Background

The world is changing: we know that oil production cannot keep up with
demand.  People's attitudes to consumerism are changing, fortunately,
as the effects of recession bite.  However, there hasn't yet been a shift
in manufacturing and design strategy to take into account these world-shifting
changes, and we are in danger of clinging onto computers that will be difficult
to run, let alone maintain.  This article therefore advocates the combination
of three powerful approaches, to meet the upcoming very challenging scenario:

* The near-exclusive use of Free Software, for its community diversity 
* Environmentally-conscious designs such as the Bloom Laptop
* Modular Electronics, based around high-end Embedded "System on Chip" CPUs.

By following this approach, the days of consumerism (where products are
deliberately designed around a "one-shot deal" with failure and obsolescence
built-in) are numbered, and are replaced by products that can be upgraded
as finances, need and resources allow.  Older parts can be replaced, or reused,
or sold second-hand or donated for use elsewhere in the world.

Note to people following the FreedomBox: the same modular design means re-use of
electronic components to create a Desktop PC, Plug Computer, Router Box or NAS
Storage Box - anything.  all with re-useable, upgradable parts, especially and
including the CPU card. Yes, you can get SheevaPlug - it costs $99 not $30,
and it overheats.  If it was modular, a new chassis could have been made,
without needing to redesign the electronics.  If it was modular, the "plug"
chassis could be re-used when a cheaper CPU was available...

Free Software Laptop

There does not exist, anywhere in the world, an affordable yet non-Intel-based
useable Laptop.  "Useable" is defined, strictly, by having a processor that
is capable of running an entirely Free Software Operating System, and having
a screen of resolution exceeding 1024x600.  There do exist laptops
with this consumer-grade screen resolution:

* The Leemote (10in LCD, 900mhz 64-bit MIPS Loongson 2F)
* The HP Compaq Airlife 100 (10in LCD, Qualcomm "Snapdragon" - Cortex A8)
* The Toshiba AC100 (10in LCD, 1ghz Dual-Core Cortex A9 Tegra 250)
* The AlwaysInnovating Touchbook (9in LCD, 720mhz Cortex A8 OMAP3530)
* The Genesi-USA Ekiga (9in LCD, 800mhz Cortex A8 iMX515)
* The OLPC XO-1.75 (9in LCD, 1.2ghz Marvell Armada 610)
* The FirstView PC101v (10in LCD, unspecified CPU)
* The Next SurferPro (10in LCD - product withdrawn due to mis-advertising!)

The manufacturers of these laptops are wished every success with their
products, in fulfilling their market goals.

The Free Software Laptop recognises that the cost of a laptop can be
brought significantly down by replacing the internal Intel-based design
with a far simpler, lower-power and lower-cost "embedded" processor.
The OLPC XO-1.75 demonstrates that it is even possible to achieve a
total power budget of 2 watts, to outperform an Intel Laptop running
the same Operating System - and to achieve a Bill of Materials that is
at least 20% lower than an equivalent Intel Laptop.

It is quite easy to find reports that show the 1024x600 resolution to
be woefully inadequate for day-to-day usage, and so the Free Software
Laptop will have a decent screen size (12.1in) and a decent resolution
(1280x800).

In order to keep the costs down, an existing "no-brand" Laptop Chassis
will be chosen (by the simple, arbitrary and expedient criteria of selecting
the first Chinese Manufacturer willing to collaborate with the project).
As the OpenPandora Project shows, the cost and time consumed in creating
a laptop plastic mould is... extreme.  Given that there have been
thousands of laptop designs ever made, it is a significant waste of
time, money and resources to create a new one.  Thus, an existing one
will simply be adapted.

Costs will be further reduced by keeping to a 2-board design.  The
CPU, NAND and DDR2 RAM will be on a 35x67mm 200-pin SO-DIMM, which will
need to be at least a 6-layer PCB.  This very small board will be the
most complex part; all its interfaces will come off the connector
as balanced line pairs (Ethernet is 100mb/sec, USB-2 is 480mb/sec),
or as low-speed signals (24-pin RGB/TTL at 65mhz is probably the fastest).
The SO-DIMM will be inserted into a standard socket on a much simpler
4-layer motherboard.  The most complex and highest speed parts on this
board will be the USB-2 Hub ICs.  Even the WIFI will be a separate
low-cost (and Free Software-compliant) self-contained and off-the-shelf
USB-2 module, thus sticking to the simplest possible development path,
minimising costs and increasing reliability and maintainability.

Further reductions in cost are achieved by utilising a low-power CPU.
The reduction in power consumption is so high that it is unnecessary
to even have a heat-sink, let alone a fan.  Cost is also reduced by
providing some NAND Flash and the option to boot from an SD-Card,
instead of a mandatory Hard Drive.  The end product therefore relies
on passive cooling, has no moving parts, and is thus again far simpler,
easier to design, easier to assemble, has a longer battery life - the
list of comparative superlatives goes on until it becomes bewildering
and close to being "too good to be true".

What has prevented and prohibited anyone from creating such a laptop,
before, when the benefits and cost savings are so blatantly obvious?

The answer is clear: the monopoly situation on Operating Systems.
When there is only one dominant Operating System, which only runs
on x86-compatible hardware, it is very difficult to justify the creation
of a laptop based around hardware that is incapable of running that
dominant OS!

So what has changed?

