Personal computers are now ubiquitous and are something you expect to find in almost every home. With over a billion PCs in use throughout the world things like power consumption can really add up on a global basis. Windows 7 includes great power management features that extend the time you get ‘on battery’ with a laptop and reduces the overall power consumption on a desktop. Have you ever set your system to never sleep and stay on all the time though? Have you ever wondered how much that power management is saving on your electric bill?
If you are like me you’ve probably got more than one PC around the house. If you are tempted to leave any of them powered up 24x7 then read on. One of the hidden costs of a technology hobby can be buried in your power bill at the end of the month. Some systems can consume a fair amount of electricity but just because it’s not readily visible doesn’t mean it can’t be estimated or better yet, measured. I thought I’d spend a few minutes to consolidate some of what I’ve learned about minimized my energy use while still spreading plenty of technology around my home.
This is more than being ‘green’. After all, nobody wants to spend more than they have to on their power bill. Armed with a little knowledge it can be pretty easy to save enough on the monthly electric bill to cover a few song downloads. For this post I used a machine I have that I built a couple years ago. It’s been a great machine and is probably similar to systems in use by many PC enthusiasts. The actual configuration of the machine is not that important though. The power your system actually uses will vary from this one, perhaps significantly. The methods to measure and calculate though are the same regardless of the actual system used.
When we’re talking about power consumption we’re talking about watts. To calculate watts we use the power formula, P=VI where P is the power consumed, V is the voltage across the load and I is current through it.
Above: My system consuming 157 Watts (157.25W = 117V * 1.344A)
Most people won’t have the necessary test equipment laying around to measure voltage and current individually but luckily there are inexpensive devices available that will measure both and do the power calculation for you. They are inexpensive enough that they could pay for themselves in reduced energy costs if you make some changes based on your test results. In some places your electric utility might even loan you one like they do here in the City of Seattle where you can check one out from the public library. This is a fun gadget to have as it helps expose the operating costs of many electrical devices. The one I’ll be using is a P3 International Kill-A-Watt P4400.
First, here are some basics about power and billing. As I mentioned power consumption is measured in watts. Electrical utilities in the U.S. generally bill in kilowatt hours (KWH). One kilowatt hour is a load consuming one thousand watts for one hour. The same amount of KWH is consumed whether it’s a 2,000 watt load that runs for 30 minutes or a 100 watt load that runs for 10 hours. One way to think about this is like water. 2,000 gallons per hour flowing for 30 minutes will result in 1,000 gallons being consumed. 100 gallons per hour flowing for 10 hours does as well.
Rates charged vary by region and utility but I pay around $.10/KWH on average for most electricity I use. You can look on your electric bill for your actual rate. It will look something like this:
For an example we’ll assume the above measured 157W reading is constant and calculate what it costs to operate that system 1 hour by taking the following steps.
1) Make sure watts consumed is in kilowatts
(Measured Watts/1,000 = Kilowatts)
157 / 1000 = .157
2) Calculate cost based on the load, number of hours spent at that load and cost per KWH.
(Measured KWH * Hours Spent * Cost Per KWH)
.157 * 1 * .10 = $.0157
The above system would cost me a little over a penny and a half per hour to run. Though that sounds pretty inexpensive you’ll see below that it can add up fast.
Here is the configuration of the system on my test bench that I’m using for this post
The system power plan was set to High Performance and for simplicity I’m measuring only the system itself here and not any of the peripherals such as the monitor. You can apply the same techniques in this post to devices and peripherals if you wish.
The first and most relevant measurement I did was that with the PC idle. After booting and logging in I gave the system plenty of time to complete post boot activity then I checked the load and measured 102 watts.
If you were to leave this system on 24 hours a day this is the state it would likely be in most of the time. With the Kill-A-Watt device I’m using it would be difficult to measure and calculate spikes and widely varying workloads to get a precise consumption figure. The idle state represents a baseline that assumes that if the system is left powered up this is the least amount of power it would consume.
Precision measuring aside, it’s interesting to see how different loads might impact power use. To check the effect of some specific scenarios I made use of a couple handy utilities to load down the system. As you can see below I used HD Tune to make the disk extremely busy and then got a reading of 106 Watts which is only slightly higher than the load at idle. A busy disk in my system doesn’t appear to impact consumption that much.
There was however a measureable impact from placing a load on the CPU. Using a tool called MaxCPU I put a 25%, 50%, 75% and 100% load on the Q9550 and saw 125, 140, 152 and 165 watts respectively. It’s clear that in this system the CPU load plays a big role in determining instantaneous power consumption. This measurement illustrates that Windows 7 and modern CPU’s work together to provide processing power when needed but avoid wasting energy when it’s not.
