[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Anyone here made a "TV computer"?

On Sat, Nov 5, 2011 at 12:12 AM, David Given <dg@cowlark.com> wrote:
> On 04/11/11 17:50, Luke Kenneth Casson Leighton wrote:
>> On Fri, Nov 4, 2011 at 1:32 PM, David Given <dg@cowlark.com> wrote:
> [...]
>>> Well, I know why *I* would get one: it's cheap.
>>  would you be interested in anything that has a higher bang-per-buck
>> ratio?  or just cheaper?  (or both)
> Absolutely! Note that I was *criticising*, merely pointing out different
> value frameworks.

 oo sorry if i gave the wrong impression (happens a lot with me, i
don't understand why, but just have to live with that, move on... beh)

 no.  yes.  great!

>>  details on the allwinner A10 are emerging, but they include:
> [...]
>>  so, what would you be prepared to pay for something with the above
>> specs, especially given that it will have the GPL source code
>> available in advance, and you'd have an indirect line to the
>> manufacturer and the factory to get relevant questions answered if you
>> wanted "in" at the very early stages.
> Hmm. $0-$50, I'd probably fling cash at you.

 :)  that might actually happen, but not for the very first run.

on the basis of the stunningly-low CPU price (in large volume) the
factory are just going ahead with the prototyping regardless: once
they have the components agreed they'll be able to give me a rough

> $50-$100, out of my
> spending bracket for toys so would need to consider my precise
> requirements.

 i intend to make sure that, based on the feedback i've received, the
PCB covers as much as i've received input on, as well as making some

> >$100, I'd still be interested but would probably wait for
> the next iteration to make sure that no nasty first gen faults appeared.

 yeah sensible.  if you've been following what dr schaller is doing
for the GTA04 we're basically following the stages that they're doing,
except of course factoring in the construction being in Shenzen, fuun.

 1) very first prototypes.  very limited run.  get it in under "Green
Carnet", send it back with working source code.
 2) early prototypes, out to free software developers.

 there's lots of other steps

> All subject to shipping, of course.


> My particular use case is that I run cowlark.com off a single
> SheevaPlug, which is doing a sterling job but suffers badly from too
> much I/O going down its single USB2 port, and isn't a particularly meaty
> processor to begin with. I'd very much like to upgrade to something
> with, like, an FPU.

 ok, yeah i think this CPU will do the job for you, with room to spare.

>>  *sigh* it's a matter of power.  the architecture of separate ICs
>> (usually northbridge/southbridge) simply doesn't work for this level
>> of ultra-low-power.
> I was actually rather thinking of integrated GPUs; I'm aware the
> external ones suck, being mostly designed for the PC market where you're
> only cool if your PC is doubling as a space heater.

 apologies, i jumped the gun on the direction of the conversation.

> For my day job I do a lot of work with the more exotic Android devices
> (including MIPS ones). The one thing these all have in common is the
> total lack of quality of the vendor-provided OpenGL stacks. Prior to
> Android Gingerbread, for example, it was trivial to crash the entire
> phone by e.g. for one chipset, calling glReadPixels() on a 32-bit EGL
> config. (Gingerbread's much more robust. I've had a few strange reboots
> but nothing I could replicate.)


> Go back a few years and ask game developers about OpenGL stacks on Brew
> and Symbian and things are even worse; you'll probably end up with them
> curled up in a corner.


> So (even though I don't have an immediate requirement for one), the idea
> of a GPU, integrated or otherwise, which has a properly open source GL
> stack where these bugs can be *found* and *fixed* is very attractive.
> Plus, of course, if there's a programmable shader pipeline, having a
> documented ISA for the GPU processor makes it easier to abuse.

 yeah, i did set up a reverse-engineering proposal both on CELF and
with the FSF, nothing's come of it - it was for PowerVR SGX.  where
the heck are they, now?? since ARM bought it from MediaTek (or
licensed it, more like), everyone's gone with MALI.

 is CELF back up - it was hosted on kernel.org (whoops) which as you
know got hacked a while back.. oops :)


Reply to: