Power measurements of home computers

 Update: new measurements are lower. 10/4/07

At our house we try to be energy efficient. We've replaced most of our lights with compact fluorescent bulbs, have a Toyota Prius, and have rooftop photovoltaic panels that supply roughly 100% of our annual electricity usage. But how efficiently is the electricity being used in our high-tech gear? I took some measurements with a cheap Radio Shack clamp-on ammeter around a split extension cord to get a general idea of what different computers and home entertainment gear used. I had plenty of pre-conceptions about what used the most. But there were enough surprises that the only generalization I can make is that you really need to measure your own equipment. If I were doing it again I'd pay a few dollars more for the more accurate and convenient Watts Up meter. I'll write about PC's today. Later I'll write about TV's, and about a couple of other computer related items, my cable modem and router.

Our oldest computer (PC1999) is a Sony with 500 MHz Pentium III. The fan sounds like a lawnmower, and thinking that fan equals heat equals power dissipation, I expected this to be among the least efficient PC's. We replaced the original CRT monitor with a small LCD monitor, and its speakers are external. Next is a Sharp laptop (M2000) with 300 MHz Mobile Pentium II. Next is a Compaq (PC2002) with 1.9 GHz Athlon, which we bought largely on the basis of product specifications showing low power consumption by the processor relative to its competition, and anecdotal reports on how quiet this Compaq series was. We were right about the sound; you can hardly hear it running. I expected it to be our most efficient PC. It has a 17 inch LCD monitor with integrated speakers. Next is an HP (PC2006) with 2.2 GHz dual core Athlon 64, which we bought on the basis of power specifications and good experience with the Compaq. It is even quieter than the Compaq. It has a 19 inch LCD monitor with integrated speakers. Last is a Nokia tablet (M2007) with 330 MHz TI OMAP 2420.

 

Light Load 

First I looked at how many watts they used in normal home, light load operation.


OS Computer Monitor Speakers Total
PC1999 Linux 90 130 10 230
M2000 Linux 160 - - 160
M2000 Windows 100 - - 100
PC2002 Linux 300 120 - 420
PC2002 Windows 210 120 - 330
PC2006 Solaris 290 70 - 360
PC2006 Windows 220 70 - 290
M2007 Linux 5 - - 5

 

 

Surprise 1! Loud PC1999 is actually our lowest power PC, but its old monitor though small uses more power than the big new monitors. Surprise 2! PC2002 with the slower single core Athlon uses more power than PC2006 with the faster dual core Athlon, as I normally use them.

Surprise 3! Windows uses significantly less power than Linux or Solaris. There may be some power management settings I could activate in the various Linux distros. And I'm running an old version of Solaris Nevada, so this makes me very interested in tracking the progress of the power management project at OpenSolaris.org. But out of the box, Windows is doing a better job with power for consumer equipment. I'm no Windows bigot! I can tell you the top 20 things I hate about Windows - but energy efficiency isn't one of them.

 

Heavy Load 

Next I looked at how many watts they used under a heavy load. These weren't the same loads. PC2006 after all is many times faster than PC1999. It's just whatever load I could generate with audio/video playback, disk, and network activity, to use as much power as I could.
 


OS Computer Monitor Speakers Total
PC1999 Linux 100 130 10 240
M2000 Linux 230 - - 230
PC2002 Linux 320 120 - 440
PC2002 Windows 380 120 - 500
PC2006 Solaris 310 70 - 380
PC2006 Windows 310 70 - 380
M2007 Linux 10 - - 10

 

 

Surprise! Windows power efficiency advantage disappeared.

 

Suspend 

Next I suspended each PC. This isn't a full power off, suspend to disk, hibernate mode. It's a reduced power, suspend to memory, that you get when you press the sleep button as the system is configured out of the box. How well suspend mode works is a function of the OS and the hardware. On my systems, suspend mode isn't available on most of my Linux distros, nor on Solaris. Suspend works reliably on Windows XP, and on Linspire. Windows 98 always suspends correctly; it just sometimes never wakes up again :-( For these measurements I also show how the monitor's power consumption is reduced when it goes into its power saving suspend mode.


