HP Proliant MicroServer

Table of Contents

The MicroServer is a unique SFF server in the HP Proliant line that looks more like a small PC or multi-HDD NAS box. Powered not by Atom but a more powerful dual-core, 12W TDP AMD Athlon II, it is the smallest server we’ve ever seen. We gave our sample the complete SPCR treatment.

April 13, 2011 by Mike Chin

Proliant MicroServer
Street Price
Starts at $300~350 w/ 1GB RAM, 250GB

The HP Proliant MicroServer caused a ripple of excitement among some SPCR forum
members when it first came to light late last year. Ostensibly a part of HP’s
extensive server line, the MicroServer is aptly named, being truly small, and
nothing like the big noisy racks the term server brings to mind. It is
much closer in size and price to 4-HDD Network Attached Storage (NAS) boxes
or Windows Home Server servers (like the now discontinued HP MediaSmart Servers)
but with the promise of considerbly higher performance hardware than the Intel
Atom or ARM chips most often used in such boxes.

These snippets from the
press release at the launch of the MicroServer
make HP’s target market crystal

"Designed for companies with less than ten employees, the HP ProLiant
MicroServer enables them to share and secure critical business information.
Ideal when needing to stay connected, whether in the office or on the road,
the server simplifies how users access data differently than on desktop or
notebook PCs.

"The HP ProLiant MicroServer’s compact, quiet and sleek design
is about half the size and 50 percent quieter than most entry-level servers.
It also offers lower power consumption for energy-conscious businesses.(1)

"HP has the top position in the x86 server market with 39.2 percent
factory revenue share, according to IDC’s Worldwide Quarterly Server

"Based on current buying trends and the continued advancement of technology,
approximately 1.7 million small businesses will purchase their first server
in the next five years, according to AMI Partners, a global strategy consulting

" ‘Small businesses have traditionally relied on interconnected PCs
to share files and resources, which doesn’t help them keep pace with
a competitive environment,’ said Jim Ganthier, vice president, Marketing,
Industry Standard Servers, HP. ‘Leveraging over 20 years in server leadership
and innovation, HP is introducing a no-compromise, compact server designed
to fit into office environments with a versatile appearance and quiet operations.’

It is a server marketed to small and home businesses, but some of its features
are obviously attractive to sophisticated home network users as well.

The starting price for a base unit with 1 GB RAM and a single 160 HDD was a
modest $300 at launch, making it easily affordable for most home users. Currently,
the MSP has risen to US$350 (with a larger 250 GB drive, it seems) presumably
due to the weakness of the US dollar. In the UK, £100 rebates from HP
are being offered during the first quarter of 2011 (extended to end of April),
making the current £209 price a very sweet deal.

The key features that makes this product enticing for energy- and noise-conscious
users include:

  • Low starting price
  • Very low power (12W TDP CPU)
  • Potential for low noise (just two fans, a 120mm and a smaller one)
  • PCI-E x 1 slot for TV card
  • PCI-E x16 slot for discrete graphics
  • Full size DDR3 slots (not expensive laptop or server) for cheap upgrade
    to 2 or 4 GB
  • ECC RAM support (not found in HP Home Servers)
  • 4 HDD easy to use "pluggable" drive bay (not hot swap)
  • Sturdy, tidy build quality — even the motherboard slides out easily
    on a tray
  • Internal power supply rather than yet another AC/DC adapter to add to the
  • Full size optical drive (for lower cost)
  • Gigabit ethernet

The processor in the MicroServer is an AMD Neo N36L, a low power (12W TDP)
variant of the Athlon II designed for embedded computing. That’s right, this
dual core 1.3 GHz CPU is soldered to the motherboard; the absence of a CPU socket
is part of the reason for the low cost.

SPCR received a review sample with 2 GB of non-ECC RAM, preinstalled with Microsoft
Windows Server 2008: R2 Foundation preinstalled on one of two 160 GB Seagate
7200 RPM hard drives. A keyboard and mouse were also supplied, but the MicroServer
is meant normally to run headless via another PC in the network. This configuration
is a significant step up from the base product. The Windows Server 2008: R2
Foundation operating system adds around $200 to the overall cost, and the additional
software might cost an additional $100 or so.

