Energy comparison

[dark_slayer]

Greetings,

I've talk myself into and out of setting up an Ubuntu server several times. The one thing I continually *want* to believe is something I can't find any good comparative data on. That is simply windows vs a linux server distro idle power consumption.

At one point using this site I installed and ran a 12.04 server from a new ssd using all my ntfs data drives, and liked the experience. I had an opportunity to sell an HTPC, and that brought me back down to 2 boxes. Both need to be htpcs, and the reasoning (for me) at the time for taking the opportunity to sell was that apartment living limits me to a server in the living room. Since it always has to be there, why not make it double duty? That's where I went and I'm at a point where I've set up W8 as a server / all in one / htpc. I have taken a kill a watt to measure the power output and it does well at idle. Here are the "server" specs

Asrock z77 Pro4-m
i5-3570k
2x4GB DDR3-1600 (Corsair Vengeance)
128GB Vertex 4 SSD
1TB WD Blue 7200
4x3TB Seagate ST3000DM001
3x4TB Hitachi 7k4000
Silencer MKIII 400W
Thermaltake Armor with 4x120mm fans, 1x230mm fan
2x Coolermaster 4in3 HDD cages each with 120mm fan

Running W8 w/ disks set to spin down after 5 minutes gives
idle - 49W

Running w/o disk spin down gives
idle - 62W

Playback w/ disk spin down in XBMC "fresh out of idle" (w/ other stuff running in background like MCM, Plex, Flexraid)
1080p Prometheus - 54W
1080i cable TV - 56W

I've been on and off the fence about never spinning down the drives (extra $10 / year but some say extends life)

In the end, if I can find an accommodating place to tuck such a device out of sight (which I have in the living room, only need a USB extension for Flirc and LAN/HDMI cables coming through the wall) is there anything further compelling me to switch over on the basis of energy consumption?

Has anyone measured their server power consumption relative to a windows install? Unfortunately I had the ubuntu server install on different hardware and couldn't find my kill-a-watt during that time.

[Ian]

Hi there and welcome to the forums

That's a nice setup you've got there and the low power consumption you've achieved is impressive. Re your question, I've never thought about this before. Why do you think the OS would make a difference to the power consumption? On idle I'd reckon they'd both be the same. After all, the hardware is just sitting there doing nothing, it's simply powered up?

Ian.

[dark_slayer]

That's a nice setup you've got there and the low power consumption you've achieved is impressive. Re your question, I've never thought about this before. Why do you think the OS would make a difference to the power consumption? On idle I'd reckon they'd both be the same. After all, the hardware is just sitting there doing nothing, it's simply powered up?

Thanks, and of course credit goes to Intel I was amazed that enabling dxva in XBMC with just the i5's hd4000 could handle a full (26GB) 1080p rip at super low utilization

I guess I thought that if it was a dedicated server running headless would save some power. Your question already makes me realize it's not an extremely fruitful endeavor It's just dawning on me that the spun-down 49W is at the wall with at least a 20% conversion loss. The 400W silencer is just 80+ bronze, which isn't really giving a guaranteed 82% eff since the idle operates under the 20% (80W) threshold.

Thanks for bringing me to sanity

It feels like the monkey wrench in the whole thing is ARM. I like using Plex, and I don't believe ARM as a server is up to the task of transcoding with plex just yet. I'm basing that opinion from owners over at synology forums http://forum.synology.com/enu/viewforum.php?f=193&sid=ebe8a336a9039358500c86506e246569

Without getting content to devices (which may be quite unimportant to some but not to me) the power savings of ARM as a server are pointless. On the flipside of the argument is bringing up the capabilities of devices I use. I believe, for example, that if I had Nexus10 level hardware for my phone, laptop, and/or tablet that I wouldn't need transcode-level horsepower in the server and could scale back to something like a Celeron 847 (or next gen) or ARM.

I still run quite a few things in ubuntu server Just within vmware player

[Ian]

test

[dark_slayer]

test

Are you saying I should test it? You are correct, but wondering if anyone else ever has before I go fiddling around

[Ian]

Oops, sorry, that post wasn't mean for you. I sent a "test post" to my forum but couldn't find out where it went. Now I know

I was thinking about your question today and I've now reached the conclusion that the energy consumption could vary between one OS and another. Windows, for example, runs housekeeping routines when a computer is idle (disk defrags etc) and so such routines could vary from one OS to another. I'm sure it'd only amount to a few watts worth of computing power but it's a difference nevertheless.

When I get the time I'll install a couple of different OS's as virtual machines and see what happens when they're powered up and idle. It'd probably not be a true test given the host may cause the guest OS's to behave differently but it'd be an interesting exercise nevetherless

Ian.