Again, the answer is simple: the price-performance metric is crossing
a threshold that makes it very very challenging for consumer electronics
manufacturing and retail companies to ignore.  If an entire laptop has
a Bill of Materials of $USD 90, and has a 1ghz processor, how long can the
commercial opportunity be ignored!  (And, crucially, how can a proprietary
OS with a pricetag of $30 and above ever possibly be accommodated?)

From a technical perspective, Processor Geometries (translating to raw
speed, effectively) have reached the point where the old "RISC"
architectures (MIPS, ARM, PowerPC) are, pure and simple, capable of
operating at undeniably acceptable speeds.  The OLPC XO-1.75, as well as
recent proprietary products from Apple, along-side modern "Smart Phones"
with the Android OS, prove without a shadow of doubt that these so-called
"embedded" processors can do the job of "full-blown" x86 equivalent CPUs.

In fact, the new 28nm geometry, recently announced as having been proven
by GlobalFoundries, is capable of making these RISC CPUs run at a whopping
2.5ghz, and the fantastic thing is that they will consume half the power of
the same CPU in 45nm!  Given that MIPS has a Quad-Core 64-bit CPU design,
and so does ARM, it can be clearly seen that there is truly no longer
an overwhelming and pressing need to squeeze out that last drop of
performance.  (Intel's CPUs use techniques such as having 5 Floating Point
Units designed by separate teams, having them all on-board the CPU, and
asking every unit to perform the same calculation.  As the units
are all by different teams, with different designs, they have different
advantages, given different ranges of inputs.  The fastest to complete the
calculation is chosen, each time, and all other partial answers abandoned.
The resources wasted, all in the name of "speed"...)

So, it is clear: Processor Geometries from 20 years ago were so crude that
x86 CPUs were comparable to RISC CPUs, and so the OSes that ran on them
were dead-level (from Atari, Commodore etc.).  The intervening years allowed
CISC CPU designs to gain at least an order of magnitude advantage, and any OS
that didn't shift to x86 was dead (i.e. all but one).  Now, IC Geometries
allow RISC CPU designs to reach "good enough" speeds, whilst at the same time
being far more power-efficient.  Whilst x86 CISC CPUs can also benefit from
the same Geometry benefits, the sheer overwhelming complexity of the
resultant design and its complete overkill performance puts them into
an ever-shrinking specialist market.  (It will be interesting, in particular,
to see what happens to AMD...)

It is ironic that technological advances, driven by consumers'
"need for speed", has been far more effective at ending the ongoing
OS Monopoly situation of the past two decades than any "Monopolies and
Mergers" or Cartel investigation by the EU or the USA could ever be.

Specifications

* CPU: likely to be a 1ghz ARM Cortex A8 - DM3725.
  (options being evaluated here)
* 1gb RAM
* 1gb NAND Flash (TBD)
* 12in 1280x800 LED-Backlit LCD from Chimei
* OEM "no-brand" Standard Laptop chassis
* Option for any standard SATA-II 2.5in Hard Drive to be fitted

Bill of Materials

* 35x67mm 200-pin SO-DIMM PCB (8-layer?)
* 1x DM3725
* 1x Power Management IC
* 4x H5PS1G43EFR-S6? 1gb DDR2, Hynix
* 1x TBD, NAND Flash IC
* 1x SN75LVDS83B (RGB/TTL-LDVS), TI
* 1x GL831A, PATA-to-SATA, Genesys Logic

* 60mmx100mm (approx) Motherboard (4-layer)
* 2x GL850G, 4-port USB Hubs, Genesys Logic
* 1x TBD, Battery Charger IC
* 1x TBD, LED Backlight Management (DW8400?)
* 1x TBD, I2C EEPROM
* 1x GL632, USB Audio IC (Stereo), Genesys Logic
* 1x SO-DIMM Socket, (FoxConn?)
* 3x USB-2 Socket
* 1x HDMI out Socket
* 1x SATA-II Socket
* 1x RJ-45 (Ethernet) Socket

* 1x OEM "no-brand" Laptop Chassis (w/keyboard, battery etc.)
* 1x N121I6-L01, 1280x800 12.1in LCD, Chimei

Approximate Costs

These are estimates excluding NRE (setup / development) costs:

* SO-DIMM: approximately $USD 45.
* Motherboard: approximately $12.
* RA-Link RT2070 USB WIFI: appx $5.
* Chimei 1280x800 LCD: appx $65.
* Laptop Chassis: approx $5
* Battery: approx $10
 
Estimated total cost (excluding labour, tax, profit, shipping): $143

Note: Low-volume quantitiy markups could increase this amount by at
least 20-25%.  Regardless, this BOM cost makes it perfectly reasonable
to expect a final retail price of somewhere around $230 to $270.
Given that 12in Intel Laptops can retail for $400 or more, it is a
significant eye-opener to see these kinds of figures, even if a
Hard Drive and a DVD Drive are excluded from the BOM.

Sponsorship

If you would like to sponsor this project, please do email me.
So far, as of Feb 1st, two offers have very kindly been received: one
for $1,000, and another slightly larger one which is conditional on the end
product being entirely Free-Software-compatible.

Links

Some very useful information can be obtained from the design of the Odroid,
including convenient links to components and datasheets, as well as
significant resources from Texas Instruments for the Beagleboard-XM
and the DM37x Schematic Checklist

Forum Discussions and Threads
* gp32x
* Debian-ARM
* ubuntu hardware forum
* fedora arm