What?

Future computers will be fairly powerful, but they will excell in their ability to ab-so-lute-ly nothing if it isnot needed. When current computers are switched on they continually draw current, making it necessary to use funny constructs such as standby. All this improves with technology that is currently under development.

We would like to have computers in just about anything we use, including even gadgets like photo frames. Such a frame is a good example of a gadget that need not draw any current as long as it displays a stable picture. Since the current generation of photo frames does not work that way, better wait for proper ones to appear. Actually, why would you want to buy one anyway, if your TV and computer screens can also display a photo of your children playing when the devices are switched off?

Future computers are likely to not have an on/off button anymore. Instead, they will just be there, like a chest, or perhaps a better example, like a phone. Why would you want to switch off a device if it uses nothing? A real standby mode, which is automatically triggered when a device has nothing left to do, works much more intuitively -- and it conserves more energy.

The new devices probably cannot be rebooted, so they must be crash-proof. In addition, not every photo frame should come with software license costs. This makes it very likely that open source software such as Linux will be used for those devices.

The following list mentions the technology to watch with future purchases. Much of it is expensive at the moment, but even now it is possible that you will earn it back by saving energy during its lifetime.

• bistable screen (elektronic paper)
• bistable memory
• solid state disc
• open source operating system
• working without clock interrupts
• asynchronous elektronics
• bluetooth instead of WiFi...?

Details of all this can be found under How?.

Why?

Although most of us have accepted the need to reduce their use of energy, the overall consumption is stil rising. The main light points in our homes are still not replaced with TL or LED lighting, and there is a constant demand for mobile devices. Countries still have to increase their production of energy, instead of lowering it.

When fossile fuels are burnt to produce electricity, we decompose the carbon and hydrogen in it to form water (as steam) and carbon dioxyde, a.k.a. CO₂. The latter causes global warming, and its direct cause is that we pump up fossile fuels and emit them into the air after it has spent some time on the surface of the Earth.

If devices are used a lot for a long time, it is often clever to employ energy-saving technology. Replacing a hard disc with a Solid State Disc (a sort of flash-drive) may cost a few hundred Euro more, but it saves about the half in current --about 15 Watt-- saving you about EUR 20 a year. Since these alternative drives usually last for decennia, it is possible to earn it back. Furthermore, the lack of moving parts in an SSD means that it won't crash as a result of physical damage; this faith is sealed for every plain hard disc.

We have an everlasting hunger for more, more, more. Many of us believe brochures that claim that 160 GB is "not done" as hard disc in a normal PC. This is blatant nonsense, as every disc will sooner or later fill up if unless your remove old rubbish once in a while. Using more electricity just to store more rubbish that you would not even recognise if you were to bounce into it back again later seems like nonsense, but it is actually very common, both in homes and in companies. Contrast this with the fact that Linux runs smoothly on 10 GB and a well-managed Windows system could work on 25 GB. Add a few GB for digital music and a few GB for any movies you may have backed up on your disc temporarily, and you're set.

The wish for ever-more is mainly of interest for manufacturers, who like nothing more than introducing new technology and sell it to all of us. Much of this is unnecessary for normal applications at home or in the office. Most brute computing power in homes is used for games, and in offices it may be put to good use for busy servers like spam filters or database application servers. But even in offices the choices are not always made in a reasonable manner: the habit of choosing a powerful webserver is often nonsense, as the web protocol is fairly lightweight, and chances are the true bottle neck is the uplink to the Internet, not the computational power of the web server.

How?

We now turn to technology that put our hopes up, because it saves energy and at the same time is full of commercial opportunities. The energy needed may be so little that the simplest devices can be fed by wrist movements or similar "free" sources of energy!

This explanation is a bit technical, but anyone who understands how computers work in general should be able to understand it.

Old and low-power? We sometimes hear of people who setup an old computer as a server. The idea that old computers are slow often make people think that this will also conserve energy. This is not true, because the power unit may be dimensioned a bit smaller, but the average energy used is fairly high, especially when viewed as the energy per computing step. Older technology generally uses more energy to do the same thing than a newer computer. It is better to use a new computer that is specially designed to conserve energy for your server or desktop PC. A laptop with a broken screen backlight could be an excellent home server, to be managed through remote login or its external VGA port.

Bistability. Many parts in a current computer use energy to remain stable, and the most interesting technological changes are those that avoid this minimum load. The term "bistable" means that something is stable in two different states, so that energy is only needed to move from one state into the other. Think of a ball that lies either to the left or right of a hill, but that will not roll to the other side without external force:

If you call one side "1" and the other side "0" you can store digital information in this system without a need to burn energy to retain it. Many interesting developments use this generic principle.

