[Arm-netbook] proposal for GNU Netbook project

[Luke Kenneth Casson Leighton]

On Fri, Mar 26, 2010 at 4:07 PM, Mark Constable <markc at renta.net> wrote:
> On 26th March 2010, Luke Kenneth Casson Leighton wrote:
>> > Maybe we should REALY design a simpe Netbook our own
>>
>> yes - i have contacted richard stallman to see if it could
>> be listed as a GNU Project.
>> ...
>> the reason for doing that - listing it as a GNU project -
>> is to get some PR.  the point is that we need to reach
>> "critical mass" of at least 1,000 to 10,000 units, with
>> at least 60% preorders.
>
> Going the GNU accreditation route may, or may not, help with
> getting minimal critical mass. It's one path. To me, if I can
> offer a few thoughts, it simply comes down to price/performance
> and in the case of an open startup concept device, whatever is
> the simplest base level of servicibility. Once an initial product
> is actually put into the hands of end users then the followup
> products can, obviously, become more featureful with actual
> experience and hindsight dealing with the whole production chain.
> The trick is to get *anything* off the ground to start with.
>
> a) so what is the minimum useful hardware device that could be
>   produced at the cheapest manufacturing costs?

 in netbook terms?  833mhz S5PC100, 256mb RAM, 9in or 10in LCD, 2gb
NAND: $102 for 20k and make a small profit, maybe 3%.

 in "pc" terms? ncomputing are going to be doing a dual-core 1ghz ARM
SoC for $20 called the "nuva". with other components being $10, you'll
get a stonking HDMI video capable thin client for $30.  add a bit more
and you've got something a bit more useful.

 in "available now" pc terms, i.e. using e.g. 833mhz S5PC100, you
could possibly get away with around $60 (remove LCD, keyboard, mouse,
and lose the LVDS chip, from netbook above, basically)

 in "available _really_ right now" pc terms, that'll be the
sheevaplug, priced retail at $99.

 but it has to be said - this list is called "arm netbooks" for a reason :)

 also it has to be said, you _really_ want to watch out for tooling
costs.  $50,000 for a single case design is not unreasonable, because
getting the plastics right (as they shrink) is a black art.  you can't
just "make a straight 1.1-to-1 mould" and heat-shrink it, you have to
make a "curvy" mould that, when it is shrunk, the stresses pull
inwards and against each other, the end result of which is that the
plastic case _happens_ to be straight!

> b) what is the cheapest way to get that product to end users
>   anywhere in the world?

 if you're doing an entire container-load - 50,000 units - there's a
company i have heard of that will do door-to-door delivery for $5 and
door-to-shop (multi-box) delivery for $1.  goes by sea,


> I'd love to buy, *right now*, a sub $200 USD so called smartbook
> that does 1080p playback at 1920x1080 on a 10" screen.

 that's very unlikely to happen, and if it does, the cost will be well
above $200.  the reason is simple: the transistor in each cell is a
fixed size.  each cell size on e.g. a 1024x600 10in LCD, that's
actually 3096x600 cells, sized 0.25mm x 0.08mm.  the transistor takes
up ... what... 25% of that cell? (just guessing).  now you want approx
double the dimensions in each direction, so that's a cell of size 0.12
x 0.04mm, and that's almost the size of the transistor itself.  you've
no room for the actual liquid crystal.

then, even if you could shrink the transistor, you have the issue of
forcing the liquid crystal between two glass panes so that it spreads
across and fills all those 0.12 x 0.04mm cells, of which there are
just over two million on a 1920x1080 screen.  the number of times in
which this is successful is going to determine the "yield" rate
(factories simply smash the glass and melt it down rather than sell
"faulty" screens).

the end result is that the $32 worth of glass + liquid crystal +
quarter of a million transistors in a 10in 1024x600 screen goes up to
"being reprocessed dozens of times; two million transistors _and_ the
liquid crystal gets thrown away each time", and now you have something
that costs $200 to $250 _even_ if it could be made successfully.

_then_ there's the fact that 1920x1080p video playback requires
hardware acceleration, with not even 2.8ghz dual-core pentiums being
able to keep up with the 60fps playback rate!  so, ironically, it's
the high-end ARM cpus with hardware MPEG decoder blocks that stand a
better chance of achieving what you're asking for, but even then
you're pushing your luck for ooo... about the next...
*finger-in-air*... six to twelve months.  (an OMAP3530 720mhz _just_
about does 720p, and it has one of the most powerful and affordable
DSPs in the world; you _might_ get lucky with an OMAP43xx series, and
guess what - those will start being commercially available somewhere
in the next twelve to 18 months).

> So, as a startup suggestion, a) could be a small lightweight
> minimalist ARM cpu (1/2 dozen to chose from), with 512mb of ram
> (256Mb x 4 density ram from Samsung is too expensive at this point

 hynix DDR2 1gigabit is $2.30 P/N H5PS1G43EFR

> I vaguely estimate this part could be $60USD landed in HK at 1000
> parts if some factory in China (or Taiwan) happens to have even a
> devel prototype board and case that could almost fit the need.

 i'd agree with the $60 estimate.  i'm not up to "second-guessing"
what people want: i'd rather it was a netbook but hey if there are
9900 more people who want a low-cost... "thing", then great! who am i
to stand in the way :)

 the thing is that $60 FOB _turns into_ $99 retail (ok maybe $85 to
$95) when you add on a small but reasonable development cost, small
but reasonable profit margin, plus low-volume shipping costs plus
import duty.

 so to get that down, i'm proposing local assembly of "parts" thus
doing away with import duty; even (if michelle can get the machines
and if we can justify the cost with enough orders - 20k+ would do it,
we buy them for her) local manufacture of the PCB and even the case;
free software licensed designs to encourage people to build their own
machine, thus keeping costs down; etc. etc.

l.