We love our computers for all the ways they make our lives (and the world) better -- the wealth of knowledge (and democratizing force) of the Internet, the instantaneous communication, the sophisticated tools that help us work and create and share. But do our computers love us back?
The modern world's greatest tool is among our most disposable and resource-heavy. Performance-wise, computer design has progressed staggeringly well and astonishingly fast. But looking at it from a green perspective, the work has barely begun. When our laptop dies and we toss it, it either rots in a landfill, or children in the developing world end up wrestling its components apart by hand, melting toxic bits to recover traces of heavy metals. Did someone forget to design for them? And of the $250 billion per year spent on powering computers worldwide, only about 15% of that power was spent computing--the rest was wasted idling. Did we really get what we paid for?
We've written about various aspects of green computers before, but here's an attempt to give a whole-picture view of what the bright green computer of tomorrow will be like: efficiency, manufacturing & materials, recyclability, service model, self-powering, and other trends.
Power: Watts vs. Flops
It has been said that it takes a pound of coal to create, package, store, and move every 2 megabytes of data. This is wrong--Lawrence Berkeley Labs checked the numbers and found it was more like one pound of coal for every 10 or 20 megabytes--but even so, that's a lot of energy. Even with energy prices as cheap as they are now, Google projects it will soon cost more to power a computer for four years than it does to buy a new one. Faster processors historically use more power, and you may recall some generations of Intel chips being called "toasters" for the enormous heat they generated. But that heat's not necessary, it's inefficiency. Inefficient CPU's are a double hit because they both use too much power themselves and their waste heat increases air conditioning needs, especially in server farms--between the computers and the HVAC, data centers use almost ten times the energy per square foot as office space. The waste heat also causes reliability problems, as CPU's crash much more often at higher temperatures. Besides CPU's, inefficiencies also crop up in power supplies, monitors (though LCD's are an order of magnitude better than the CRT's we used to use), and some peripherals. Many people have been working for years to slice this inefficiency out of machines, particularly laptops and other mobile devices whose battery life depends on wise power use.
Transmeta was the pioneer in power-efficient CPU's, with a brilliant design that provided 30% - 170% improvement in performance-per-watt from Intel chips. The market didn't really appreciate Transmeta, but they're still plugging away, and will hopefully be more successful over time. Even more impressive, P. A. Semi is a new chipmaker who claims their CPUs have ten times the performance-per-watt of other computers. The biggest player involved is Sun Microsystems, which has bet a lot on efficient CPUs. They claim their "Niagra" processor gives the best performance-per-watt of any CPU on the market, and said in a 2005 press release that if just half of their entry-level servers from the last three years had been Niagras, the impact would be equivalent to pulling a million SUV's off the road. Sun's CEO Scott McNealy said about their Eco-Responsibility Initiative, were here to make money. Understand that. But we think we can do that in a way that pays off not only for our shareholders but for everybody on the planet.
CPU's are not the only sources of inefficiency in a computer. Power supplies are notoriously bad, generally as little as 47% efficient. And since everything in your computer runs off the power supply, nothing can be efficient without a good one. As we recently mentioned, "80 Plus" certified power supplies are helping fix this by running at 80% efficiency or better. Software also helps. Many computers are only in use 10-20% of the time; laptops use power-management software to sleep or hibernate when not in use, and although desktops lag behind, almost all now have some kind of power management software.
On the far horizon, reversible computing (which also includes quantum computing) promises to reduce power consumption by a factor of several thousand, but such systems are still very much in the lab. It will probably be a decade or two before anything like that makes it to the marketplace.
Computers' power needs will decrease over time one way or another, and at the same time, our ability to generate power renewably on-site will increase. I think within the next ten years we will hit the crossover point where computers can generate all their own power. Solar panels can already cost as little as a dollar a watt, and can generate 5 - 10 watts per square foot. Laptops currently use between 15 and 45 watts. If you assume photovoltaic efficiency increases 50% and computer efficiency doubles (modest goals for 10 years), every laptop could power itself with a PV panel on its top. Alternatively, an average person can generate 75 watts continuously pedaling a bike, or 5 watts turning a hand-crank; computer efficiency gains of 10x could make this a viable method. This is theoretically already designed into the One Laptop Per Child project, and has been tried elsewhere as well.
Sugar and Spice and All Things Nice--well, actually not.
