Rhombus Tech 2nd Revision A10 EOMA68 Card Working Samples

lkcl writes "Rhombus Tech and QiMod have working samples of the first EOMA-68 CPU Card, featuring 1GByte of RAM, an A10 processor and stand-alone (USB-OTG-powered with HDMI output) operation. Upgrades will include the new Dual-Core ARM Cortex A7, the pin-compatible A20. This is the first CPU Card in the EOMA-68 range: there are others in the pipeline (A31, iMX6, jz4760 and a recent discovery of the Realtek RTD1186 is also being investigated). The first product in the EOMA-68 family, also nearing a critical phase in its development, will be the KDE Flying Squirrel, a 7-in, user-upgradeable tablet featuring the KDE Plasma Active operating system. Laptops, desktops, game consoles, user-upgradeable LCD monitors and other products are to follow. And every CPU that goes into the products will be pre-vetted for full GPL compliance, with software releases even before the product goes out the door. That's what we've promised to do: to provide Free Software developers with the opportunity to be involved with mass-volume product development every step of the way. We're also on the look-out for an FSF-Endorseable processor which also meets mass-volume criteria, which is proving... challenging."

Comments

Definitions, please?

[gman003]

It's funny how, on articles about things everyone here knows about, like BitCoin or the Raspberry Pi, the summary wastes space explaining the context (ie. what BitCoin or RaspPi is), but on an article about something relatively obscure, it just throws model numbers and acronyms at you.

As far as I can discern without reading TFA, this is just some new ARM system-on-a-chip, not particularly revolutionary or powerful, but aimed at use in open-source environments.

Re:Definitions, please?

[Anonymous Coward]

Yeah, the A10 is a chip that's popular with mid-range unbranded chinese tablet PC manufacturers. It's a slightly better chip than the more common A13 from the same manufacturer (which is what most of the current crop of dirt cheap tablets are based on). It's usually advertised as being clockable at up to 1.5GHz (although 1.2 seems to be the fastest they reliably work at), and is a single core ARM chip with fully open source drivers for all of the integrated hardware, which is somewhat unusual. I have an Olimex OLinuxino A13, which is a somewhat similar board based on the cheaper A13. They're a little more expensive than a Raspberry Pi, but substantially more capable in many ways.

Re:Definitions, please?

[citizenr]

Yeah, the A10 is a chip that's popular with mid-range unbranded chinese tablet PC manufacturers. It's a slightly better chip than the more common A13 from the same manufacturer (which is what most of the current crop of dirt cheap tablets are based on). It's usually advertised as being clockable at up to 1.5GHz (although 1.2 seems to be the fastest they reliably work at), and is a single core ARM chip with fully open source drivers for all of the integrated hardware

Lie. No GPU, no Video Decoder, no Video Encoder.
Its as "open source" as Broadcom sitting in RPi.

Re:Definitions, please?

[AdamHaun]

As far as I can discern without reading TFA, this is just some new ARM system-on-a-chip

No, it's much sillier than that. This is the latest in a long-running series of Slashvertisements by the submitter, lkcl. They chronicle his journey towards creating an "industry standard" for swappable processors for tablets based on the PCMCIA form factor. Nobody asked for this, nobody wants it, and lkcl has next to no experience with hardware development, but he's convinced it's going to change the world! To help the world along, he's working on-- actually, it looks like various Chinese companies are doing all the work. Anyway, lkcl is the funding conduit for an example card based on an existing ARM SoC. Today's story is about getting the first samples of the "2nd revision" of this card. Future samples are approved for sale as a standalone product because "they boot", which obviously qualifies them to ship.

In our last episode, lkcl digressed from his main project to announce a funding drive for a totally unrealistic project to build a free software-friendly SoC with a custom CPU in six months without doing any "design" work. Except for speeding up the processor, adding a bunch of peripherals, and implementing it on a cutting-edge semiconductor process. And then getting to market by Christmas. Just a small side project, right?

lkcl is pretty prolific on his own stories, so I'm sure his dozens of comment responses will answer all of your questions.

Previous episodes:
Live Interview: Luke Leighton of Rhombus Tech Dec 11, 2012: Live interview that nobody saw. There doesn't seem to be a transcript.

Rhombus Tech A10 EOMA-68 CPU Card Schematics Completed Sept 7, 2012: PCB schematics (for the first revision -- prototype?) completed.