Elektronic paper. The first few devices with bistable screens have appeared on the market, even with colour displays. Such displays do not burn energy to show an image or document, and could therefore replace paper as a zero-energy display. For this reason, they are sometimes called electronic paper.

They are less suitable for games with rolling screens, but they do lend themselves well for many other popular applications, such as text processing and spreadsheets. For applications, think of the note books that are used a lot in Star Trek. It is to be expected that modern laptops will come with similar screens, to make the battery-stored energy last longer.

Digital movies often follow the MPEG format (like on a DVD, but also in DVB for digital television broadcast and satellite TV). This format is constructed so that a minimum amount need chnage on the screen, so even a digital television showing such moving images could conserve some power if it had a bistable screen. To make all this possible, it is required to write the data to the screen in a different form than the current TV-ish norm that rewrites the screen at least fifty times a second. A nice add-on of bistable screens is that the screen won't flicker anymore, making them as relaxing to view as a 100Hz-television.

Main memory. The RAM in a computer acts as its main memory, storing all sorts of intermediate results. This includes the documents that you are working on. You may know RAM in the shape of DIMM's or DDR-memory. All this memory is based on a component called a "condensor", which can be charged with a voltage. The disadvantage of using this component is that it is not bistable: the memory leaks and will eventually forget what was stored in it. This is why its contents are refreshed a few thousand times a second, by reading it and then recharging it. This means that a computer with more memory burns more energy in keeping the memory from forgetting its contents.

A few alternative technologies are in development to make RAM a bistable storage medium. This means that no more energy is wasted on storing data. The technologies are new, but are all better than flash, to which it does have some resemblence. Flash memory is not very fast, and it can only be rewritten a few million times. When bistable memory technologies become commonplace, then energy will not constrain the amount of main memory and nothing is lost when the power drops. This is pleasant during energy supplier problems, but it also gives an opportunity to save a lot of energy.

This chance to suspend the computer's operations does not have to be limited to RAM; the computer is full of little buffers that store intermediate data, including in the central processor. All components that function according to bistable technology can be switched on and off at any time, and just continue their work as if no power interruption took place. Using ACPI it is already possible to switch a computer on or off in parts, so we do seem to be ready for such smart components. It is a matter of time until a stable and massive production process is setup for one of the many ways of achieving bistable RAM, and until it finds its way into mainstream computer components.

Solid State Disc. The main difference between RAM and disk memory is that a disc is a mechanical device. A few magnetisable discs rotate and a read/write head moves over it between circular inner and outer tracks. Except for the risks of damage to this fragile construction, the mechanics also have the disadvantage of delays. The head must move back and forth, and await the right data to rush by it. It takes quite a bit of energy to swiftly reposition the head, and the motor that spins the discs also consumes electricity. That is why a technology named Solid State Disc is currently rolling onto the market. It is flash memory that interfaces towards the computer as though it were a disc drive. Clever use of the available capacity means that the limitation of several million rewrites can support continuous writing for 25 years or more. On top of this, an SSD consumes about half the power of a normal disc drive. SSDs are relatively expensive at the moment, but they will undoubtly break through in the coming years, certainly for use in laptops.

All memory on one pile. Although SSD appears to be a good solution, it probably still is a temporary solution. If bistable RAM becomes commonplace, it can also overtake the functions of the hard disc. The somewhat arbitrary separation between main memory and disc as we currently know it will disappear, because every change is stored in non-volatile memory anyway. This means that your text processor will not expect you to Save your documents, but simply keep everythin that you type. To support getting to older versions there will be a powerful Undo possibility, or better even, Version Control to get back to older versions of your text and to compare what has changed since then.

Interestingly, non-volatile memories used to be quite common. In ancient times the main memory of computers were made of magnetic rings that formed ring core memory. This memory also retained its contents because it too was bistable: a ring was either polarised clockwise or counter-clockwise.

The vision given above becomes possible as soon as all memory is electronic and bistable. Devices can always remain switched on, or they can switch themselves off until an external events switches tham back on to continue as though nothing had happened. As an example, it would be possible to switch off between two keystrokes, as the time between those is enormous to a computer, and it could take a nap instead of actively waiting for the next keystroke. An interesting result of all this is that a very quick computer would not be a waste of energy, as the power used would entirely depend on the number of actions to perform, rather than on the time the computer is switched on.

