Building a Green Windows Home Server: The 37 Watt Server

Sunday, September 14, 2008 – 6:49 PM

Which uses more power, the AeroCool Green Home Server or a 40 Watt light bulb? You guessed it, the light bulb looses.

A few hours playing with a Smart-Watt meter hooked up to my home server I came up with some power usage numbers. While idle the Home Server consumes less power than a 40 Watt light bulb. It only consumes a few watts more under normal loads associated with backups and streaming data.

Mode Power Usage
Idle 37 W
Backing up * 44 W
Streaming music/data ** 42 W
Under full load *** 68 W (72 W peak)

The average use over a 24 hour period was just under 0.9 kWh. Huge sigh of relief! After all that guesswork and effort it would be horrible to end up with a system that used more energy than an electric cooker.

The green home server compares very favorably with other systems I was able to dig out numbers for.


These are all two HDD systems with a total storage between one and one and a half terabytes. A typical disk uses about 5-10W so comparing systems with significantly different storage isn’t reasonable. The Green KPC K45 build is from Home Server Hacks, power ratings for a similar KPC based system is in blog comments.

I’ve not included the Shuttle X27 (24W idle and 36W under full load) because it’s not a comparable system. Firstly, Shuttle don’t give any information as to the disks attached to the system they measured so I’d have to assume a single HDD. Adding HDDs has significant effect on power usage ****, about 8W per disk. Secondly, the Shuttle’s Intel Atom 230 1.6GHz gets a PassMark CPU score of 245 compared to the HP MediaSmart’s score of 358 with a 1.8GHz AMD Sempron 3400+. Assuming HP used the smallest processor they could to get the job done then the X27 might not be up to the job. The Atom is certainly borderline as desktop solution and the power hungry motherboard lets it down somewhat in the power consumption stakes too. An AMD Athlon 64 might be a better choice.

The AeroCool WHS uses less power because it uses energy efficient disks and power supply in addition to a low power processor and motherboard. For example, the Seagate drives in the MediaSmart idle at 8W compared to 2.8W for the Western Digital drives.

Is it possible to do better? Yes, definitely. I’m no expert I just assembled a system from the best off the shelf parts I could find. Tranquil PC, based in the UK, make a home server box based on the Atom processor but don’t give power consumption numbers for it. Some of their other systems come in at under 20W idle for a single disk system. Tranquil are worth checking out anyway as they seem to be doing some interesting things in the Green IT space.

What about performance?

How does the performance compare to other Windows Home Server systems? The raw system specs are:

  • AMD Athlon X2 BE-2400 Brisbane 2.3GHz dual core CPU
  • 2Gb RAM
  • 2 x 700Gb HDD drives

Running the PassMark suites gives the following results:


These numbers are comparable to the HP MediaSmart PassMark performance with significantly better CPU performance due to the somewhat over the top Athlon BE-2400. This was a deal at NewEgg, a slightly better choice would have been a 35W rated processor. I think the savings would have been small as it is the overall power consumption of the processor and motherboard that really counts. AMD’s Cool’n’Quiet technology really helps here.

If you have another system and want to share your power usage numbers post then them as a comment here. Remember systems with more or less disks will have significantly different power usage.

In the final post in these series I’ll go over some reconfiguration I did so that the server runs as quietly as possible. With a bit of tweaking you can end up with a near silent machine.

Next… Part 6: A Quieter Server
Index: all posts in this series

Currently listening to:
Therapy? – Troublegum

* Backing up a single machine over a wired network. Typical incremental backups take under ten minutes.

** Simulated user access by streaming music from the WHS \Music\ share. Two different files to separate instances of media player. Light use in other words.

*** I used the PassMark CPU test suit and disk test suites to simulate heavy load running them multiple times simultaneously while measuring power consumption. The peak figure is the highest power usage observed in a single minute during the tests.

**** HP MediaSmart requires another 8W per drive. See: page 154.

  1. 10 Responses to “Building a Green Windows Home Server: The 37 Watt Server”

  2. Nice article! I’m curious about the noise level with your setup. How loud does it get with the fans and hard-drives running?

    By George on Sep 19, 2008

  3. George,

    It’s hard to judge noise level and I don’t have a dB meter. Currently I have the server sitting on one corner of my desk, about two feet from me. It’s makes a low hum which is entirely fan noise. Acceptably quite but not silent. If it were sitting on the floor behind my desk it would be pretty much unnoticable.

    I actually want to go back and fiddle with the fans some more. The unit puts out so little heat that I should be able to configure the fans to run intermittently or at least very slowly which would significantly reduce the noice.

    Hope this helps.


    By Ade Miller on Sep 22, 2008

  4. I am building a similar system and I was curious where you mounted the temperature probe and how you did it since I did not see any instructions. Any help would be appreciated.

    By Scott Schaeffer on Sep 23, 2008

  5. Scott,

    I didn’t bother to mount the temperature probe. It’s lying loose in the middle of the case with the plastic tube around it. It seems to give reasonable values.

    I’m actually writing a followup post around cooling and noise levels which will go live in a couple of days.


    By Ade Miller on Sep 23, 2008

  6. George, Scott,

    Here’s a new post around noise levels. With some judicious fan configuration and BIOS settings you can get the CPU cooler running at lower speeds and the case fans to only run when things get hot. At which point the server is very quiet. While setting this up I ended up having to open the case because I thought I’d actually turned all the fans off by mistake.

    Building a Green Windows Home Server: A Quieter Server


    By Ade Miller on Sep 24, 2008

  7. Great article. One problem that I’ve had is tracking down the power requirements of individual motherboards. How did you determine that this motherboard would be power efficient? I’m looking for a similar motherboard with Giga-ethernet… however it’s difficult to determine the exact chipsets used, power usage, etc. For instance would the Biostar NF520 A2G+ be a good replacement for the MCP6P-M2 you used?

    – Jeff

    By Jeff on Nov 29, 2008

  8. Jeff,

    In part the choice of motherboard was driven by NewEgg’s combo deal at the time.

    The motherboard I used supports the EPC Green PC standard and RoHS Standard but beyond that there’s not much to say.

    It’s difficult to get power consumption numbers and hard to measure as a chipset on it’s own doesn’t mean a lot although people have tried.

    As a general rule I’d try and avoid motherboards with lots of high end integrated components like graphics chipsets as there will draw power. The BIOSTAR board I used has a pretty low end graphics card for this reason.

    Sorry I don’t have a better answer. Hope this helps.


    By Ade Miller on Dec 2, 2008

  1. 3 Trackback(s)

  2. Sep 14, 2008: Building a Green Windows Home Server | #2782 - Agile software development and best practices for building Microsoft .NET applications.
  3. Sep 14, 2008: Building a Green Windows Home Server: Conclusions | #2782 - Agile software development and best practices for building Microsoft .NET applications.
  4. Sep 14, 2008: Building a Green Windows Home Server: Choosing the Hardware | #2782 - Agile software development and best practices for building Microsoft .NET applications.

Sorry, comments for this entry are closed at this time.