Electronics make up 70 percent of all hazardous waste. Manufacturing computers produces about two pounds of waste for each pound of computer, and between a quarter and a third of this is hazardous waste (the only numbers I have are from ten years ago; can someone provide a link to recent studies?) This includes things like lead in solder, Selenium in circuit boards and power supplies, mercury in LCD backlights, polybrominated flame retardants and antimony trioxide flame retardant in plastics, cadmium in circuit boards and semiconductors, and more. In addition, chip manufacturing uses some of the deadliest gases known to man (such as arsine, silane, chlorine, and phosgene) and several greenhouse gases (like nitrous oxide and carbon dioxide). Not to mention the few thousand gallons of water and over 6000 megajoules of energy a computer requires to fabricate. This is more than the average American's energy use for a month and a half, and Americans use a lot of energy. (Again, if someone has recent numbers on water & energy, please send.) But the computer of the future will be cleaner and greener.
NECs 2002 PowerMate Eco was the first computer to use lead-free solder, a fully recyclable plastic case, monitors without harmful gases, and no toxic flame retardants. (It also used a very efficient CPU, the Transmeta Crusoe.) Little has happened since then, but Dell's OptiPlex GX520 and GX620 desktops are nearly lead-free, and by this summer all computers entering the EU are required to be also. A smaller Oregon company, Computer Technology Link (CTL), sells lead-free PC's that also reduce power use by 15-25%, CTL will even recycle your old computer for free, as well as the computer you buy from them when you're done with it. In 2004 HP prototyped a printer with a corn-based plastic case that could biodegrade. The jurys still out about whether biodegradable cases will actually go to the compost pile rather than the landfill, but it's worth a shot, and last year NEC created a mobile phone with a compostable case, made from a composite of corn plastic and natural fibers.
Circuit boards could be created with a fraction of the waste produced now; they no longer have to be acid-etched copper on fiberglass, we have new technologies such as inkjet-printed conductive polymers which cause only a tiny fraction of the waste. The technology is still in the lab, but on a production scale it should be both greener and cheaper than etching. Even with existing technology there are many toxins which can be eliminated from circuitboard production, and the European Union's pending RoHS toxics-reduction legislation will begin to force companies to build better boards.
Chip fabrication could be radically cleaned up. It's a hard problem--manufacturers push the edges of materials science to get all those tiny features on the chips, so processes are not well-enough understood to easily swap toxic reagents for nontoxic ones--but it is a solvable problem. The lack of progress is mostly due to lack of priority in R&D--consumers don't care about production impacts, they care about speed. Some fabs are working on making their buildings and support systems better, though. In 2004 Texas Instruments worked with Rocky Mountain Institute to design a greener fab in Texas that saved so much money in water, power, and construction costs that it was cheaper than building overseas. That success simply involved better facility design; imagine if we redesigned the whole process.
In comparison to reformulating the computer's chips, reformulating the outside shell is a snap. The shining, coated plastics many computers have now are cobbled from a stew of materials that are not only unrecyclable, but also somewhat expensive. As mentioned above, NEC and HP have made forays into greener case design using exotic bio-plastics. Even simpler methods work as well--sleek aluminum cases are more robust and far more recyclable than plastic ones.
Giving Your Computer Nine Lives
For years, computer manufacturers have blithely packed hazardous substances into their gadgets, then let the whole pile of materials speedily obsolesce; but legislation and consumer activism is changing that. Their new laws requiring manufacturers to take back their own products when customers discard them is a simple but powerful force for changing the production process. Faced with disassembling parts and cycling them back into the fabrication process, companies are making more careful decisions about how those parts are assembled in the first place. Faced with disposal of their own toxic waste, companies are more motivated to eliminate the ingredients from production.
Companies in Europe who've been legally obliged to take back their productsthough they initially balked at the ideahave found that it actually helps their bottom lines, increasing efficiency in labor and materials, and thus saving money. A Panasonic television from 1984 had 39 plastic parts made from 13 different plastics, and took 140 seconds to disassemble. Their 2000-model television contained a mere eight plastic parts made from two types of plastic, and took 78 seconds to take apart (with a corresponding advantage in the time to assemble). Panasonic has a great illustration of this on their website.
Dell recovered 36 million kilos of their products last year for reuse or recycling, and they plan to triple their product recovery by 2009. Their takebacks were not brought on by legislation, but by consumer activism--the Silicon Valley Toxics Coalition's "Toxic Dude" campaign, which had college students pressuring Dell to take back old computers, worked so well that Dell now gets it, and has exceeded its own recycling goals. Hewlett-Packard is another leader in recycling old machines--both their own and other companies'. They reuse what they can, then downcycle and dispose of the rest. It's far from being a perfect cradle-to-cradle system, but when all companies have to take back their products, much more design effort will go into creating self-perpetuating cycles of components and materials.
