That Open, Upgradeable ARM Dev Board Is Trying To Make A Comeback

[starshipeleven]

no, i'm chipping in here because i feel that what you describe duby229 is evidence of the "planned obsolescence by design" meme,

Look, another idiot that did not understand what iGPUs are about. They are not meant for gaming, and this is the only reason they don't become obsolete so fast as dedicated GPUs (around 2 years).

[duby229]

there was an article on slashdot yesterday about how hard it is to maintain an x86 games pc machine (the reviewer amusing concluded with "and that's why i buy apple!" o dear...) which you basically have to customise, research, and constantly upgrade to keep up with their "minimum recommended requirements" - it's a huge and expensive investment basically.

no, i'm chipping in here because i feel that what you describe duby229 is evidence of the "planned obsolescence by design" meme, which, although you probably *won't* find any evidence of direct collusion - not anything that would stick in a Monopolies investigation - but is more "implicity understood as being the end-goal by all major parties who could and have to profit from such implicit and tacitly-understood collusion".

what you're seeing is, i feel, this "strategy" coming unstuck, causing massive inconvenience to people as they're forced to upgrade, upgrade, upgrade, where they're just getting more and more absolutely sick of it.

so this is what i'm tackling. i can't say that the strategy i've come up with is perfect, but i am at least doing... *something*.

You make a really good point. I really hope DX12 and Vulkan, really any compute-like API, can solve that. Game developers should be the ones designing "features" so that hardware designers can concentrate on enough compute performance to accelerate it. I think Intel should especially like that concept.

[ecloud]

1366 x 768 will persist because the cost will come significantly down

Fine so you picked a cheapie for the first version because you think some of the proles (people who aren't in the target market yet anyway, but maybe eventually) could maybe get by. But designing the digital video output so that there is no possibility of upgrading makes this a no-go, I'm afraid. I hope you don't really mean that the VGA output on that nice wood case is also limited to something less than full HD. Or that we can't install better screens in customized laptops which run off this card, without resorting to using the HDMI output on the "wrong" end of the card.

for everyone else - those who are doing just some email, a bit of internet browsing, and editing word documents? no. 1366x768 is not only tolerable but is actually desirable in large screen resolutions, to give them the larger and clearer text

I'm so sick and tired of that particular trope. All those people need is a control panel that lets them scale the default text sizes simultaneously, with a slider or something, and that the apps would respect it. MacOS, MS Windows and Android all have such settings, although they're a little too discrete: small medium and large, basically. But if we're talking about Linux, you can easily have a slightly customized desktop with that control readily available, perhaps as a slider accessed from a tray icon or from a right-click/long-press menu on the desktop or whatever. I guess it will be mandatory if you really think people will be swapping this card back and forth between different housings and screen sizes all day long. (Although, the PCMCIA pins won't survive for many years if they do.) But developers have backwards habits like specifying fonts sometimes in pixels and sometimes in points, and then because of hidpi and all the surrounding disagreements, the perverse conclusion for Qt has been that neither unit can possibly mean directly what it used to anymore. Or thinking that fonts should have only discrete sizes in the first place, rather than being infinitely scalable, ultimately because hand-set type or Linotype molds were only available in discrete sizes, I suppose. We will get there eventually: the distance field rendering technique helps. (Only if you use the GPU though. This is another reason you need it.) App-specific scaling should have the control-mousewheel and control-plus-minus methods working universally, like they already do in the browser, in many editors, in Konsole, most graphics applications, etc.

Full HD screens will probably be around longer than anything else, because of the TV standard. You can play a movie fullscreen without downscaling, which ought to save some power somewhere. That's something the proles do a lot of.

[lkcl]

Look, another idiot that did not understand what iGPUs are about. They are not meant for gaming, and this is the only reason they don't become obsolete so fast as dedicated GPUs (around 2 years).

interesting. no, i have no idea what iGPUs are. i'll go look them up. [edit: oh: "Integrated" GPUs - sorry, new abbreviation. yes, familiar with those. where any memory access by the GPU stops the main CPU dead in its tracks. standard fare for embedded SoCs with a shared memory bus, as it's too expensive to put dual memory buses into such a small package given that it costs 150 pins to provide a 32-bit lane DDR3 interface and there's usually only between 300 and 500 pins on the whole SoC. what did you believe i don't understand?]

