Zerofootprint: Black is the new green: Sources for computer power consumption
by Ron Dembo, Zerofootprint on 12.15.06
We’ve got a heap of responses to “Black is the New Green,” and they tend to fall into three themes. The first is to commend the idea of making computers more energy efficient through power-saving software. The second is to suggest that much of what we suggest is already possible for anyone who takes the time to configure their machine; in other words, the goal should be education, rather than software development. The third is to express incredulity at our numbers.
There’s not much point responding to the first set of responses, except to say thanks. We appreciate it.
As for the second, we agree: education can go a long way. But not far enough. The fact is that we need to be educated about a lot of things right now: about making our cities sustainable, about the environmental consequences of our driving, of our travel, of our consumer choices, of the food we eat, of our landscaping decisions, and so on. The power we waste on our computers is a small piece of the puzzle. That’s not to say that it’s not important. But we have to acknowledge that a lot of educators are vying for our attention, and it could be a while before people get around to changing the settings on their computers. Why not create a tool to make it easier for them?
A software solution has something else to recommend it: whatever we develop will have the ability to quantify the electricity savings, and aggregate those numbers centrally. We will be able to keep a running count of the “nega-watts” - that is, the power that doesn’t get used. This is a powerful tool.
For one thing, just seeing the immense savings (in kilowatt-hours, in dollars, in CO2 emissions) will be a powerful educational tool. What gets measured, gets managed. In other words, the more you save, the more incentive you have to save more.
For another, if we can quantify rigorously enough, we’ll be able to issue and sell carbon credits to create revenue to plough back into the project. In other words, the more power we save, the more we’ll be able to save.
Neither of these tools is available if we just show people how to configure their machines.
Finally, a response to those who have concerns about our numbers. It’s true that much of the math is based on speculation. It is meant first of all to show the sheer scale of the problem. We can’t know how many of the world’s 650 million computers are turned on but idling, or what their screensaver settings are, or even exactly how much power all the world’s graphics cards draw. There are too many variables to get an exact number. But we do know the number is huge.
Still, though we did have to use some proxy values, the math is based on real numbers, which we have in some cases discounted to ensure that we’re not overestimating.
First, here is a link to a simple calculator that will give a sense of the power requirements of a system of just about any combination of hardware, not including monitor. It is very easy to get to 200W, which is higher than the number we used.
The graphics card can make a huge difference, in some cases jacking the power usage up over 450W. Again, this is three times the number we used in our calculations.
The monitor also makes a difference, of course. Switching from CRT to LCD offers a savings of about 60%. That’s a drop from 100W to 40W. Imagine what shutting it off altogether would save, and giving the CPU and graphics card a break when you’re not using the machine.
We’re talking about a lot of energy going to waste. That means squandered money for those paying the bills, and it means unnecessary greenhouse gas emissions from the plants that generate the power. The CO2 emissions will vary depending on the source of the power, of course. Computers wasting electricity from a nuclear plant won’t warm the globe as much as machines wasting electricity from a coal-fired plantbut it still doesn’t make sense to build expensive, risky nuclear plants to keep computers idling. And let’s take into account that some computer-intensive companies draw more power than the local utility can provide, and are forced to install their own diesel generators, which raise the environmental cost quickly.
With more computers drawing more electricity (to say nothing of all the video games out there—a PS3 uses more than an order of magnitude more power than the first Playstation), the situation is only going to get worse. We can begin the process of turning all that around, we can do it in such a way that the project sustains itself, and in such a way that it helps educate people about energy and its environmental costs.
Let’s not delay.
[Ron Dembo, Zerofootprint]
This issue has been discussed a lot on ecoIron as well. Power consumption is a big expense, accounting for maybe 10-20 percent of the typical IT budget, and this figure is expected to rise to 40-50 percent in ten years time. Companies can take the first step by holding their IT department accountable for the power they use; currently, only about 5 percent of IT departments pay for their own juice.