A 160 GB drive is highly restrictive by today’s standards. Some businesses
might be OK with such a small amount of storage per drive in a server, but at
a time when 2 TB drives routinely sell for no more than ~$75, 160 or 250 GB
drives seem like throwaways. It also seems remiss not to take advantage of the
MicroServer support for ECC RAM, which does improve data integrity. Why not
offer it without HDDs or RAM and let the users choose for themselves?

Part of the answer could be that the MicroServer is meant to be sold by HP
or its resellers configured as a server with an operating system preinstalled.
In other words, the base price for the near-barebones system is a gesture for
die-hard DIYers only.

HP MicroServer package examined by SPCR’s Maxi Miss Kitty. The second
box contained HP-branded USB keyboard and mouse.

The main box held the MicroServer, the AC cord and optical discs for
HP documentation and drivers, and the Microsoft Windows Server 2008: R2
Foundation OS software.

Review Methodology

We have covered the topic of home servers in the past, but we have never
actually covered a server, per se. The MicroServer isn’t meant a home server,
but HP discontinued its popular WHS-based MediaSmart Server line at the end
of 2010, with some web speculation that the main cause was the elimination of
Microsoft’s user-friendly Drive Extender HDD/storage management in the 2011
edition of WHS. Given that the MediaSmart Servers were probably the most successful
WHS-based products around, their departure likely leaves a gap in the marketplace.
It’s easy to envisage DIY users installing the earlier version of WHS on the
MicroServer, even with the prospect of the eventual cessation of Microsoft support
for the old software.

In any case, there are enough different server packages to make us complete
newbies with them. The various popular distributions of Linux, FreeNas, and
Windows offerings can all be run on the HP Proliant MicroServer. Server experts
we are not, but we do have a pretty good handle on computer hardware in general,
on acoustics and on thermal and power issues, so as you might have expected,
this review will focus primarily on hardware. Naturally, some references will
be made to our sample and its Windows Server 2008: R2 Foundation system.


We begin with a rundown of features and technical specifications. The information
below is from HP’s web page about the MicroServer, and the many supporting documents
for the product.

HP Proliant MicroServer Features (w/ our comments)
Packed with all the essential server features, the
MicroServer is a perfect first server

++ Designed for growth, you can add memory, hard drives and peripherals
with lesser effort. (True, compared to more conventional servers.)
++ Intended for small and micro organizations—its compact design fits
easily within your work space. (It is very small.)
++ Suited for an ergonomic working environment, it offers a low acoustic
level of 21 DBA. (We’ll check and let you know.)
Proven HP ProLiant reliability and support
++ Leveraging 20+ years of server experience, HP now offers a simplified
solution for small and micro businesses—HP ProLiant MicroServer.
++ Offering a worldwide network of HP trained service professionals—you
can take advantage of HP services and support.
HP Proliant MicroServer Specifications
Processor Dual
Core 1.3 GHz AMD Athlon II (Neo N36L)
with 2 MB cache
Memory PC3-10600E DDR3, operating
at maximum of 800 MHz
Memory slots 2 DIMM slots
Maximum memory 8 GB (2 x 4 GB) (Our
sample came with 2 x 1 GB, non-ECC
Advanced memory protection Unbuffered ECC Storage
Maximum internal drives 4 pluggable drives (Our
sample came with 2 x 160 GB
Maximum internal storage 8 TB (4 x 2 TB 3.5”
SATA drives)
Storage controller Embedded SATA controller
RAID 0/1
Expansion slots Expansion 2 PCIe (x16 and
Form factor Ultra Micro Tower
Dimension 10.5 x 10.2 x 8.3 inch
(26.7 x 26.0 x 21.0 cm) H x D x W
Power supply 200 Watt
Network Controller Embedded NC107i PCIe Gigabit
Ethernet Server Adaptor
Remote management Optional Remote Access Card
(like a real server!)
Optical drive Optional (Our sample
came with a standard DVD/CD writer)
USB 4 front, 2 rear, 1 internal
+ 1 eSATA
Warranty (parts/labor/onsite) 1-year parts, 0-year onsite
support, 0-year next business day response
OS options * Windows Server 2008 R2
Foundation (installed on our sample)
* Windows Server 2008 R2 Standard
* Red Hat 5

The operating system options leaves no doubt that the MicroServer was intended
to be a small business server, not a home server. The promotional data sheet
sings the sharing, security, and expandability benefits of the MicroServer over
a peer-to-peer network.


