[Arm-netbook] Anyone here made a "TV computer"?

[Luke Kenneth Casson Leighton]

On Mon, Nov 7, 2011 at 5:20 PM, Bari Ari <bari at onelabs.com> wrote:
> Am I missing something here or is the Allwinner A10 ARM SOC just another
> ARM SOC design that has no public docs, and with just another board
> design that green engineers are learning from?

 naah, that pretty much sums it up :)

 the whole point of the exercise is to cut the "core" out - almost
cookie-cutter style - and get it onto a smaller PCB.

 see below for insights into the situation, but the problem is that
there's a really _really_ limited window of opportunity in which to
take advantage of this CPU.   we haven't got _time_ to worry about
these things.

 and, also, we haven't got time to do "total product redesigns", hence
the modular architecture.  just do the main CPU board, shove it in and
sell it.

 but yes, as this is the first, we just have to see how it goes.

> What are the problems with doing a respin of the Freescale IMX53QSB?
> http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=IMX53QSB

 it would be lovely to have that... additionally (not instead of:
that's the beauty of modular architectures).  the pricing is higher,
comparatively, but if it happened to be available _great_.

> or the BeagleBone?
> http://beagleboard.org/bone
>> The BeagleBone uses a TI AM3358 ARM Cortex-A8-based microprocessor.

 yes, that one again would be great to have as well, but right now
that $5 price applies to the low-speed version (500mhz?) and that's
DDR2 RAM, the allwinner CPU can take the (faster, cheaper) DDR3 RAM.


> Are the docs not publicly available? Are the devices priced to high or
> not readily available? Open software support also appears to be
> available except for the GPU. Is this not the case?

 ah, it [allwinner a10] uses MALI.  so, it's supported.  even if it's
a non-free "system library" which qualifies as acceptable under the
GPL's "system library" exemptions.  not great, but then every single
GPU is in the same boat, so... *shrugs*.


> I'm just trying to find out what is actually missing from all these ARM
> SOC's.

 :)

 it's a price-performance comparison, the allwinner just comes out on
top.... for now.  the risk is actually that _another_ CPU comes along
before the prototypes are completed!

 to give you some idea of how seriously the allwinner a10 is being
taken: *everyone* in china has cancelled orders of everybody else's
CPUs, reneged on contracts and ordered this CPU instead.  that's
causing some quite serious problems because loads of companies are now
holding stock of parts that they can't shift.  they'll go eventually
at rock-bottom prices because that's how china works, but anyone who
has completed systems has slashed prices in an attempt to get rid of
them before allwinner a10 systems _really_ take a hold.

 this is going to keep on happening.  next it will be some other CPU.
three months ago it was the AML-8726-M: even they have "had their
moment".  six months before that it was the Telechips TCC8901.
remember - there are over a hundred ARM licensees, some of them will
succeed and when they do it will cause massive disruption each and
every time to the entire market, as every factory jumps on board.

 anyway, karen's got the source code, we're just working out how to
get it out - she isn't really familiar with software or tarballs, and
a 100mb archive is very challenging for her to send by systems that
she can understand and use [email].

 ... we'll get there :)

 l.