The best way to recycle a computer, however, is to keep it and upgrade it. After all, what do you really need a new computer for? A faster CPU, more memory, and more disk space. All these can be bought individually and replaced, in a well-designed machine. Most brand-name manufacturers are laggards here, making little effort to allow upgradeability of the CPU, though RAM and drives are easy to add. The vanguards of reusing computers are small service shops and resellers, making what are called "white box" systems cobbled together out of generic components, or souped-up older PCs. A few--like Berkeley, Californias Immaculate Computers--even make an effort to make their white boxes green by using lead-free solder. White-box systems are by their nature easy to upgrade or reconfigure, which also means they are easy to disassemble and recycle when you can't upgrade them anymore.
Service: Bits, Not Atoms
Have you ever watched the CPU usage of your machine? If so, you've noticed that it's twiddling its thumbs at just a few percent almost all the time, and only cranks up to full usage for a couple minutes here and there when you're using some fancy Photoshop filter or playing a game. Why have that idle horsepower at all? Because we want it to be there when we need it. Some efficient CPU's are starting to use power-management within them to shut down unused areas without the user noticing. But the ultimate way to have even more horsepower, and only pay for the time you're putting it to use, is to use someone else's machine. A "thin-client" or "dumb terminal" system is a service that connects you to a big industrial-strength computer somewhere else. Basically all you need is a screen, a keyboard, and a fast net connection; the computers brains and disk storage are kept on the server. Sun Microsystems is trying to push the world towards systems like this; they say existing thin-client systems use a tenth of the power and reduce raw material usage by a factor of 150. And that's even using new machines. Server farms can be just as powerful using a large number of older, slower processors as they can be using a small number of fast new processors; while each individual CPU might be too slow for you to want in your own machine, a rack of a hundred reclaimed from the trash bin could be faster than the best machine you could afford. This will help green computing by allowing old hardware to be useful for much longer--easily triple, possibly five times as long as it's currently useful. Sun points out that their thin-client system also allow people to work from home, which reduces the need to build offices and commute; these gains could have as much or more environmental impact as the improved computers. There's financial incentive, too: in addition to the green benefits, Sun says using thin-client systems in their own company saves them $2.8 million per year in energy, $65 million per year in real estate, and $21.2 million in other costs.
Sun's model involves centralized server farms running proprietary hardware and software, but theres no reason it couldn't be done running open-source software on millions of generic machines scattered in living rooms across the world. With strong encryption, redundancy, and reliable connections, it doesn't matter if your data is located in your living room or scattered in stranger's homes in Bangalore and Berlin. The Seti@Home project has already proven the potential for large numbers of ordinary (even obsolete) net-connected computers in people's living rooms, using only spare CPU cycles, to become a powerful supercomputer.
The Bright Green Computer
All this is just the beginning. So far, consumers haven't cared about ecological impact when buying computers, they've cared only about speed and price. But as Moore's Law marches on and computers commoditize, consumers will become pickier about being green. (This means you!) Tomorrow's bright green computers will be different from today's in many ways. So far its rare to see a product that pulls on more than one solution, but the industry is starting to move forward--sometimes of its own accord, sometimes dragged kicking and screaming by regulations and consumer activism. Devices use less and less power while renewable energy gets more and more portable and effective. New green materials are developed every year, and many toxic ones are already being replaced by them. The greenest computer will not miraculously fall from the sky one day, itll be the product of years of improvements to this and that along the way. Designers can now make many small changes converge into a full rethinking of the industry, into directions as yet unknown. And when they do, we will have on our desk a machine that not only connects us to the world through information technology, but to a cycle of manufacturing that doesn't hurt us or future generations; to companies that consider both the people who make their computers and those who take them apart; and all in a product that is lighter, sleeker and more elegant than any we've yet seen.
Are there places that we can take our monitors and other electronic devices if they are broken, and GoodWill wouldn't even want them? Rather than throw them in the trash? How would I find out if my city does something with them?
Earth911.org has an easy to navigate database of recycling programs in cities all across the U.S. You can find out if your city/state has a program there.
A great post; I'd just point out that another ecological aspect to be considered here is the amount of energy that goes into making a computer--it's very high, much higher than some thing. I blogged it, if anyone's interested.
Err... the energy cost is higher than some think. Not just higher than something. My bad.
Re: "Manufacturing computers produces about two pounds of waste for each pound of computer, and between a quarter and a third of this is hazardous waste"
I work for a computer manufacturer and I understand every part of the manufacturing process. Your claim is just plain wrong! Electronics manufacturing produces very little waste and current laws already require that the small amount of hazardous waste that is a by-product of the process be recycled. Where did you get this bogus information?
Hey, TIm, does your assessment count the waste created right from the moment the first capacitor/resistor/circuit board is manufactured?
Or just waste at the component assembly stage? I think the article is counting the former, which the real waste of course.