The fascia is not dissimilar to that of bigger breadbox style PCs,
like the LianLi PC-Q08 we reviewed
last Sept
. The depth of the MicroServer is far shorter at just 10".
The overall volume of 14.5 liters makes it a good 25% smaller than the
PC-Q08, but a bit larger than the 11 liter Silverstone
SG05/SG06 SFF mini-ITX cases
. Its footprint small at just 10.2"
x 8.3". Note obvious venting across the lock-equipped front door,
and 4 USB ports with power button on the top right edge.

The back panel sports a 120mm exhaust fan for the main components and
drives, and a tiny 3~4cm fan for the power supply. The latter is not promising
for low noise. There are two half height expansion slot openings, an eSATA
port, two USB 2.0 ports, gigabit ethernet port, and a single D-sub VGA
connector. A single captive spring loaded thumbscrew (our favorite kind!)
provides access under the top cover.

Back to the front: Unlocking the dust-filtered, vented, hinged front
door gains access to the four HDD bays, and the motherboard mounted on
a sliding tray at the bottom. Wiring is tidily arranged, with reusable
plastic locks for them on either side of the case.


Slide the top cover forward and up once the thumbscrew is loosened, and it
comes off easily. The front door lifts neatly off its hinges when the top cover
is off. Most of the screws used in the MicroServer are Torx types, and amazingly,
an L-shaped wrench screwdriver with two different size Torx heads is included,
clipped in a purpose-built spot on the inside of the machine’s front door.
(Mind you, using it to remove the 120mm fan screws proved slow and painful to
the fingers; the lab’s cheap set of Torx screwdrivers with proper handles proved
much easier to use.)

The quality and thoughtfulness of the MicroServer’s construction and design
is excellent. Everything fits well in modular fashion, and the double-layer
steel panels provide an assuring weight and sense of strength. It is clear that
this product has more in common with real servers than with the consumer-oriented
SFF PCs or NAS boxes it resembles on the outside.

Here’s a view with both the top and front cover removed, with one drive
partially slid out. The plastic clamping mechanisms work well enough,
and a SATA plane on the inside accepts the bare connections directly when
a drive is slid into place. As with a hot swap bay, there is no need to
fiddle with connectors and cables.

A WD Green and the frame used for mounting. The HDD attaches to the plastic
frame, which seems sturdy enough, with 4 ordinary drive screws.

The 120mm fan slips out easily once the four Torx head screws securing
it are removed from the back panel.

The 120mm fan is a Delta, rated at 0.6A. A web search turned up a Delta
spec sheet that listed the AFB1212VH as a 0.4A fan with max speed of 3100
RPM and 44.5 dBA SPL. Since ours is rated at 0.6A, it could be an even
faster, noisier variant.

All drives removed. Small PSU uses 4-pin Molex connectors to power
the SAS backplane. Venting on the back of the HDD cage is decent, as is
the space between the drives. Even when all four bay are populated, cooling
airflow should be good enough. Optical drive slides out easily with single
lock snap. Two blue captive spring loaded thumbscrews, one on either side
of the front edge, are all that need to be undone to remove the motherboard
— that plus removal of all the connectors to the board.

The motherboard pulled out just far enough to remove all the connectors
so that it can slide all the way out.

The motherboard is about 16 x 20 cm (W x L), which is close to the
20 x 17 cm (W x L) of mini-DTX. It appears proprietary. The Athlon II
Neo CPU is under the large center heatsink. The single SATA port is used
only for the optical drive. The 4 HDDs connect to the server backplane,
which is connected to the motherboard through the mini SAS port (near
the bottom right corner in the photo above).