Secondly, about 80 percent of the total energy going into a desktop computer is used in its manufacture. This, followed by the fact that even high end users will only use 1-2 percent of their processing power over the lifetime of the machine suggests that the issue is more of a resource allocation problem than a power problem. This can be solved by introducing thin clients, virtualization, and extending the life of the machines - studies suggests a six year life cycle is not unreasonable.
Finally, with 80 percent of a typical IT budget going into maintenance of the equipment, the sad fact is that most mid level IT managers concentrate on perpetuating the wasteful cycle to keep their staff and their jobs. As evidenced by firms in Asia and Africa, the reality is that 80 percent maintenance figure can now be brought into to 50 percent range by introducing the appropriate technologies. Ultimately, the major problem may very well be a lack of leadership in your IT department that is preventing your organization from going green.
I'm sorry, but you're wrong.
The site you reference for total power requirements for a computer doesn't calculate what your computer's power consumption is - it calculates how large a power supply you would need to power all of the computer's components in the *worst case*, i.e. every component being used to the maximum extent simultaneously. A computer sitting idly doesn't require that maximum amount of power, and if you are using it in a way that requires all that power, then it's not sitting idly since you're using it. The same is true for graphics cards - the tests you cite employ software that exercises the system and the graphics card to the greatest degree possible. A computer sitting idly won't use anywhere near that much power, and if it is drawing that much power, it's because it's being used for something, and that's not wasting the power.
Consider my main desktop system. The calculator you reference dates from 2001, so it's not adequate for accurately determining my maximum power requirements. Checking off the closest equivalents to what I have, it recommends a power supply capable of producing 160W, and I'm guessing that if it were up to date it would specify that I need a 350W power supply. The *actual* steady-state power consumption of my system at idle is 80W; that's not a theoretical or computed value, that's a value measured directly with a power meter! Computer power supplies don't use their maximum rated power all the time - they throttle up and down depending on what the system's current power requirements are. Because most of the software I run is standard stuff like word processing, simple graphics and the like, that power consumption stays pretty much the same even during operation - it may spike momentarily if I do an operation that requires some extra processing power, but then it drops back down again.
In short, you're taking the maximum power requirements for a system operating at full capacity, and you're taking that as the power consumption of an idling computer. That's just not valid.
You also write:
"A software solution has something else to recommend it: whatever we develop will have the ability to quantify the electricity savings, and aggregate those numbers centrally. We will be able to keep a running count of the “nega-watts” - that is, the power that doesn’t get used. This is a powerful tool.
For one thing, just seeing the immense savings (in kilowatt-hours, in dollars, in CO2 emissions) will be a powerful educational tool. What gets measured, gets managed. In other words, the more you save, the more incentive you have to save more."
Did you take a look at the Local Cooling application? It pretty much does exactly what you're asking for. It configures your system to reduce power consumption, aggregates numbers centrally, and shows your individual savings.
I am extremely cognizant of the need to keep my computer power consumption as low as possible. My main desktop system is never idling - it's only on when I'm using it, and it and the monitor are powered off completely when not in use. My principal working computer is a notebook that uses less power, and which goes into standby when it's not being used. I have two other computers that are also only turned on when absolutely needed, and turned off when no longer needed. All of my power bricks ("energy vampires") are plugged into a power strip that in turn is plugged into a 24-hour timer that shuts them all off automatically at night when they're not going to be used. Cutting unnecessary computer use can and will save significant amounts of power, but not nearly as much as you suggest.
Here's another system power consumption measurement, this one done on a more modern system than mine with a "high-power" graphics card.
http://techreport.com/onearticle.x/10945
Runnings Windows Vista with the full 3D interface enabled, the system consumes 107 watts at idle. Nowhere near the 450 watt figure quote above, because the system doesn't need that capability when it's idling.
(Hat tip to the EcoIron blog for the link)
Trehugger covered Local Cooling on December 1st. I agree with Leszek that you seem to want to reinvent the wheel here.
http://www.treehugger.com/files/2006/12/local_cooling_t.php
I'm a desktop and web app programmer and love the idea, I just don't think you need 2 of the same thing.