By James Turner
Senior Editor
The Final Design of the XO
I have seen it, touched it, and played with it. The final
industrial design prototype for the XO, the device that the One
Laptop Per Child (OLPC) Initiative is going to start shipping to
countries across the world this summer. AMD hosted a luncheon on
Monday to give the press an update on the project, and to unveil
the completed design.
Although the exterior form factor is now pretty much set
(“Unless,” said OLPC official Michalis Bletsas half-jokingly,
“Nicholas has another late-night inspiration.”), there is still
work to be done on the software package and the hardware internals
before the planned summer volume rollout of the unit.
AMD’s Chief Innovations Officer, Billy
Edwards, Lays Out Their Strategy for 50% of the Earth’s Population
to Have Internet Access By 2015.
According the Bletsas, the Beta 1 units we saw had
manufacturing screwups by several of the integrators. “That’s why
we have betas,” he commented. For example, the touchpad buttons had
been recessed rather than protruding, something that I noticed
immediately when I tried to use the XO. With two more betas to go
before the summer, Bletsas was unfazed by the glitches. He also
called the current state of the software “barely useable,” but
again was confident that it would be where it needed to be by
launch.
Michalis Bletsas, Chief Connectivity Officer
for the OLPC
Only 750 of the Beta 1 machines were built, a limitation forced
by the scarcity of a programmable logic chip used on the
motherboard. Most of the Beta 1s went to beta testers in the field
to gain comments. Bletsas commented that by doing this, they gained
much more useful feedback than they would have with an in-house
beta.
Many of the specifics of the OLPC XO were discussed at the
luncheon. The final selection for power generation has yet to be
made; it will be a yo-yo-like device that can be pulled by hand or
foot, with a strap that can attached it to a belt or table. The
yo-yo generates around 10 watts, while the XO consumes a mere 3
watts in non-intensive computing. This means, for example, that
that for every 10 minutes of power generation, a child should be
able to surf the internet for a half hour. The yo-yo is designed by
Squid Labs, the first engineering prototype was delivered two days
before the meeting. Bletsas said that the yo-yo was also being
developed for commercial sale. “You could charge your cell phone in
like five minutes,” he said. The charger is on course to cost
around $10 in parts, he indicated that a commercial unit wouldn’t
cost a lot more than that.
The XO Running in Tablet / e-Reader
Mode
Inside the XO is a 23-watt Nickel-Metal Hydride battery than is
good for about 15,000 cycles. Running at the highest consumption
mode (watching video, playing games, or listening to music) the
battery can operate the laptop for around six hours on a full
charge. When the unit is idle and providing mesh network service,
the usage time rises to over a day. The battery can be fully
charged in around two and a half hours using the yo-yo, according
to Bletsas. It’s also designed to burn at only 106 degrees C, a
much safer option than a Lithium Ion battery if the XO is thrown in
a fire or abused.
As mentioned, the power profile of the unit changes dramatically
depending on what the child is doing. Peak consumption is around 5
watts for high-demand media applications, it falls to around 3
watts for browsing, under a watt when used as an e-Reader in black
and white mode, and only 350 milliwatts to participate in the mesh
network. Keeping the power needs low as a mesh repeater was
critical, because the chosen networking design works better the
more nodes are available, and the longer they stay online (“stay as
dense as possible as long as possible”). Because the network can
operate without requiring the main processor to run, children won’t
need to worry that letting their XO participate in the mesh will
drain the battery significantly. The radio itself is under $10 in
production costs.
The Meshes Connect to Each Other and the
Internet Through a Variety of Means
Keeping the 377-Mhz AMD processor processor from having to sip
from the battery was a key concern. When used as an eReader, a
separate frame buffer keeps the LCD updated, rather than having the
processor do the job. This drastically reduced the power needed to
keep a page displayed while a child reads. In general, the XO uses
what Bletsas calls “Extreme Suspend,” going to sleep after two
seconds of inactivity, but waking up within 300 milliseconds of an
action.
The innovative display design serves two purposes. By overlaying
a lower resolution (1024×768) color screen over a very high
resolution (200 DPI, 1200×900) black and white one, then can get
what to the eye appears to be a much higher color resolution. The
color display is transmissive and requires a 1-watt backlight,
which is provided by power-efficient LEDs. The black and white
display is reflective, and actually performs better the brighter
the ambient light is. This makes it ideal for rural teaching
settings, where classes may be held outdoors. Switching from color
to black and white is simply a matter of turning off the backlight.
The ultra-high resolution black and white display is meant to make
the eReader highly useable for textbooks. Bletsas notes that
although the display cost only about one-third what a typical
laptop LCD costs, it will have a higher resolution that 95% of the
laptops on the market.
The software is based on Fedora Cora 6, put on an diet to reduce
it to 150 MB, and leverages Python heavily. According to Bletsas,
both Microsoft (WinCE) and Apple (OS X) offered their operating
systems, but neither fit the footprint or security requirements
that the XO demanded. In addition, the closed-source nature of
those operating systems wasn’t a good fit to the OLPC
philosophy.