The BIOS provides hardware monitoring of temperatures, fan speed and voltages,
but no user control. Fan speed control is automated, tied to a temperature sensor(s?)
somewhere within the system. Given these conditions, baseline system noise was
defined to be the MicroServer running with no hard drives. In actual use, this
server is not likely to get the extremely high, extended CPU or GPU loads used
routinely for benchmarking desktop PCs; idle or near idle conditions are likely
to predominate, which means the fans are not likely to ramp up in speed often.
The more significant variable is probably going to the noise made by the hard
drives, in idle and in seek.

System Configurations

As delivered with 2 GB RAM and…

  1. no drives (Baseline)
  2. 2 Seagate Barracuda 7200.12 160GB SATA drives
  3. 2 Western Digital Green 2 TB AV drives (5400 RPM)
  4. 4 Western Digital Green 2 TB AV drives (5400 RPM)

Measurement and Analysis Tools


The sample HP MicroServer, as delivered with MS Windows Server 2008: R2 Foundation
installed, was hooked into the SPCR gigabit wired network, which has 4 unmanaged
switches. It was used for a week mainly as a file server and backup, roles the
MicroServer fulfilled smoothly without any issues. Transfer speeds varied depending
on the type of files and the other machines involved, but it was as speedy as
any other system in the network, reaching a maximum of 100~110 MB/s. We did
not see any significant benefit from the dual 7200 RPM drives compared to the
mostly 5400 RPM WD Green and Samsung Eco Green HDDs running on the network storage
boxes and PCs. The unit was placed in a closet, where a fanless single 2TB HDD
NAS box resides. This system is inaudble outside the closet; the Seagate-equipped
HP is audible, though subdued.

Other functions such as e-mail, database or intranet Web server were not explored.
We have no doubt that the MicroServer can handle all such functions with the
user number suggested by HP: 10 or fewer. Users will likely experience slowdowns
at the higher end of the recommended number.


1. Baseline Noise: This is defined as the lowest possible noise level
of the system. Adding any other components can only increase the noise. The
hard drives were removed, and the system booted into the BIOS menu. It was left
running for about five minutes before any measurements were taken.

23 dBA @ 1m in our anechoic chamber.

Our baseline measurement is 2 dBA higher than HP’s claim of 21 dBA for the
MicroServer, which is pretty close, given that HP used the >1m ISO 7779 "By-stander"
position for the mic. The overall noise level is modest, and the noise quality
is not bad. Though it has distinct tonal aspects and is several decibels louder
than the noisiest PC in the lab (~16 dBA@1m) we could live with the noise.

2. As Delivered, with two Seagate Barracuda 7200.12 160 GB HDDs: Surprisingly,
the measured SPLwas almost identical, whether one or both drives were engaged.
The difference in perceived noise between no drives and with these drives running
was night and day, however. The acoustic signature of the sample as delivered
is best described as extremely tonal, annoying and difficult to tolerate for
any length of time in a quiet environment. Perhaps in a busy office with lots
of other machines running, its noise would fade into the background buzz, but
in our quiet lab, no way. The fault lies much more with HP’s choice of hard
drives than in the MicroServer.

Still 23 dBA @ 1m but with a completely different sonic signature.
Without the Seagate drives, there was no output above 3 kHz. With them,
the upper range is riddled with highly audible tonal spikes.

3. With two WD Green 2TB HDDs: This was to confirm our perception that
the Seagate Barracudas caused much of the annoying acoustic character of the
sample system, as delivered. WD Greens are currently our favorite quiet HDDs,
widely available at low prices, and used by many SPCR enthusiasts. As expected,
with the WD Green drives, the overall acoustic character became much closer
to the baseline, even though the measured SPL was slightly higher than with
the Barracuda drives. Again, while the overall level is a bit higher than we
prefer, we can live with this noise.

With WD Greens, the same measured SPL as the Barracudas, but virtually
no increased high frequency noise. The recording shows slightly higher
broadband noise but no increased tonality over the baseline noise.

