideal. Some heavy machinery was being operated in the building, and the background noise level in the chamber was 20 dBA (sound pressure level or SPL), a bit close to the levels we might encounter with this system.

The Stealth PC was placed on a stool. A sophisticated handheld sound level meter (SLM), accurate to below 20 dBA and valued at $7000, was calibrated moments before each of the two sets of measurements:

  1. All 4 fans set to max speed via Zalman Fanmate (~11V). Fan inside Seasonic PSU idles at 4.3~4.5V
  2. All 4 fans set to minimum speed via Zalman Fanmate (~5V). Fan inside Seasonic PSU idles at 4.3~4.5V
Fan Settings
1 M from front
1 M from right side
32 dBA
28 dBA
23 dBA
22 dBA

The second test results were so close to the ambient noise that it is possible the actual PC noise was lower than measured. Note that the drive remained at idle throughout testing.

Subjective Listening

With all the fans set to minimum speed, the ARM Systems Stealth is a quiet machine. Compared to any store-bought PCs I've encountered in the last couple of years, it is substantially quieter, and the level of drive seek noise is notable by its virtual absence. The primary noise is the low level whooshing of fans, with no high pitched sounds at all, along with a trace of lower frequency noise from the hard drive, audible only on a hard uncarpeted floor.

When the Fanmate on the HSF is turned up to maximum (11V), the system is still pretty quiet. My guess is that in the anechoic chamber it might measure 24-25 dBA SPL @ 1 meter. Most visitors who heard it in operation commented on its low noise even in my acoustically lively test lab -- which is now a converted, otherwise-unused kitchen with vinyl tile flooring.

It is noisier than my self-assembled machines, which probably measure 4~8 dBA quieter, but that is to be expected. My PCs cost an absolute fortune in sheer time, never mind the parts, and are likely to be less stable under high duress. I don't bother torture testing them; CPU stress testing is unrealistically strenuous for my needs. When hot weather comes, I usually end up under the hood hacking and adjusting to ensure better cooling. And I would not, could not, ship them as they are.


There is some of debate on the MSI forums whether this motherboard actually reads the temperature diode in the XP processor. The general consensus is that it is supposed to. I think it is, judging by the relatively quick CPU temperature swings.

Under 100% CPU load with stress testing, the minimum voltage setting (5V) for the heatsink fan may not adequate to keep the CPU comfortably cool. In a room ambient temperature of 18-20C, the CPU temp reported by Motherboard Monitor 5 (MBM5) stabilized at 64C after 20 of Prime95. The case fans and the VGA cooling fan were all at the minimum 5V.

I use the phrase comfortably cool because even though 64C seems high, the AMD XP is not in mortal danger till the core hits around 90C. Keep in mind, too, that 20~30 minutes of constant 100% CPU utilization is much more stress than most people will apply to a system in normal use. Even 3D shooter games don't do that... or do they? (I'm sure someone will correct me if I am wrong...) Most DIY silent computing enthusiasts are used to higher temps, anyway.

Of course, there's no need to keep the CPU HSF at 5V if you're not happy with the resulting temperature. Just turn the Fanmate knob up a notch!

In the 20C ambient temperature of the testing room, the best balance of low noise and stability under Prime 95 (or similar CPU stress testing for long periods) is with the CPU HSF Fanmate at ~7V and the three other fans in the case at minimum (5V). With this setting, after 30 minutes of Prime 95, the CPU temp reported by the motherboard stabilized at 57C.

The noise difference between the minimum 5V and the 7V setting for the CPU fan is not audible, likely swamped by the 4 other fans spinning at 5V. Even going to 9V, the added noise is only just perceivable at the nominal 3~4 feet distance if you place the PC on the floor under your desk like most folks.

The PSU fan may have speeded up a bit and become nicer during the stress test. The maximum steady AC power drawn was 140W, which translates to just under 100W of delivered DC power. AC power did surge to as high as 155W on momentary peaks. If the fan did speed up, the difference was too small to be noticed.

Temps with case fans set at 5V in 20C room, after 20+ mins of Prime95:

CPU HSF voltage*

*NOTE - about setting fan speed: Most people do not have the resources to check the voltage fed to the fan, and the Fanmate1 does not even have a calibrated scale. So how can the above data be used? I suggest MBM5 be installed, then set the Fanmates on case fans and VGA fan to minimum. The Fanmate for the CPU HSF should be set to max. Work the control down slowly till you hear the first significant decrease in fan noise. That is usually 8~9V. It is probably a good place to leave it. By considering the three factors ? CPU temp, your perception of noise and the Fanmate control knob ? you will be able to find a happy balance that is right for you. And nothing says you can't take days or weeks to find that happy medium.

With extreme room temperature, the case fans may need to be increased in speed. No instability was noted on video even with the fan over the VGA turned completely off.

The hard drive temperature never exceeded 37C at any time, even at one point when room temperature rose to 27C. The location of the hard drive, just above the front case fan, ensures a steady flow of air that keeps it cool.


These benchmarks were all run with the motherboard at the MSI default optimized BIOS settings. As expected, the results were squarely in line with other similarly equipped PCs.

MadOnion 3DMark 2001
MadOnion PC2002







In general, the system felt fast and responsive, easily the equal of the P4 (overclocked to 2.2G) I use as my main PC these days. Another 256 MB of RAM would help with multitasking and large image files, but this is obviously a simple option that can be implemented at any time.


The core membership of SPCR are willing and able to hack and mod to achieve the goal of PC silence; many others are not. In the Stealth series, ARM Systems has created an attractive, viable option for those who seek a ready-made quiet PC. A wide range of processor speeds and other options makes it possible to custom tailor the PC precisely to your needs. A nice bonus, as many quiet PC makers do not offer such a broad choice.

Given the fine cooling performance with the XP2000+ Palomino core, I would expect the AMD Stealth systems to ramp up nicely, albeit at the cost of graduated increases in noise or heat, all the way to the XP2600+. The P4 Stealth systems are similarly configured; the main difference is the use of a Zalman 5500CU HSF and Zalman 80mm fans as standard or an Alpha PAL8942 heatsink with all Panaflo fans option.

Some pros and cons. (Well it's true I had to stretch just a bit to fill up the Cons column.)

Pros Cons

System is quiet and stable

Not as quiet as quietest DIY efforts

Quiet PSU and fans

Could use better damped feet

Great case design

Case not sexy enough

Secure noise reduced HDD mounting

Case doesn't allow larger fans

Speed of each fan individually adjustable

HDD grommets not as quiet as suspension

Nothing to stop a modder from making further reductions in noise

Too easy

Preloaded OS a nice luxury

No Linux option -- oh, I've been corrected, there are several flavors of Linux!

Easy to obtain: just order (if in US or Canada)

Might be too noisy with faster, hotter CPUs?*

Wide range of options and prices

Too easy?

*With faster CPUs: ARMS says up to the fastest P4 and XP CPU machines currently offered, when the Alpha HS + Panaflo fan option is used, the CPU fan may have to be set to 10-11V, but the other fans are OK at 7~8V. This will result in only a small increase in noise from the review sample quiet / cool optimized settings.

Seriously, it's a quiet, solid machine at a decent price.

Much thanks to ARM Systems for their patience with this review. (But it is our first system review, after all...) The machine finally gets to go home now.

* * * * *

Discuss this article in our Forums.

Previous 1 2 3

Complete|Mobile Systems - Article Index
Help support this site, buy from one of our affiliate retailers!