PCMCIA Computer Project Aims Even Higher (and Cheaper) Than Raspberry Pi Dec 17, 2011: Project announced? This is as far back as the Rhombus Tech news page goes.

Re:Definitions, please?

[EmperorArthur]

It's hidden behind the fourth link. http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68

It's a computer in the form factor of a PCMCIA card. The whole goal of the project is to have upgradable tablets and laptops. You just pop the old card out, and put a new one in.

It's a laudable goal, but more hardware manufacturers would need to support it. Honestly, at some point they need to come out with a high performance x86 one.

Re:Definitions, please?

[asola]

The big news here is that the first EOMA-68 formatted computing cards and accepting cases are nearing completion. This is an open-source project with a strong emphasis on embedded Linux.

This will make it possible to upgrade the computing core of your tablet or set-top-box or all-in-one computers, smart-tv when higher-performance computing cores become available.

This makes sense since, for example, there are a lot of parts on modern tablets that need no upgrade at all (e.g. an IPS retina display is already at the edge of the capabilities of the human eye).

I, for one, would be very happy if my NAS was upgradable like this.

Re:Definitions, please?

[dublin]

Not only that, it's been done before: One of the Japanese companies (Sharp or maybe more likely Epson?) tried to push a PCMCIA-based CPU module back in the early '90s. It was a decent idea then, but the form factor is not as reasonable a choice today, given the increased availability of smaller-scale SMD components....

Re:Definitions, please?

[steelfood]

That's because the editors only include explanations for things they know about. What, were you expecting them to read TFA before front-paging it?

Re:Definitions, please?

[K. S. Kyosuke]

but he's convinced it's going to change the world

So are the Raspberry Pi people. So what?

Re:Definitions, please?

[lkcl]

It's funny how, on articles about things everyone here knows about, like BitCoin or the Raspberry Pi, the summary wastes space explaining the context (ie. what BitCoin or RaspPi is), but on an article about something relatively obscure, it just throws model numbers and acronyms at you.

As far as I can discern without reading TFA, this is just some new ARM system-on-a-chip, not particularly revolutionary or powerful, but aimed at use in open-source environments.

absolutely! i'm a link whore, what can i say :)

ok - it's not limited to one particular ARM SoC, it's about a standard. actually, it's not about the standard. the actual goal is to bring about a revolution in the mass-volume computing appliances sector. i don't mention that too often because it sounds absolutely madly insanely ambitious, but it's what's actually going on here.

so let's look a bit further down the chain, at say rhombus tech. rhombus tech is the web site designed to help link free software developers with chinese factories. if you're familiar with the rampant level of GPL violations (google "mjg59 android gpl violations") in the chinese hardware sector you'll know how limiting that is. i have two friends - both engineers - who are VERY frustrated at being able to buy such incredibly cheap hardware, yet they CANNOT RE-PROGRAM it because not even the factories have the GPL source code! so the promise of cheap hardware is there, but the reality is that you can't use it.

this insane situation is what rhombus tech is setting out to break, by ONLY working with factories and SoC vendors that are willing to be GPL compliant right from the start. my role here is therefore to pre-vet the SoCs, get the documentation and source code sufficient to at least get something up and running, find a substitute low-cost affordable development system (e.g. the Mele A1000 and now the Mele A100G Quad in the case of the Allwinner SoCs), then get it into modular form and get it out there.

the modular form-factor - which includes the development of standards such as EOMA68 - is the real kicker though. i don't know if you're aware of the situation with regard to development of tablets and other hardware but it's insane. due to the high level of integration on ARM and MIPS SoCs, EACH AND EVERY PRODUCT REQUIRES TOTAL CUSTOMISATION. there is no BIOS: there is only u-boot (which must be customised) and the linux kernel (which must be customised). the turn-around time on products is such that in some cases the introduction of a new SoC can cause a MAJOR recession in the electronics industry in Shenzen. it happened before when the Allwinner A10 came out, because its price was (is) only around $7.50 when all of its competition was around $13, and it'll happen again.