4. With four WD Green 2TB HDDs: The measured SPL barely went up, to
24 dBA, and the subjective impression was the same. The fact that the noise
level rose so little tells us that other sources of noise in the system set
the noise floor, not the very quiet WD Green drives.


Normal fan speed monitoring proved unachievable. Although the fan RPM showed
up in the BIOS, no software utility could access the sensor in MWS 2008, or
later, in Windows 7. The SPCR
forum discussion about the MicroServer
pointed to a reason why this might
be so. Although the HP uses a 4-pin WPM fan, the pinout configuration is non-standard.
The software utilities are programmed to read the on-board terminals in a standard
configuration, thus cannot get the data correctly from the HP board.

We resorted to tagging one of the blades of the fan with a piece of white plastic
tape so that it could be monitored optically with an IR tachometer or calibrated
strobe through the back panel grill while the system was running, under different
loads. This was a slightly tedious task.

Prime95 was use to stress the CPU. Various tools to stress the HDDs were tried,
but we found none that could work all the drive at once, for any length of time.
HDD Scan was identified as the most promising tool for our purpose, but again,
it could not run any high stress test on the drives simultaneously. Even when
run for more than an hour on just one drive, none of the utilities raised the
HDD temperature beyond 37°C, the highest seend during Prime95 testing.

Windows 7 was installed, on a WD Green drive, simply to expedite testing.
The security features in MWS Server 2008 makes it odious to turn the system on/off
as many times as we must do in our intensive hardware testing. All the hardware
montoring runs on MWS Server 2008 as well as on Windows 7; the same temperature
results were seen in both OSes.

All of the Prime95 results were recorded after 60 minutes of continuous load.

HP MicroServer: Cooling Results
with 2 Seagate Barracuda 7200.12 160 GB drives
(MS Windows Server 2008)
23 dBA
23 dBA
with 2 WD Green 2TB AV drives
(Windows 7)
23 dBA
23 dBA
with 4 WD Green 2TB AV drives
(Windows 7)
24 dBA
24 dBA
* The temperatures of the hottest and the coolest

The 120mm fan starts at an elevated speed for a few seconds (under 10)
when the system powers up, but it quickly settles down to a slower speed. In
MWS 2008, the fan speed never changed in any of our testing, staying at a constant
1060~1100 RPM. The room temperature never exceed 24°C throughout the testing;
it’s possible that in hotter weather, with hard drives that draw more power,
the fan could speed up. This aspect was not checked. The low temperatures suggest
that the 120mm fan is providing much more airflow than is actually needed in
most conditions.

The 40mm fan in the power supply might have sped up or down a bit during testing,
but it was difficult to be sure with the hard drives and the 120mm fan all running.
The small fan obviously did not introduce much — if any — audible

One important note: Stopping the 120mm fan or unplugging its connector from
the motherboard immediately shuts the MicroServer down. The safety-conscious
speed sense feature appears to be undefeatable.

The component temperatures remained perfectly safe under high load.
Hard drives are specified for safe operation to 55~60°C, and the AMD Neo
CPU is rated for 95°C maximum, so there is plenty of headroom for much hotter

The power draw of the system is quite modest. With two WD Green drives,
the power draw ranged 24~35W in normal use, with the average not much higher
than 32W. With the more likely 4-drive setup , it went up about 10W, to around
40~43W average.

Surpringly,with the two Seagate Barracuda 7200.12 160 GB drives, the server
idled (in MWS 2008) 9W higher than the two WD Greens. Peak power went up by
the same amount, so it seems clear that the Barracudas are the cause the of
extra power consumption. However, note that the idle CPU temperature in MWS
2008 was some 6°C higher than in Windows 7. So…

Windows 7 w/ 2 WD Green & 2 Seagate 7200.12 drives
23 dBA
23 dBA
* The temperatures of the hottest and the coolest

Another test was run, this time with the Window 7 loaded WD Green as the boot
drive, another WD Green and the two Barracudas. This showed 5W higher idle power
than with all 4 WD Greens, about what you might expect. CPU activity at idle
was virtually zero as in all the previous tests. But the power draw at full
CPU load was the same, which is odd. We expected the extra power for the HDD
to add to the total AC power at high CPU load, too.