OS Computer Monitor Speakers Total
PC1999 Linux 70 40 10 120
M2000 Windows 10 - - 10
PC2002 Windows 20 10 - 30
PC2006 Windows 140 5 - 145

 

 

Surprise 1! PC1999 saves only 20W by suspending. Is that because of the hardware or because Linspire has a less effective suspend mode than Windows? More testing is needed to answer that question. Surprise 2! Although PC2002 saves 190W by suspending, PC2006 saves only 80W. Is that because of the hardware or because of differences between Windows XP Home and Windows XP Media? I do notice that I can't rely on PC2006 to stay suspended, and suspect that it stays awake enough to watch the network and wake up whenever it sees something “interesting.”

 

Turned Off

Next I turned all equipment off, but left it plugged in.


Computer Monitor Speakers Total
PC1999 5 10 0 15
M2000 5 - - 5
PC2002 10 10 - 20
PC2006 8 5 - 13
M2007 0 - - 0

 

Surprise! It's not much in absolute terms, but it's a lot of power to do nothing but wait for the ON button.

 

Configurations

Here are some details of my configurations.


OS Computer Monitor Speakers
PC1999 Primary: Linspire; Secondary Windows 98SE PC, Sony Vaio, Pentium III, 500MHz NEC, 14” LCD, 1024x768 Sony
M2000 Primary: Sun JDS3 Linux; Secondary: Windows 98SE Laptop, Sharp Actius, Mobile Pentium II, 300MHz integrated integrated
PC2002 Primary: Sun JDS2 Linux; Secondary Windows XP Home PC, Compaq S6000Z, Athlon 2600XP, 1.9GHz Samsung 170MP, 17” LCD, 1280x1024 integrated
PC2006 Primary: Solaris Nevada; Secondary: Windows XP Media PC, HP a1540n, Athlon 64 x2 4200 dual core, 2.2GHz Sharp LL193A, 19” LCD, 1280x1024 integrated
M2007 Debian Linux Tablet, Nokia N800, TI OMAP 2420, 330MHz integrated integrated

 

 

Conclusions

The main conclusion is that you probably can't guess what power your equipment is using without measuring it yourself.

These are mainstream consumer PC's, not top end gaming machines, and they use a significant amount of power when compared with home entertainment equipment or as a fraction of total home energy. It's worth worrying about, but there aren't any good sources of information to help you make a good selection when you buy. The newer equipment tends to use less power than the old equipment, so that's another reason to talk yourself into springing for some new toys.

By a huge margin the most frugal device is the Nokia tablet computer. Many times at home I want to just quickly look up something like a recipe, who directed a movie, how the Mars rovers are doing, or to check my email. I used to always do that with PC2006 which boots Solaris very quickly (Windows is much slower). Now instead I usually just pick up the Nokia tablet which uses almost no power when asleep and wakes up instantly. Then I'm using 5W instead of 360W, and sitting comfortably in the family room or in the garden.

When you're talking about corporate servers among the first thing you want to do to conserve energy is use virtualization to consolidate workloads and use your servers more efficiently. What do you do for home computer use when you want a big screen? Workload consolidation is not an option for PC's at home where you need one monitor per pair of eyes. Or is it?

What if I kept my monitor and got rid of the computer? A Sun Ray thin client uses 4-7W. Add my monitor and it would be about 75W - quite a savings from 360W. Running a thin client over home broadband is more challenging but some of my colleagues are using Sun Ray at home, in a Sun administered program. Zonbu is renting a Linux PC for $99 plus $12.50 to $19.95 per month rental. It saves cost and power just like my Nokia tablet by using flash memory instead of a hard drive, so it uses just 15W of power.

What if Time Warner or AT&T offered to rent a Sun Ray the same way they rent cable boxes? They'd probably be cheaper and use less power than the Zonbu. How about it?


Comments:

Very interesting. I'm finding out quickly that you can't guess power use as I'm sizing a UPS for about 10 machines, a combination of servers, desktops, NAS and SAN, ranging from 6 months to 7 years old. Thanks for the info

Posted by Al on August 06, 2007 at 07:58 AM PDT #

I noticed and fixed an error in my table of configurations. The laptop runs Sun JDS3 Linux and the old deskside PC runs Linspire. My boy loves the games included in Linspire and won't use Windows.

Posted by Walter Bays on August 06, 2007 at 11:00 AM PDT #

Post a Comment:
Comments are closed for this entry.
About

I am a software engineer in San Diego, president of the Standard Performance Evaluation Corporation (spec.org), formerly a mathematician and a violist.

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today