It is to be expected that any such computers work under open source software like Linux. Not only is it very stable, it is also the place where most work is done to reduce energy consumption. Attention for these aspects has always existed, and that makes sense: A smart programmer building an energy-saving trick into Linux can see these savings multiplied by millions, if not billions!

Budgetting. The architecture sketched so far makes it possible to apply budgetting to computational power, for example to better spread the available energy. This could enable devices to limit their energy to that won from wrist movements, ambient light or perhaps even the signals on a network interface. An energy-efficient device could become a zero-power computing device if it can restrain itself to these external sources of energy. Although not all devices will be able to achieve this, it still is an extra option caused by the ability of a device to switch itself off without trouble.

What's the time? It has been a property of computers since the very first decent-sized models that every period of the electricity network causes an interrupt; this means that the computer temporary tends to some regular maintenance like incrementing its internal clock and perhaps switching to another process. Such interrupts occur 50 to 120 times a second, and most of the time no work is done. This means that it is usually a wasted effort, especially if it wakes the processor from a low-power resting mode. This conflicts with the vision of a computer that consumes no power if it has no work to do. Luckily, it also causes other problems.

For these reasons, work is done (at least within Linux) to come up with an alternative technique, in which the computer need not generate 50 to 120 of those interrupts a second, but only if there actually is some work to do. All hardware generates an interrupt when, say, a key is pressed, a mouse moved or data sent to a network interface. For its internal tasks the processor has timers available, which too can trigger an interrupt on the processor.

All this will be done with Linux on the central processor(s), but perhaps not on the rest of the electronics. This leaves another opening for improvement, in devices like keyboards and mice that could also be designed for zero-power consumption when in rest. This is currently not the case; minimal computers called "microcontrollers" are embedded in such devices. These microcontrollers constantly scan if work is ready for them. It could be necessary to use a special network adaptions to make it even possible for the protocols between a computer and an interface device (perhaps ATM?) but it does look like a place where savings can be made. Also the area of routers on the Internet, and the modems used in our homes could transition to a zero-power resting situation which could save a lot, globally. This is likely to be a very long-term plan because it takes quite a lot of work to make the transition, but there is no reason not to look for it as a long-term goal. ATM run over fibre networks is technically a sensible technology, and appears to be the proper path towards the future.

Asynchronous hardware. At a basic level, hardware uses a regular heartbeat on which it operates: the clock speed that makes the processor tick. This speed is often specified in GHz, or billions of steps the hardware takes each second. That is quite a lot! Especially considering that a processor consumes a bit of current to toggle from "1" to "0" and/or back. The ideal is to avoid this whenever possible.

Asynchronous electronics makes this possible, in principke. This technology is more difficult for designers, but it is possible to build similar devices with it. The idea is that pieces of electronics tell each other explicitly if some data is available, instead of pumping it around at every clock tick. Cautious steps are being taken towards this technology. Philips has developed a processor along those lines, and it has been shown to stay a lot cooler (read: consume less energy) than the comparable processor designed under the standard design regime. The first manufacturers are offering asynchronous, programmable hardware (FPGAs) and a lot of research is devoted to design methods for asynchronous electronics. You won't find it at your local computershop just yet, but it is definately coming. This is one more development for which we can thank mobile device development.

Wireless. A rather noticeable energy consumer in many devices is WiFi. It is probably possible to do something about that, but we currently lack information on this. A mobile device that uses no energy while nothing is sent to it would still make it possible to be reachable; wires are even more suitable but do not cover all application areas.

An aspect that plays a role in wireless power consumption is the distance that we want to travel with a device. The shape of the antenna is another factor; an aimed antenna consumes less energy to bridge a distance, because the radio waves are transmitted only to the intended recipient and not to everything around it. A combination of aimed transmissions and technology with a limited range could be the key to wireless networks which consume less power. BlueTooth may well be the technology to implement this.

Where?

• A more detailed report by us on Computing without Breaking a Sweat.
• Electronic paper on WikiPedia.
• Solid State Drive on WikiPedia; also note the remarks about possibly using more power when badly implemented on Tom's Hardware.
• Flash memory and modern non-volatile memory on WikiPedia; especially interesting are FeRAM, MRAM, NRAM, PRAM, etc.
• The Mini ITX is a series of very energy-efficient mother boards.
• Linux Devices shows a lot of devices with modest energy use for embedded applications.
• Information about Interrupts such as the clock pulse (in the context of Linux).
• Information about Asynchronous FPGAs -- generic chips onto which asynchronous circuits can be burnt.
• News story: ARM-Powered Linux Laptops