so we've come up with EOMA-68 as a way to mitigate against this (and other) problems. example: i received the PCB PADS files from Allwinner for the A31 SoC. i got a basic EOMA-68 layout done in FIVE DAYS because all i had to do was throw together a set of libraries and circuits that i already had from the other 4 boards that i've done. so with this approach we could basically upgrade an ENTIRE product range - tablets, laptops, desktop systems, anything-that-takes-a-CPU-card - in about 3 weeks flat. that's just... unheard-of! imagine going to a factory and saying "we can upgrade all your product in 3 weeks! all your base are belong to us!" they would be telling that story for years down at the local pubs, about the day some crazy guy came in and said they could upgrade their entire product range in 3 weeks... ... but that's literally what can be done here. do you see how powerful that is?

then there is the fact that the products, because they don't fundamentally change (it's only the CPU that gets faster), the factory can LITERALLY make them until the cows come home. or, more to the point, until a component goes end-of-life. but if they're making a 7in tablet chassis by the bucket-load,

Re:Definitions, please?

[lkcl]

Not only that, it's been done before: One of the Japanese companies (Sharp or maybe more likely Epson?) tried to push a PCMCIA-based CPU module back in the early '90s.

the difference here is that this is re-use. it's *NOT* backwards-compatible with PCMCIA. the idea of having a computer-that-can-dock-with-a-computer is great, but nowhere near as revolutionary. i did quite a long post to one of the other questions on here, which explains a bit more of the background: http://hardware.slashdot.org/comments.pl?sid=3643131&cid=43435507

but yes, you're right: the EOMA-68 form-factor, which is credit-card-sized and 5mm in height, is a bit too big to put into devices such as smartphones, some of which are now only 6mm thick themselves! that's why we also created EOMA-CF which, surpriiise, re-uses Compact Flash. however that's *really* small, and will need us to finance the tooling as well as get access to SoCs that have Package-on-Package RAM and so on, so we made a conscious decision to focus on EOMA-68 first.

and that's fine, because EOMA-68 covers a *huge* range of products. we have a guy in Spain who's designing a hand-held games console. the KDE Team is sponsoring the development of a 7in tablet with a 1024x600 IPS screen (actually the same panel from LG that's in the Kindle Fire). once we've got actual demo products we'll go to netbooks, laptops, LCD TVs and Desktop computers next. my favourite product i'm really looking forward to is a Digital SLR camera. no, really! a camera with decent lenses with a user-replaceable CPU Card, how cool would that be? rather than swap the memory card out, you'd actually swap the *entire processor* :) and upgrade it later to a faster version. or put in a CPU Card with a built-in 3G Modem, so you can upload pictures automatically and in real-time. journalists (professional and amateur) would love that.

so. yeah. all good stuff.

Re:Definitions, please?

[lkcl]

I, for one, would be very happy if my NAS was upgradable like this.

yeah me too! NAS boxes is on the list of products, as well as routers. as an open hardware project where the CPU Card is guaranteed to have full GPL Source Code, you'd be able to do anything - re-use the older cards as a router, and cycle them down over the years, between products. last year's "latest tablet CPU Card" becomes "the kid's games console CPU Card" becomes "the NAS / router" for your home or the "Freedom Box" for your grandma.

Re:Definitions, please?

[Archangel Michael]

All of that is fine, until you can't upgrade the bus, and that becomes the limiting factor. And, when you upgrade the bus, you usually cannot use existing cards (easily). Upgradable items are usually good for two, perhaps three generations before the rest of the device is obsolete. Which may be a fine goal, but if the cost of the device doubles between upgradability and the upgrades themselves, it becomes a wash at two upgrade cycles and only profitable at three. That is a risk, and one I've seen burn people when they are caught buying an upgradable item that has no upgrades made for it.

Re:Definitions, please?

[lkcl]

All of that is fine, until you can't upgrade the bus, and that becomes the limiting factor. And, when you upgrade the bus, you usually cannot use existing cards (easily). Upgradable items are usually good for two, perhaps three generations before the rest of the device is obsolete. Which may be a fine goal, but if the cost of the device doubles between upgradability and the upgrades themselves, it becomes a wash at two upgrade cycles and only profitable at three. That is a risk, and one I've seen burn people when they are caught buying an upgradable item that has no upgrades made for it.

