Wednesday, February 04, 2009

The $20 Laptop

India thinks it needs a $20 laptop to educate its population.

India is planning to produce a laptop computer for the knockdown price of about $20, having come up with the Tata Nano, the world’s cheapest car at about $2,000.

The project, backed by New Delhi, would considerably undercut the so-called “$100 laptop”, otherwise known as the Children’s Machine or XO, that was designed by the Massachusetts Institute of Technology of the US.
Electronic Design News has some interesting views on the subject:
There were no details available in the story about just how the Indian designers intended to hit that price point. But apparently it's not vaporware. The story claims that a prototype will be available on display at an education conference tomorrow. There appears to be no commercial backer to build the design in quantity yet, however.

None the less, and even if the project ultimately proves too ambitious, there is an important message here. When we are addressing the developing-world market, we cannot afford to make the assumptions that we in the US don't even recognize as assumptions any more.

Such as? For instance, take the assumption that a computer necessarily implies an Intel/Microsoft computer, or even an x86-based CPU. Clearly, if you consider the tasks necessary for e-mail, facebook, google, or distance learning, nothing from the Intel or Microsoft camps can be even remotely justified. A sufficient CPU costs pennies, lodged in the corner of an SoC that costs a few dollars. A sufficient amount of memory costs a few dollars more. Take out the mediocre mechanical keyboard and the pointless mouse, replaced by a cheap membrane keyboard, and the bulk of the bill-of-materials cost of a really lean netbook design will go into the display and power supply.

But all these costly components are things we assume must be in a notebook PC, and therefore in a netbook as well. When we do so we are wrong. The message here is not that Intel and Microsoft are soaking the industrial economies for a fortune in unnecessary costs (although that might be an interesting discussion. Please feel free with your comments.) The point is that when we approach a developing-world market, we must reason from first principles. And those principles are based on the actual user's needs, not on how we would do it in Silicon Valley or Cambridge.
OK. Where would I start? With the processor. The SeaFORTH chip looks like a good start. About 25 billion operations a second. At 360 milliwatts of power. And that is peak. it goes down to 360 microwatts while the processor is waiting for something to do. And it only uses the power required for what ever process is running. Automatically. What would I do different? Add more RAM to each of the processors (there are 40 of them), go to a 32 bit address bus and a 32 bit data bus (at least for external access), and go to the next size smaller semiconductor lithography process node to shrink it all. Put it in a cheap package - 100 pins would be nice. Maybe 128 if it eliminates other hardware. Four external interfaces. USB, wireless, microphone jack, earphone jack. A keyboard and a touch screen display. FLASH to take the place of a hard drive. Maybe 1 G byte. Maybe 1/4 of that. RAM of about the same size.

In the classroom: Custom designed server with 40 USB connectors. Several 1 T byte hard drives in a RAID configuration. For student use and for local storage of info to reduce Internet bandwidth required to support classes.

The server could be multiprocessor if that was a convenient way of getting it up faster. Ultimately server power must be reduced. If each of 40 laptops consumed 1/2 watt on average it would be nice to have a server that averaged 40 watts or less. All powered by a solar/wind power plant capable of 2,000 watt hours a day. At 6 hours a day charging time a 500 watt array should be able to keep a 100 Amp-hour deep cycle 12 volt battery charged up without any trouble most of the year. That should suffice to keep the server up and to charge the laptops as well. A solar array on the cover of the laptop might be interesting as well. I wonder why no one is doing that? Which kind of reminds me of something I had worked on earlier with somewhat more grandiose ends in mind. The Neighborhood Development Package. Evidently they want to start smaller. Which is probably a very good idea.

Cross Posted at Classical Values

1 comment:

Habib said...

Before I start jumping out of joy - I would like to know:
1. Can the child use it in bright sunlight?
2. Can it be used under rainshowers
3. What's the battery life?
4. Does it have features that surpass OLPC features?

Thank you

Dr Habib Khan (PhD Harvard)