[lkcl]

Fine so you picked a cheapie for the first version because you think some of the proles (people who aren't in the target market yet anyway, but maybe eventually) could maybe get by. But designing the digital video output so that there is no possibility of upgrading makes this a no-go, I'm afraid. I hope you don't really mean that the VGA output on that nice wood case is also limited to something less than full HD. Or that we can't install better screens in customized laptops which run off this card, without resorting to using the HDMI output on the "wrong" end of the card.

it's also not about what the high-end community wants, it's about what we can *get hold of*. i give a glimpse of this in an update here: https://www.crowdsupply.com/eoma68/m...ng-a-processor

tried to get in contact with AMD: no response.

tried to get in contact with NVidia (via two separate routes): no response.

tried to get in touch with marvell (and to also find some form of low-power PCIe GPU, which doesn't exist): no response.

it's just completely insane. even if we got hold of these people the costs involved in creating a PCB based around their SoCs is like... $50k and above.

so we have to start somewhere, and i chose to walk the "path of least resistance" with the possibilty of upgrading later.

... how would it be done otherwise? would you prefer that i choose a goal that is financially near-impossible to achieve and extremely risky design-wise (so is near 100% likely to not even get off the ground let alone succeed), or one that i can achieve on an (appx) $USD 3k/month budget, set up a viable set of designs, bring them to people and *then* come back to more achievable targets later, once i'd got the experience and the larger budget to do it?

serious question: what plan could you come up with that will allow community-driven hardware to be created with the specifications that you seek?

[ecloud]

All I am saying is the A20 can do 1920x1080, so I don't see why it should be impossible for a "maker" to make a customized laptop with a full-HD LCD even though yours isn't. And it ought to be possible on the VGA output of the "desktop" enclosure too, or are you having trouble finding a converter chip which can do that? Or did I misunderstand, and the card is not really crippled to a 1366 x 768 limit? I got your earlier point that that's the minimum spec for any EOMA68 CPU card, because you want to support even lower-end SOCs.

[lkcl]

All I am saying is the A20 can do 1920x1080, so I don't see why it should be impossible for a "maker" to make a customized laptop with a full-HD LCD even though yours isn't.

yes - they could perfectly well do that.... they just wouldn't be permitted to sell it or promote it as "EOMA68 compatible", and i would require them to not use the word "EOMA68" in any online publications, otherwise they could cause huge confusion and thus bring the standard into disrepute. remember this isn't software, it's hardware: it would be like someone trying to create cheapo nasty USB3 cables that aren't properly compatible with the USB3 power budget, or AMD trying to pull too much power from a PCIe slot. it's *dangerous*.

And it ought to be possible on the VGA output of the "desktop" enclosure too, or are you having trouble finding a converter chip which can do that?

the $1 worth of discrete components used on the micro-desktop are rated at around 100mhz: 1024x768 is arouuund... 72mhz clock rate? and that's ok. 1366x768 is just about ok for these low-cost buffer ICs. yes, finding a converter chip that's (a) not under NDA (b) findable at all (c) has a public datasheet (d) isn't completely insane cost i.e. would double the cost of the micro-desktop board.... you get the general idea. yes we found an analog devices triple-output D2A converter IC... it was $12 in *volume*. the *entire BOM* is $12. an IC from china didn't have a datasheet. no response from the company. yes you can find the ICs on taobao but it's no damn good if you can't tell what pins are which.

etc. etc. and that's just *one chip*. now go through that process for a hundred different ICs and you start to appreciate why i've gone for the approach that i have.

Or did I misunderstand, and the card is not really crippled to a 1366 x 768 limit? I got your earlier point that that's the minimum spec for any EOMA68 CPU card, because you want to support even lower-end SOCs.

... and that 1920x1080 is supported for the 3.3mm variant of EOMA68, which will require large up-front investment to redo the PCBs to 0.8mm height and to have custom-made tooling.... yes.

[robclark]

Just in case people didn't know, HTC phones can have the bootloader unlocked officially.

http://www.htcdev.com/bootloader

It's the reason I like HTC so much.

sony/xperia too.. they also have contributed a lot of bits of snapdragon support to upstream kernel. If only more handset vendors did the same..

[robclark]

it's also not about what the high-end community wants, it's about what we can *get hold of*. i give a glimpse of this in an update here: https://www.crowdsupply.com/eoma68/m...ng-a-processor

jfyi, for sd 410, I don't think it is upstream yet, but there is a gpl kernel v4l driver for video dec/enc (well, not 100% sure if the enc side of things is supported yet, but afaiu that is the plan).. also, for the 'tech support' column, I guess you could call ##linux-msm and #96boards on freenode as "Community-based" support.. (you could probably also get support via arrow/etc.. but at least on the sw / upstream-kernel side of things you're probably better off asking on IRC ;-))

[lkcl]

jfyi, for sd 410, I don't think it is upstream yet, but there is a gpl kernel v4l driver for video dec/enc (well, not 100% sure if the enc side of things is supported yet, but afaiu that is the plan).. also, for the 'tech support' column, I guess you could call ##linux-msm and #96boards on freenode as "Community-based" support.. (you could probably also get support via arrow/etc.. but at least on the sw / upstream-kernel side of things you're probably better off asking on IRC ;-))

nniiiice. good leads, thanks rob.