Let’s say your normal miditower server that you have running 24/7 in the closet draws 150W which is not unusual for a pretty powerful computer. What if you can lower it to 50W using the Mini File Server and a Mini-ITX motherboard instead? Read on and you’ll get an example how to calculate what you save.

• What you would save is 100W (150W-50W)
• That gives 2400W per day (100W*24)
• That gives 2.4 kWH (2400/1000)
• 2.4 * <whatever the cost per kWH is>. Here in Sweden that would be around 0,1€ per KWh
• 0.24€ per day * 365 days = 87,6€
• You would save almost 90€ a year alone on the electricity bill.

And not to mention that you also save the environment – priceless!

An ATX-computer with a powerful CPU and graphics card could draw hundres of watts – even on idle. Some graphics card consume even 100W even on their own. So how does it compare to the Mini File Server running the Mini-ITX motherboards? There’s also a post on how much money you save on the electricity bill.

Here are the results when running tests while idle, but not in sleep mode:

Mini File Server: 62W
Mini-ITX VIA EPIA EN12000EG, 5×3.5″ PATA/SATA harddrives, extra PCI-controller
At high load (using HeavyLoad that reads on all harddrives and 100% CPU), the consumtion went up 8W.

Mini File Server: 30W
Mini-ITX VIA EPIA M6000, 1×3.5″ PATA harddrive

Mini File Server: 40W
Mini-ITX Commell LV-667, Intel Core2Duo 1.83 GHz T5600, 2 GB DDR677, 1×3.5″ SATA harddrive

Normal Miditower: 100W
ATX ASUS P4T-E, Pentium4 2.2 GHz, NVIDIA GeForce 6600GT, 2 GB memory, 1×3.5″ PATA harddrive

20″ Widescreen Monitor: 40W
Philips Brilliance 20″ widescreen

Network Hub: 10W
3Com Office Connect Dual Speed Hub

Rule of thumb is that a VIA Mini-ITX motherboard in a case with a fan does not draw more than 25W, less than a lightbulb. Each harddrive will add on that. For example, Western Digital’s high performance drives WD Caviar SE16 has the following specifications:

Read/Write: 8.77 Watts
Idle: 8.40 Watts
Standby: 0.97 Watts
Sleep: 0.97 Watts

But if you go for their power saving WD Caviar GP models, you get:

Read/Write: 7.50 Watts
Idle: 4.0 Watts
Standby: 0.30 Watts
Sleep: 0.30 Watts

Also make sure to read the post on how much money you save on the electricity bill.

• Possible to add PCI / PCI-Express card. Not the biggest ones though since space is very limited. But our plan is to be able to fit any usual SATA PCI-card and some moderate graphics card.
• Different colors. This time we made a prototype of four different colors and the front cover having the same color as the rest of the case
• Testing air filter. Since the server will be on 24/7, you will get dust in it so we’re trying out a dust filter. However, this might stop the airflow a little bit and make the harddrives more hot. We will test this soon.
• Better mounting for harddrives and other small changes to the actual mounting design

Here are some pictures:

We have tested the case with several different fans. All tests are available in this post. In general, all tests were performed with the following hardware but each step usually involves different prototypes:

• VIA EPIA-EN12000EG motherboard which is fanless
• Morex 80W (luckily, the actual PSU is outside the box. Only power board is inside)
• 5 harddrives (4 x 7200 RPM and 1 x 5400 RPM different brands). Unfortunately, one of the drives didn’t support temperature reading from SMART.
• Windows 2003 operating system
• SpeedFan 4.31 to measure temperatures
• HeavyLoad and HDTach and good old .bat-files to simulate load. Tip: To see your HD performance, you can use HD Tune.

Test using Scythe 120mm S-FDB

* Prototype case 1.0
* Scythe 120mm S-FDB

Here are the temperatures before we got started when the server had been idle for many hours:

HD0: 38C
HD1: 33C
HD2: 32C
HD3: 33C

Here are two charts showing the HDs and motherboard temperatures during idle which showed pretty stable:

Then we started some testing with HeavyLoad. However, none of the harddrives went up more than 2-3C so we decided to go a step further. We created numreous looping batfiles that copied 200 MB files between all the different harddrives. We ran them all simultaneous. To break the risk of just using sequentual reading, we also started some looping “dir /s” that continously made a dir /s on all harddrives. At this time, also the CPU went up 100%.

We left that running until we saw the temperatures stabilizing (around 1 hour) monitoring the temperatures and here are the maximum results we got to. Here are two examples:

HD0: 42C
HD1: 39C
HD2: 36C
HD3: 36C

So in general, the temperature went up 3-6C (4C average).

If you look at the manufactures specifications (Examples: Samsung=60C, IBM=55C, WD=55C, VIA=50C), we didn’t cross any of the recommended limits. However, we’re pretty close. Do remember that this was the most silent, and therefore not the most efficient, fan we could find. There are some very quiet fans out there that will be able to move more air that would probably lower the temperature even more. One reason why the case keeps the harddrives pretty cool is because there are about 5-10mm’s between the harddrives which lets the airflow through.

Temperature testing using Noctua NF-S12-1200

* Prototype case 4.0. This also included a dustfilter which will stop the airflow a bit
* Noctua NF-S12-1200 120mm fan

Thank you very much Noctua for sending us a sample of your fan. The NF-S12-1200 comes with a 6 years warranty and MTBF of >150.000 hours so we expect it to keep on running for a very long time in our server. Read the detailed datasheet.

We used the included rubber vibration compensators to attach the fan more easily and also for less noise.

We started out with a normal server running under low load and these were the temperatures of the harddrives:

HD0: 45C
HD1: 39C
HD2: 35C
HD3: 40C

One thing to notice regarding the fan was that we couldn’t even hear the noise from the fan, even though we changed it from 10% to 100% fan speed. All we could hear was the harddrives, the fan was extremely silent.

Then we put some load on the system using HeavyLoad and normal file copying.

HD0: 49C
HD1: 45C
HD2: 37C
HD3: 41C

So in general, the temperature went up 2-6C (3C average) – less than with the other fan. The reason why the temperature in general is higher is probably because of the dustfilter.

Here are some picture with the front cover:

Here are some pictures without the front cover:

A problem with this is that there’s no standard on mounting or form factor on these DCDC-converter boards. Morex is one of the larger suppliers, then there’s Lex, Universal and the powerful 220W model.

This is how we fitted the DCDC converterboard in front of the mini fileserver.

One alternative would be to use the picoPSU model which doesn’t take up much place at all. However, then you’re limited to 120W and that could be close to the minimum if you have the case fully fitted. However, our prototype is fitted with a motherboard + 5 harddrives and a 120W power supply can handle the load. Here are some pictures of the picoPSU:

It’s possible to change front plate to have different designs. Quite handy when testing airflow but also a good thing when considering the fact that not all people have the same taste. I prefer basic holes whereas Daniel prefer flames or a logo.

Why use such a big fan if you want to keep the size to a minimum? Well we figure that to keep the noise down at the same time being able to cool the hardworking harddrives and also the motherboard, a big fan would be the best solution. First of all we can keep the speed down, keeping the noise to a minimum and still being able pull a lot of air thru the box.

After a few minor changes to the design the first steps to get a working prototype was made. Here you see a picture of the huge fan in the back:

This was the first prototype so we didn’t fit the backplane for the connectors, but a new protoype is coming up soon!