Did you know that your PC can be immediately available while consuming less than 1/25th the power? Read on!
(If your PC already is configured to go to the proper S3 sleep, look below for some power consumption numbers for the various levels)
Coinciding with the recent, widely-linked WSJ article on power usage around the home, I happened to be anxiously awaiting my own power meter’s arrival in the mail. In my case I wanted it for a different purpose (a complete article on that topic will be published shortly. For now I just wanted to author this post — on the topic of S1-S4 power modes — so I could reference it from the other post, avoiding long parenthetical sidetracks like this), however in doing research for the other article, some rather surprizing analysis was completed that I thought worthwhile to share.
My primary development desktop PC is rather obsolete: Athlon XP 3200+ on an nforce2 motherboard, 1GB, two hard drive (totaling 200GB), network card, nvidia 6600GT 256MB AGP video card affair. From an energy usage perspective, it’s very comparable to most current PCs.
I’ve been meaning to replace it with a shiny new Core 2 Duo for some time, but the hassle of setting up all of my software again, configuring everything the way I like it, and so on, is a huge
disincentive. Add the fact that my current PC never leaves me really begging for more power, so there’s not enough drive to
On the bright side, while I’ve procrastinated about upgrading, the state of mainstream computing has continued to march forward, so what I’m upgrading improves with every passing quarter.
I use my laptop quite a bit, but the machine described above is really the workhorse, and I generally leave it on throughout the day. I leave it on partly because I sometimes need to remotely access it, but more so because I tend to have the need to jump onto it to do short tasks throughout the day. Booting up on a need basis isn’t acceptable as even the fast boots of today are still too slow, not to mention having to start all of the various tools that I use.
I used to rely upon S4 (hibernation), but even there the time to restore is too time consuming for these quick-hit usages. Add to that the occasional failure of hibernation to actual recover from a hibernation, locking up when the load is complete, forcing me to dump the state and reboot from scratch – makes me very wary of the feature as a whole, though I realize it’s more than likely an issue with a specific driver or piece of hardware that I’m using.
So instead lately I’ve been resorting to standby — auto initiated at a preset interval of non-usage, and manually triggered when I know I’m leaving the PC for a bit. I presumed this was a substantial power savings.
With delivered power meter in hand, I finally had a chance to prove it one way or other. First I measured the “base load”, when the PC is sitting on and ready for use (note that actually
doing something causes this number of spike, sometimes substantially, so I’m intentionally indicating the idle power usage). This measurement is only the mid-tower consumption, and does not include monitors, speakers, etc.
Idle Power Usage: 129 Watts.
Wow. 129W just sitting there idling. That’s about $10 of electricity a month. An ignorable cost, but when you’ve had conservation beaten into you it really makes me feel quite guilty thinking that my PC was consuming upwards of 100kWh a month largely doing nothing. What a waste.
Next I measured standby mode. In this mode the machine takes just a second or two to recover, ready for use (which I do by either sending it a Wake-On-Lan packet from remote, or by clicking
Standby (S1) Power Usage: 112 Watts.
Still 112W? Admittedly I was quite shocked that a non-functioning, non-computing PC, with hard drives powered down, could consume this much power. Sure this was a fluke; I did the same test on several of PCs and found the same marginal power savings in standby (S1) mode. This finding was entirely contrary to many sources that claim much greater efficiency of S1 mode.
All of the above was in Windows 2003 (and is the same behaviour on several XP machines I tested). I rebooted the same machine into Vista, and was shocked to find the standby mode consuming far less power (not to mention that it was audibly different — it actually shut down the fans and such).
Back in Windows 2003, I started investigating why the disparity exists, and why I’m not seeing the power savings I should.
It turns out that XP, and thus 2003, identifies the power support on first installation — determining which of the S# modes your machine supports — S1 is the weakest power mode, S3 is a
“suspend to RAM” mode (where power is cut to basically everything but the RAM modules, retaining system state with very little power), and S4 is hibernation, where the state is saved to the hard
drive. Not only will it not automatically accommodate later changes (for instance BIOS changes), but apparently it very frequently defaults to only configuring standby to S1, even where the machine
fully supports more. Furthermore, some configurations of USB devices will cause it to revert to the hoggish S1 standby mode.
My search brought me to a little Microsoft utility called dumppo.exe. With this, I could imperatively force the operating system to start using S3 sleep (given that my PC already supported suspend to RAM. Every new motherboard supports it, although some default to it being disabled in the BIOS). After running dumppo admin minsleep=S3 and rebooted, I put the machine into sleep and checked my power meter.
Sleep (S3) Power Usage: 5 Watts.
(I’ve scaled all of the power consumption values relatively, and the above says “5 Watts“)
Not only is the machine using the same power in S3 sleep as it uses when it’s completely off — yes it does use 5W powering the network card and other motherboard systems even when it’s “off” –
but the availability is impressive, with it coming completely to life instantly, and thus far reliably.
Again I use Wake-On-LAN to tell the machine to come alive remotely, making it immediately accessible from afar, and it’s far more available than hibernation. Win/Win/Win!
A little utility and suddenly this PC is using 1/20th the power than the prior standby mode. Given the average usage cycle, this will drop monthly energy consumption for this PC from 100kWh to
approximately 35kWh a month. Imagine the conservation an entire office building would achieve (consider that many IT departments unnecessarily mandate that all PCs are on 24 hours a day to allow
for automated off-hours patch deployments).
Not everyone will have this particular problem (with S3 not being recognized by the operating system), but some quick checks of PCs under my control and among friends and family has demonstrated
this to be a ridiculously common scenario, so I thought it worth an entry.