The application environment looks nothing like a typical
X-Window GUI that you or I have ever seen. Written menus are
totally replaced with icons. In one example screen, the child can
view all the other meshed XOs around them (the mesh is good point
to point to about 600 meters), and see what activities the other
children are involved in. Almost all activities can be done
collaboratively. So, for example, multiple children can work on the
same document or browse the web together. The distributions are
fully open source, and can be downloaded and played with now at
laptop.org. Also included will be
a Gecko-based browser that Bletsas told me should be capable of
displaying Flash-enabled web pages.
Viewing Other Children in the Mesh. Clusters
Indicate Children Doing Things Together.
Bletsas says that the design philosophy tries to leverage
Moore’s law in the opposite direction from that being taken by
traditional laptop manufacturers. “When I bought my first laptop,
it was $3,200. The last one I bought was $3,100. My latest on is
10,000 times as fast as my first one, but still takes the same
amount of time to boot.” Bletsas says that laptop manufacturers
have tried to cram more and more into their products, rather than
use the falling cost of existing processors to produce a cheaper
product. He acknowledges that the XO is not designed on intended
for power users in the developed world. “You have to look at this
through the needs of a child [in the developing world]. A child
doesn’t want to play the latest video games. he wants to be able to
read a book.”
The XO includes a Video Camera, Microphone,
and Speakers
Also being developed is a central server designed to sell for
around twice the cost of an XO, with a laptop hard drive and
running under 5 watts in operation. Although the intent is that
each mesh (which should correspond to a local school) will be
connected to the Internet, the OLPC is not mandating a single
solution. According to Bletsas, there are so many peculularities to
each countries information infrastructure, that it really needs to
be handled on a country by country basis. Among the connectivity
solutions that are being suggested are satellite, WiMax, and
cellular. The prototype server has already been tested with
satellite receiver boards. The key again, says Bletsas, is that the
connectivity can’t require huge amounts of power. He says that when
a WiMax provider tells him that their central facility will require
“only” about what a hair-dryer uses (1000 watts), his initial
reaction is “Where am I going to get that kind of power?” He
continues that the OLPC has been aggressively pushing its
connectivity partners to adopt the same power-miserly approach that
went into the XO.
Because of the emphasis on the XO as an eReader, content is
obviously key. One third of the 15 full-time employees at the OLPC
work on content development. According to Bletsas, Latin America is
in the best shape, due to Mexico’s aggressive initiative to produce
an electronic library of all of their text books. The OLPC is also
working with the Wikimedia Foundation and Google to make locally
cached versions of the databases available on the local school
servers.
A Closer View
Bletsas doesn’t believe that the XO is at risk of becoming the
latest target for botnets. “For one thing,” he comments, “Why would
anyone try to create a botnet on 377-Mhz laptops with a relatively
low-speed Internet connection when there are all these unprotected
computers already sitting around.” In addition, the design of the
operating system protects against infection by running each
application in its own virtual machine, preventing any potential
security loopholes in a given application from spreading to
others.
The OLPC is still working on bringing down the cost of the XO.
“The first units will be closer to 100 Euros than 100 dollars,”
admits Bletsas. They hope to drop under the magic $100 figure
during 2008. One way that costs are being kept down is to deliver
the units en-masse to governments for delivery along the same
channel as they currently use for textbooks, keeping the OLPC out
of the distribution business. “If we were selling this laptop
through normal consumer channels, it would be more like a $250
laptop.”
In addition, Bletsas indicates that the units have been designed
around low-failure operation, with no moving parts. For example,
the motherboard sits directly behind the LCD, avoiding the need for
a failure-prone connecting cable. By giving the laptops to the
children to own rather than the school, the OLPC also believes that
there will be less vandalism on careless treatment. About the only
part worth replacing if it fails is the LCD, Bletsas says that
there will likely be central repair depots set up by the
governments to handle the repairs that make sense, otherwise failed
units will just be replaced.
Bletsas also envisions the children becoming their own support
network, trading tricks and tips with each other, and actually
becoming peers with their teachers. “Teachers tend to learn much
slower than the kids themselves,” he notes. He says that waiting
for teachers to become fully trained in the use of the XO would bog
down adoption significantly. He likens the training and support
model to a peer to peer network. “We believe in empowering the
kids.”
Members of the Press Question
Bletsas
Uruguay, Brazil, Argentina, Pakistan, Nigeria, Libya, and (most
recently) Rwanda are the countries currently signed up for the
program. Bletsas indicates that the program will actually be supply
rather than demand limited in 2007. The goal is to deliver five
million laptops within a year of the summer rollout, a goal he
notes will represent a 10% increase in the total worldwide laptop
production for the year and will be the largest single deployment
of a computing platform ever. The units will be divided up between
the program countries rather than concentrating on any one to the
exclusion of the others, with countries encouraged to deploy to
entire schools rather than cherry-picking students, to avoid envy.
The only requirement being placed on the countries is that the
laptops must go directly to the children.
Bletsas acknowledges that some abuse is inevitable. “Will some
parents sell their children’s laptops on the gray market? Sure.” He
also stated that the OLPC would rather engage with countries that
might attempt to filter or censor information access, rather than
isolate them. He notes that as soon as you give any Internet access
to people, they tend to figure out how to work around whatever
restrictions the government may have placed.
Editorial note: Because of the length of this report, I will
be combining my report from Show Stoppers (there be Linux there,
arrrrgh!) with my second day show floor report, which should be
available Wednesday morning.