The best way to resolve this little puzzle would be to install Window 7 on
one of the Barracudas, and MWS 2008 on one of the WD Greens, and recheck the
power with both… but at this point, the exercise became too exhausting. Suffice
it to say that most SPCR readers are not interested in any drives that sound like these Barracudas, and the overwhelming HDDs of choice for most users will be the Samsung, WD or Seagate "eco" drives which run quieter and only a bit slower at 5400 or 5900 RPM.


The HP MicroServer is an unsual product in many ways. Its size is unique among
servers, and it is much closer to 4-drive bay NAS boxes in this regard. The
latter usually run Atom or ARM processors, which do not have the processing
power of the Athlon II Neo, even at its slower 1.3 GHz clock speed. The design
and build quality are excellent, and it really is small enough to be tucked
on a shelf, a corner of a desk, or in a closet. Its power profile is low enough
that even with four 2TB drives and high system load, the AC power is unlikely
to crest 60W peaks and average closer to 45W. Finally, its 120mm primary cooling
fan and the little 40mm fan in the PSU are quiet enough that the choice of hard
drives will have more effect on the perceived overall noise. It isn’t quiet
enough for us to blithely place atop a nearby desk, but in almost any small
business office, it should be very unobtrusive.

The real question is whether the MicroServer really meets the needs of the
intended small business market. The low entry price and small size are certainly
attractive, but does it represent an ideal choice for a small business moving
up to its first server?

Cheap servers are available from most of the major computer companies, but
none are so small and very few start at $300, $350 or £109 (after cash
back in the UK). The inclusion of a gig of RAM and a 250 GB drive seems silly,
though; most users will end up dumping both, for 4GB RAM and a couple of 2 TB
drives, at least.

The big issue for the small business target market is the absence of a user-friendly
operating system. While something like Ubuntu
might be a perfectly plausible option for a tech geek, most small
businesses do not have the qualified personnel to tackle installation, setup
and most importantly, maintenance of a Linux server. This might also be true
of Windows Server 2008 R2, even the $200 entry-level 15-user Foundation
. A new MS offering called Windows
Small Business Server 2011 Essentials
might be better suited, but its $250~300
online price for a 5-user license is close to that of the bare server.

Realistically, with a minimal setup of two 2TB drives (one for OS & data,
the other a mirror or backup), 4GB ECC RAM and a familiar MS-based server OS,
the real base price of a MicroServer in the small business role is around $800.
This is for a server capable of running centralized software, serving multiple
clients. It is getting close to a more powerful and expandable conventional
server, which might be a better choice if rapid growth in needs is anticipated.

If the role is a simple, centralized, secure file server, a better choice might
be a NAS box with a built-in, customized firmware OS that is simpler to use
and maintain than a server. The 4-bay QNAP
, for example, sells for under $500; with dual 2TB HDDs, it should
match the MicroServer in this file server role. Still, the open-source FreeNAS
OS on a MicroServer may well outperform the Atom-based QNAP for tech-savvy users,
except that the QNAP has RAID 5 functionality built in, not just the RAID 1
and 0 of the MicroServer.

All this is to say that while the MicroServer is a clever, well-engineered
product that will please a lot of people, it’s not a slam dunk choice for a
small business. There are many factors to consider.

For the home geek, the HP MicroServer is a great toy or tool. It can be a more
powerful Windows Home Server platform than the discontinued HP MediaSmart Servers.
For the more adventurous, there are the many Linux variants, available free
for download and experimentation. It can also be treated like a barebones SFF
PC. There’s the option of running the MicroServer as a combination file server
and media PC. The HD4200 integrated video is good enough for HD video play,
and the Neo CPU should be perfectly capable of HTPC duties. The two expansion
slots give it expandability as well. You can even make it SPCR-quiet with a
swap of that overzealous 120mm fan. For details continue reading. Note: Audio
recordings are on the next page as well.


Answers to some questions sent off to HP arrived just as this article
was being posted.