let me answer the profitability issue first. we chose to re-use legacy housings, sockets and assemblies precisely because to do otherwise *would* result in this becoming a profitable venture only at cycle 3. there's a company in the U.S.+Taiwan which has had $USD 100m investment to create a 100mm x 70mm x 10mm modular PC standard. we've had *zero* investment.... and haven't needed it! the CPU Card development cost us under $10k. the tablet: $6k. getting new plastic done for the card because we're re-using PCMCIA metal casework from a product that's been made for the past 10 years straight: $6k.

so you're thinking inside-the-box, i feel compelled to point out :) we'll go into "profitability" with the first 10k order!! everyone involved has been working on a commission-only basis for the past 4 years on the project. there *are* no investors or banks to pay off. the first lot of profits will go straight back into the project and will begin to fund and reward the free software developers and other people who have been helping us out over the years, and that will happen pretty much immediately.

regarding the upgradeability and the durability of the standard: there's one factor that you've not taken into consideration, and it's the power requirements of faster interfaces. 10GbE over copper takes SIX WATTS, just to push the signals over those 4 twisted-pairs that's just... insane. as people have wanted faster and higher resolution screens, VGA has fallen by the wayside because at 75 ohms impedance, driving 3 lines at 200mhz and above in *analog* is just way waaay too power-hungry.

but look closely at the interfaces selected for EOMA-68. RGB/TTL (24-pin), I2C, USB3, Gigabit Ethernet and SATA. are any of those particularly critical that they be ultra-ultra-ultra-ultra-ultra fast? no not really. what are they connected to? well, they're connected to peripherals i.e. I/O. do you really really really really need an 8096x5000 resolution LCD panel on a 7in tablet? no not really. do you need a 10000Mbytes/second SATA hard drive on a 10in $150 laptop? no, not really. do you need 10 Gigabit Ethernet on a portable device where battery life is important? no, you don't.

so you're thinking of upgradeability as being all-important and the be-all and end-all of computing appliances, and i think you'll find that it really, really isn't that critical. at the apple end of the market? sure, there will be people who will always go after apple products, and the great thing is: just like microsoft's absolute-insane-latest-and-greatest processing and memory requirements have pushed the price of RAM down to $4 for 1GByte of 800mhz DDR3 RAM, so will apple's R&D costs *also* drive down the cost of parts for the rest of us who are happy to sell in much higher volume, quietly, to the rest of the world market including China which is 10x the size of the rest of the world's markets PUT TOGETHER and nobody knows it even exists.

summary: the strategy we've pursued immediately pays off, and the EOMA-68 standard's designed around a different market focus which i believe is sound for at least the next decade. we could always develop new standards that take advantage of the latest-and-greatest innovations, but they would be limited to the latest-and-greatest products. we're going after the bigger volumes - the cash cow markets - and helping the Factories to stabilise their products, take advantage of the latest-and-greatest as it filters down.

does that make sense?

Re:Definitions, please?

[lkcl]

http://lkcl.net/articles/tiny.computers.txt

the above article may help also to give you some background about where this came from and where it's going.

Re:Definitions, please?

[AdamHaun]

The Raspberry Pi people made a toy for educational purposes. If you read the About page, you will see two key differences:

1. The Raspberry Pi people have extensive hardware experience, with key figures actually working in the industry. They do their own technical work. They can answer questions about their design. They are selling through reputable distributors. lkcl, on the other hand, is a front-man with no hardware background. He is apparently the sole advocate in the world for a China-based for-profit operation.

2. Raspberry Pi's ultimate goal is to make a small single-board computer to help kids learn programming. lkcl's goals are large and varied, but mostly involve upending industries he knows nothing about, and probably getting rich in the process.

Re:Definitions, please?

[lkcl]

not yet!!! we have a plan, here, if it all goes to hell in a hand-basket: we use Cardbus (the gold-covered Type II). the gold-coloured shielding is earthed in 8 places directly to the PCB.

if you look closely at the PCB layouts i've done, you'll see e.g. on the A31 PCB that the SATA (1.5gbit/sec) and Ethernet (100mbit/sec spread-spectrum) are within 10mm of the connector. i believe the length of the SATA and Ethernet tracks are about 6 to 8mm in length.

we'll find out, eh? :) it's all so exciting, not having access to $250m of funding and having to wing it!

Re:Definitions, please?

