Re: [Arm-netbook] Single-Core Cortex A9 1ghz, ECC DDR3 RAM available soon
On Sat, Aug 13, 2011 at 5:58 PM, Gordan Bobic <firstname.lastname@example.org> wrote:
> On 08/13/2011 02:46 PM, Luke Kenneth Casson Leighton wrote:
>> ok, right. continuing on the discussion of upcoming and/or available
>> Cortex A9 systems, i heard back from one of the CPU manufacturers
>> (can't say which one), and they are sampling a new CPU next month.
>> rough spec:
>> * Single-Core Cortex A9, 1ghz
>> * MALI 400MP 3D
>> * SATA-II
>> would anybody else be interested to see a small panda-like board (4in
>> x 3in or so) or other type of board be brought into existence based
>> around this CPU?
> Depends on the cost. Freescale are pretty competitive in this arena with
>> if so, what would you be prepared to do to make that happen and,
>> also, what retail price would you be prepared to pay for it?
> Based on the competition maybe $170 for a 2GB one if it has ECC RAM.
ok - good to know.
>> bear in mind that the CPU's manufacturer has:
>> * a general understanding of and respect for free software licenses
>> including the GPL.
> What about GPU drivers supporting all the recent Xorg ABIs? I'm not an
> OSS faschist who demans everyting be OSS - I am prepred to overlook the
> OSS-ness of drivers and detailed documentedness of the GPU if the
> drivers are of decent quality, feature-complete and continuously supported.
it's MALI 400MP-based. the git repository with the linux kernel
already has the mali "shim" linux kernel driver added: that means that
the libGLES.so.2 proprietary "system library" should work, and the
xorg x11 driver just happily sit on top of that. i don't forsee any
problems, but at some point in the next few days/weeks i'm likely to
have to test this out (using a demo unit which has the precursor
processor in it). it's currently all running android so i'll have to
hack together a debian armhf for it. deep joy.
>> they also have a "system library" for the on-board Video CODEC DSP,
>> which those who understand the subtleties of the GPL will appreciate
>> qualifies for an exemption under the GPL, even though it's
>> proprietary. they've released an example application which utilises
>> libffmpeg (enhancements to use their "system library" have already
>> been made) and is a working video player.
>> for the freedombox to fulfil its sponsor commitments (bearing in mind,
>> as we know, the freedombox project is not a hardware project despite
>> the word "box" being in the name of the project but they still have to
>> deliver a number of actual physical boxes to the pledgeware sponsors),
>> i'd say that this CPU and the fact that its manufacturer respects the
>> GPL fits the requirements far better than the chosen marvell-based
>> plug computer.
> That would depend on the quality of the current drivers/libraries that
> aren't OSS and the ongoing support for them.
yeah - proprietary's a bitch, whichever way you look at it. you
wouldn't believe the fun and games i'm hearing about how these SoC
vendors are having a blast with adobe flash... they have to wait in
line for adobe to bother to compile up flash... with *their*
proprietary hardware video extension libraries! what a hoot.
anyway, leaving that aside: there are two areas of
stupidity^H^H^H^H^H^Hproprietary. a) hardware-accelerated CODEC
library b) MALI-400MP.
the first: well... you just have to trust them. and, once you've
seen a lowly single-core processor smoke a dual-core intel processor
at 1080p30 video playback, you've seen it all. it works, or it
doesn't. and, if it works, it will _stay_ working.
the second: you have to trust and put your faith in ARM (with their
stupid deal with MediaTek). Hallelujah, brothers: _every_ CPU
manufacturer using MALI is in the same boat. so that means Samsung
Enyxos, Telechips, AMLogic - everybody.
bottom line: it's not a nice situation, but tough titty, we have to
keep the pressure up.
>> perhaps the freedombox project might like, if nothing else, to use
>> this CPU as leverage to accelerate marvell out of their self-imposed
>> stupor by threatening to pull the plug (ha ha) if they don't get with
>> the C21st?
>> please bear in mind that there is a window of opportunity lasting a
>> couple of weeks in which i can potentially persuade the CPU designer
>> to come up with a demo / engineering board that would actually be a
>> useful saleable product in its own right...
>> ... and that i can *only* do that if there is a demonstrable need for such.
> The problem is that this board would be in direct competition with this:
no it's not. he says... *reviews spec*... yes it is! :) that has a
lovely SATA interface as well. and 10/100 ethernet.
and, freescale are actually on the ball these days in terms of free
software support etc.
[ um... freedombox people... tell me... why was the shitty little
NDA-loving armv5 marvell sheeva guwwu pwug chosen over the
freedom-respecting really rather dishy little IMX53QSB? i realise
it's good to break marvell of their paranoia habit but even so... ]
> It does have an advantage in terms of A9 vs A8 and ECC, though.
yes. even with the linaro improvements to gcc i'd say that the A9
still is a step up (superscalar out-of-order instruction execution).
>> so, if you would like an affordable 1ghz Cortex A9
>> panda/beagle/imx53qsb-esque/plug-computer board with 1 or even 2gb of
>> DDR3 ECC RAM, with SATA-II and 10/100 Ethernet as well as the usual
>> other interfaces including some GPIO pins, *and* where the usual
>> battle with the CPU manufacturer over GPL violations and NDAs has
>> completely gone, now is the time to speak up and say so.
> I would love to see this succeed - ARM boards with more RAM and more
> importantly ECC RAM is something we definitely need in this arena.
> Unfortunately, I just committed to getting a load of Freescale boards of
> similar spec so I won't be needing any more for a long time to come.
*shrug* ah well can't win 'em all. so, you _would_ be in... but you
personally don't need them :) and, by the time you do, things will
have moved on.
ok - anyone else?