Q1. Is there anything in the US or Canada (or anywhere else) like
the £100 cash back in the UK? If not why not? Why in the UK?

A1. Special promotions such as the £100 cash back in the
UK are done at the country level and are typically in response to local
market conditions.

Q2. It’s been selling now for 4-5 mos. What is the typical buying
pattern? Is the base configuration the most popular? Or are there more
sales of more fully decked servers sold through resellers? I’m interested
to know whether you hit your target market (small business <10 users)
or whether the enthusiast market has been more responsive than expected.

A2. With the MicroServer, HP has created a new category and
initial offering is a base model targeted at resellers. Basically, the
MircoServer is sold primarily through resellers who perform configuration
and installation based on customer needs. Anticipate resellers to fulfill
the OS install function

Due to the large uptake in the base model, we are evaluating offering
additional configurations targeted at different segments.

Q3. Why is the fan configuration non-standard? That is, if a fan
burns out it cannot easily be replaced with one from normal outlets (ie
Frys) due to the plug incompatibility.

A3. Because of its small form-factor and acoustic goals, the
MicroServer requires a fan that is powerful yet quiet. This fan was
selected because it met the MicroServers stringent system specifications

Our thanks to HP
for providing us with the review sample.

* * *

Articles of Related Interest
Home Server Build Guide: SFF Configurations

HTPC Home Server
Mid-Tower Home
Server Configurations

* * *

this article in the SPCR forums.


It was inevitable that SPCR modify the MicroServer for lower noise. What are
the usual methods to reduce noise in an existing computer?

  • Reduce the transfer of vibration from the hard drives to the chassis.
    This is not really possible here, given the plug-in design of the HDD
  • Lower fan noise by reducing RPM or replacing fans with slower, smoother
    This is not a practical option with the PSU fan, which made very
    little acoustic contribution anyway. The 1100 RPM minimum 120mm fan is a perfect
    target for this approach.

We mentioned that this primary cooling fan is a 4-pin PWM type with a pin-pattern
that is non-standard. This means you cannot plug any old PWM fan into the CPU
fan header on the MicroServer meotherboard and expect it to work. Our first
step was to examine the connector. A close comparison revealed that it is not
quite the same as the 4-pin connector on other PWM fans used in PCs. The latter
is designed to fit either 3-pin or 4-pin fan headers; the HP connector can only
fit on a 4-pin header.

We also examined the information that SPCR forum members unearthed. Much thanks
to vlrnmz and camlin for the
information they shared
. Here is the 4-pin configuration information for
the HP MicroServer from camlin:

HP MicroServer
4-pin CPU fan header
Pin Name Color
1 Control green
2 GND black
3 +12V red
4 Sense yellow

What you need to know now is the normal pinout for 4-pin PWM
fans. From AllPinouts:

Pin Name Color
1 GND black
2 +12VDC yellow
3 Sense green
4 Control blue
Pin Name Color
1 GND black
2 +12VDC red
3 Sense yellow
4 Control blue
Header on mobo
For some AMD CPU fans

So armed with this information, a couple of possible Scythe PWM fan candidates
were identified, the connectors compared, and modifed. This requires a very
small screwdriver to unlock the conductor pins from the connector and swap them
around to the desired slots. The center guide on the plastic connector also
needs to be removed, as it will stop it from fitting on the HP fan header. This
was accomplished with a sharp utility knife.

The labels on the photo are self explanatory. The Scythe fan color
code is different from either of the ones described by AllPinouts.

Rather than installing the fan, it was tried first, by plugging it into the
fan header with the chassis door open, then powering up the system with the
fan on the outside. The first fan to be tried was a Scythe SY1225SL12LM-P. This
is a PWM version of the popular SlipStream series, with a rated maximum speed
of 1300 RPM. After the initial full speed burst at turnon, the fan slowed to
an extremely slow speed, which caused a fan error message to display and the
system to power down immediately. The system wants to see a faster fan. Since
the >3000 RPM stock fan ends up spinning at 1100 RPM, the PWM fan controller
applies a duty-cyle of about 35%. That meant the Scythe must have been spinning
at under 500 RPM.