Above: 165 watts measured with the CPU at 100% load
Up to this point the measurements have all been about what the system consumes when powered on. To find out the potential savings we need to look at what is consumed in the lower power states. By selecting the Sleep option off of the start menu shutdown option I saw that in S3 the system consumes only 4 watts. This is a significant difference from the 102 measured at idle!
Above: Various readings taken during my tests. Note that you can find more on the specifics of System Power States at MSDN.
Armed with a complete set of measurements we can calculate the cost of operating this system and the savings realized through the use of power management. For starters, what would it cost me to run this machine for a month if it were left on all the time? While the load may vary somewhat depending on what the PC is doing it’s safe to use the idle load of 102 watts as a minimum estimate.
102 Watts * 24 Hours * 30 Days = 73,440 Watt Hours or 73.44 Kilowatt hours.
73.44KWH * $.10 = $7.34 per month
If the system was configured so that it spent 8 hours a day on and 16 hours in sleep we can show the savings
102 Watts * 8 Hours * 30 Days = 24.48 Kilowatt hours
4 Watts * 16 Hours * 30 Days = 1.92 Kilowatt hours
For a total of 26.4 KWH * $.10 or about $2.64 per month.
Using sleep represents not quite $5/month in savings on this system. Since I’ve got more than one PC the savings here can actually be significant! Based on what I found with my own systems I made changes to capture some of those savings. I also located some rather power hungry peripherals that I decided to replace when I found that their consumption didn’t decrease when the PC was sleeping or even turned off.
If you get the chance to measure your system let me know here or join me over the Windows Experts Community and tell me about your results. I’d also be interested in hearing about other tips people have to maximize their energy savings.
thanks for post . i like it.,
Wonderful post. I am searching awesome news and idea. What I have found from your site, it is actually highly content. You have spent long time for this post. It's a very useful and interesting site. Thanks!
I like this
thanks for sharing your
[url=http://www.y8me.com]y8[/url]Awesome post, thanks
Also remember to size your power supply to your system. In the above system with a non-use power consumption of 100W on the mains using an 80+ power supply that would be about MAYBE 60 watts on the DC side.
80+ power supplies are most efficient at their 50% DC Consumption. With this 600 Watt Power Supply drawing 60 Watts on the DC side it is in the inefficent range of it's 80+ efficiency curve.
Having said that, I always look at the maximum current draw from the mains side and multiply that by 2.5 -3.0 to get the optimal power rating for my design. The reasoning for that is that the on-board multi-phasic DC-DC converter for the CPU can pull 50-75 amp hi speed pulses and if you don't have Quality and massive output capacitors as well as the power supply to deliver them (this applies to high end video cards as well) then you will start having anything from just occassional lock-ups to complete rebooting of the system.
For a list of REAL 80+ power supplies follow this link:
You will also want a high quality Active PFC supply as they can take over and under voltages (usally 90 to 260 VAC) without an interruption of the system.
I recommend Seasonic as they were the first to qualify for the 80+ certification, they are a quality PS but you pay a bit more for them. Just think of it as insurance of both your system and your data.
DO NOT buy the passive PFC supplies that have the 110/220 voltage selector switch as they are very low efficiency (50-75%) and will not stand an overvoltage event and possibly even pass it into your system. They are also the first to reboot during a power spike where the Active PFC do not.
I have 4 tower systems running 24x7 but all have at least RAID1 Hardware 3WARE controllers, have external backup drives and on my primary system, on top of all of that I have Carbonite encrypted back up to their cloud. Currently I have about 70GB out their and it "saved my bacon" once before so I will never go without it.
PS - the power supplies I purchase are seasonic 750 X series and weigh about 8lbs and I have yet to replace one.
@Roger - Thats an Extech AC Line Splitter that splits the hot and neutral into separate channels so I can clamp just one. It's really useful as you can easily current clamp things like this by just plugging them into the line splitter.
Love this! Going to try this ASAP!
What is the device that the clamp, multimeter and power cord connect to which is plugged into the power strip?
@mullerwh: The Kill-A-Watt will let you measure watts used over time. You could throw your weaker system on it for say a week, doing what you normally do with your WHS. Throw the other system on there for the week after assuming your backup schedule and other usage is the same from week to week. At the end of each week you'll have the amount of power each system consumed and you can determine from there what you want to do.
I have been playing around with the same power issue as you. I have 2 machines that I can use as Windows Home Server, one drawing 40w and the other 63w idle. Howeve the performance of the latter is twice that of the first. I'm struggling to work out which is best. If the first one works harder than the second it might not as efficient. The big questions of speed vs cost remains ?
There is also great app from Microsoft Research called Joulemeter. Which gave some interesting numbers about electricity consume of computer for people which have UPS connected to computer or battery running laptop.
Awesome post, thanks, i've been wondering for a while how much power 'Sleep' mode uses, 4 Watts is hardly anything compared to what it is at Idle.