Tim: as I said in the article, the numbers I have are ten years old, and I'd love to find a recent credible source. The government of Canada did a study in 2000, but that's still pretty old, and I couldn't get the full text, just the executive summary (which just did national numbers, not itemized-by-computer numbers.)
The numbers I used are from Northwest Environment Watch's book "Stuff: The Secret Lives of Everyday Things", and I haven't found a credible source more recent than that which goes into adequate detail. A good summary quote from the book is this:
"In all, the factories making my 55-pound computer generated 139 pounds of waste and used 7,300 gallons of water and 2,300 kilowatt-hours of energy (about one-fourth the energy the computer would use over its four-year lifetime). State-of-the-art factories could have made the same computer with half to two-thirds less waste. And different computers--with flat-panel displays (like those in laptop computers) instead of todays' big vacuum tube monitors, for example--could have been made with even less waste."
So that's why I roughly said 2x waste-to-computer ratio, I took the optimistic side of their numbers.
But be aware that most of the waste is from the chip fabrication, not from the case or whatnot, so if you work for a computer manufacturer as opposed to a chip fab, you won't see the bulk of the waste. I know firsthand how amazingly resource-intensive chip fabrication is, because I used to teach people how to run & repair high-density plasma dielectric etch tools (as well as other wafer fab equipment) for Applied Materials, the leading semiconductor equipment manufacturer.
Again, if you know a reliable published life cycle analysis for an average computer today, I'd love to see it.
Noam: thanks for the link to the paper! I copped out and said "a couple thousand kilowatt-hours of electricity" because of the outdated-numbers problem mentioned above. (I knew energy use would change much more unpredictably than waste, and didn't even want to hazard a guess.) ...However, the Williams paper gets its numbers from a computer with a CRT, not an LCD monitor, so it overestimates the energy use. (Few people use CRT's anymore.) Do you know a source that uses better assumptions? In the meantime, I've updated the article to include your link and say "over 6000 megajoules of energy".
Could multiple cores help on the efficiency side? Think 10 100MHz tiny-processors rather than 1GHz chip..wake them up as required. Like a shark's brain.
I found it interesting that Jeremy Faludi spent so much time quoting Sun Microsystems even while indicating that they require the use of proprietary hardware and software. He said that there is no reason why it couldn't be done using open source software. Well, it has already been done with open source software. LTSP.org has been a stand-out open source project since it was first introduced in 1999. Our company, Symbio Technologies, has been both developing and bringing server-centric, totally stateless technology to market with absolutely no vendor lock-in. A simple search of "diskless thin client" could have yielded the author a wealth of information about existing open source alternatives.
Europe passed a framework directive last summer, the Energy-using Products directive (2005/32/EC). The purpose is to quantify precisely the use of energy in both the manufacture and use of high-volume energy-using products, including PCs, then drive improvements either through regulation or voluntary initiatives. Some more recent data than you have might be available in the research reports at http://europa.eu.int/comm/enterprise/eco_design/ .
They indicate total energy use of 630PJ, of which 75% is during use as an energy metric. Other metrics are available including some waste, but not all.
Thanks for an informative article. You've convinced me to upgrade my computer rather than replace it. It's been running for 5 years now and there's no real reason why with a bit more hard drive space it shouldn't keep serving me well for a few more years.
Computers and whats in em has completely changed 3 times or so since 93-94 when they likely were doing that study.
First of all comopuuters used to use steel frames and hdds used to be heavy steel bricks.
Secondly everything has changed in every way imaginable. Bits that used to have several hundred parts now only have 5 others no longer even exist.
The crt is completely uetterly changes since 95. Back then they were very heavy think and jammed with stuff.. now they are mostly empty space hugging a think walled light tube. There is a reason they got soo cheap they use alot less stuff to make em now.
A United Nations University study claimed that a CPU took did more ecological damage to create than a whole car. Rare earths like tantallum which are used in antenna for wireless devices are even worse. This article didn't pay enough attention to the damage at extraction point.
Not likely. I know a fair amount about chip fabbing and it doesnt use all that much stuff to make chips these days.
As for wireless... the amount of rare earth metals used in a wireless rig is rather tiny if at all. And gathering rare earth metals is much easyer now that they have improves methods of extraction. Thats why rare earth magnets are so cheap now.
Alex, multiple cores help with certain kinds of problems but not others. In addition you need software written to take advantage of them. As 2 and 4-core chips become common at the high end (high power, high speed CPUs), I expect that software developers will update their software to take advantage of multiple cores, and that should help at the low end (low power, low speed) too. However not all problems can be expressed in parallel so there are some limits to this.
One other way to green your computer is to get wind power certificates specifically for your desktop or laptop: a Canadian NGO has a 20,000 wind powered computers campaign