The next fan to be tried was a Scythe SY1225SL12HPVC. This is a PWM fan rated
for 1900 RPM, equipped with a manual speed control. With the control set to
full speed, it would surely run fast enugh (~700 RPM?) for the system to boot.
And indeed, it worked.

The Scythe Slip Stream120mm PWM Adjustable fan on the left worked as
a replacement fan in the HP MicroServer. The 1300 RPM Scythe PWM fan did
not run fast enough.

Then came the task of removing the stock fan. After the four Torx screws are
removed from the back panel, follow the course of the cable to the motherboard
header and carefully snip several plastic wire ties. The Scythe fan has a much
a shorter cable, so it had to be run a bit differently. This required the removal
of the motherboard. Finally, the dongle with the PCI slot cover plate and rheostat
had to be put somewhere. The control could be removed from the plate and mounted
somewhere on the back panel, but for testing, it was left resting behind the
optical drive. Be careful that the fan wires don’t catch on the fan blades.
The fan experiments could now begin.

The Scythe fan controller was temporarily tucked behind the optical
drive. Do not leave it this way.

Many drive combinations and loads were tried with various fan speeds, with
all temperature and noise data recorded. All that is not presented here,
to reduce information overload.

First, the manual fan control was used to determine the minimum fan RPM allowed
before auto-shutdown. The BIOS fan monitoring screen was left on while the fan
knob was rotated. It turned out to be somewhere around 500 RPM; the exact value
was difficult to pinpoint. It is probably safe to consider 550 RPM as the minimum

With the Scythe fan at 550 RPM, the baselline noise was just 16 dBA. However,
a tonal peak at ~2.7 kHz which had been masked by the stock fan before was now
apparent. This comes, evidently, from the power supply or its fan. It is not
clear which is the culprit without examining the PSU by itself. In any case,
since hard drives would always add to the overall noise, there is no point in
keeping the fan speed so low. It was raised to 700 RPM, bring the SPL up to
17 dBA@1m, and this became the default for SPCR’s Scythe-fan-modded MicroServer.

The Scythe fan @ 700 RPM brought the baseline noise down 6 dB from
the stock fan baseline.

Other smooth sounding fans could be used, of course, with or without manual
control. To achieve 700 RPM speed on the motherboard’s built-in fan controller,
if we assume 35% estimated fan duty cycle, you’d need a PWM fan with rated speed
of no more than 2000 RPM. Note, however, that our 1900 RPM Scythe spun at 1080
RPM with its manual control set to maximum, which implies about 50% duty cycle.
This means that the PWM controller in the board interacts differently with different

Here is the essential cooling and noise data for the 700 RPM Scythe mod.

HP MicroServer w/ Scythe PWM fan @ 700 RPM
Windows 7 w/ 4 WD Green AV drives
20 dBA
20 dBA
* The temperatures of the hottest and the coolest

HDD temperatures are not affected by the drop in fan speed, although the CPU
ran 6°C hotter, to a still modest 59°C. This kind of extended 100% CPU
load will never happen in a NAS or home server. The overall noise signature
is much nicer than the 4 dBA reduction would suggest. You still might not want
this system on your desktop, as HDD seek noise is clearly audible, but you’ll
probably be fine with it in the same room.

The Scythe fan @ 700 RPM brought the noise down 4 dB compared to the
stock fan.

Finally, some audio recordings of the MicroServer.


These recordings were made with a high resolution, lab quality, digital recording
system inside SPCR’s own 11 dBA ambient anechoic chamber, then converted to
LAME 128kbps encoded MP3s. We’ve listened long and hard to ensure there is no
audible degradation from the original WAV files to these MP3s. They represent
a quick snapshot of what we heard during the review.

Each recording starts with ambient noise, then 10 second segments of product
at various states. For the most realistic results,
set the volume so that the starting ambient level is just barely audible, then
don’t change the volume setting again while comparing all the sound files.

* * *

this article in the SPCR forums.

Silent PC Review is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn More

Leave a Comment

Your email address will not be published. Required fields are marked *