[pjr.cc]

Say what you guys will, personally i've been following EOMA-68 for a while and for a good reason. Ultimately a long time ago, i've been wanting to do what calxeda now do (around the time the first nslu2's came out) and was really excited when calxeda started talking the talk... then they released a box that is way over engineered and way too expensive for the role (not to mention rather inflexible).

With eoma-68 you could do the same thing, only the price would be far more suitable and the server boards inside the chassis "standardised" on a simple interface anyone can implement. The backplane for putting 100 of these boxes into a 2ru chassis is quite simple also. Think: a replacement for x86 based virtualisation and vps's

So yes, you may look at it and see a slashvirt if you will, however there are those of us in tech that see a real purpose to it also..

Re:Definitions, please?

[gman003]

Ah, I remember them now. "Mini ARM computer reusing PCMCIA connector" would have sufficed to describe it.

Re:Definitions, please?

[AdamHaun]

got a problem with that? then FUCK OFF and stay out of my way. you do your thing, with your thoughts; i'll do mine.

Excuse me? You're the one who keeps bringing your pointless stories about meaningless "progress" on your pie in the sky project to Slashdot. Last time you were asking for ten million dollars for a hopeless SoC project that you yourself knew nothing about. Now you're back here, what, trolling for pre-orders and funding? What kind of con artist are you?

And let's be clear, this isn't even really *your* project. You're throwing money at companies in China to get them to do the work for you, and you can't even answer their questions. You're nothing but a sales guy with a vague idea and a big ego. Come back when you're selling in volume, until then please quit wasting our time.

Re:Definitions, please?

[jockm]

There are mounting holes on the Pi. Sadly there are only two, so you can't make it stand freely off of standoffs, but it has them

Re:Definitions, please?

[AdamHaun]

you do your thing, and i'll do mine ok? have some respect for people's desire to keep going and to encourage others to succeed.

I do not complain about other people's personal projects. But when you start asking for money, there's some responsibility that goes along with that. Part of it is being honest with others and yourself about who you are and how you operate. Part of it is responding to criticism. You can't expect the benefit of the doubt when you're selling something, especially when you barely understand your own project. You can't expect us to be trusting when we find out that nobody is working with you except the outside companies you're paying to do the design and prototyping. If your response to criticism of your credentials and your business model is to act persecuted and accuse others of bullying, how are you going to make it in the business world?

You want to make a funky SBC? Great! You had some working boards fabbed? Congrats! I am sincerely happy for you. You want to share the joy of your project with others? Go for it! (Maybe bring some technical people along, though.)

You want other people's money? Come back when you're actually shipping product.

Re:Definitions, please?

[lkcl]

PADS 9.3. it's absolutely awesome. i'd thoroughly, thoroughly recommend it. it's intuitive, it's obvious, the menus are simple yet powerful, and a heck of a lot of effort and thought has gone into the design and useability, to make sure that the context menus adjust to provide you what you *need*, at the time that you need it.

by contrast, if you've ever seen Allegro PCB design software, it's a nightmare. the menu bar has 25 options across the top!! that's just absolutely insane, and you can tell that the software team basically haven't thought about uesability - at all. you're expected to just... "know" what menu option is needed, you're expected to "know" what "mode" you're in - i can't even BEGIN to get started.

i started using PADS, and i didn't even need a tutorial in order to start doing something. sure, i made mistakes, and there were a couple of frustrating moments when i thought "ok, this isn't obvious, let's look it up" such as "what the hell is ECO mode" and it's a button that stops you from accidentally modifying the PCB from becoming out-of-sync with the schematics. if you click that button, then the software will save any differences that you make [to the netlist] from that point on, so that you can "back-import" them into the schematics.

so i was basically up-and-running in about a month. every time i look at Allegro, i just... i can't even begin to get started. big big difference.

and yeees, i really want to use KiCAD, but it simply cannot cope with these types of tasks *plus* my ignorance of PCB layout :) if i was a trained engineer with 20 years of experience in PCB design i *might* be able to use KiCAD for these tasks, but it would still be very very frustrating even with 20 years experience because KiCAD doesn't have anything like a built-in autorouter, or the Design Rules checks, or differential-pair routing or in fact anything that you'd expect to have in a professional-grade PCB design package.

Re:Definitions, please?

[lkcl]

Ah, I remember them now. "Mini ARM computer reusing PCMCIA connector" would have sufficed to describe it.

:) Mini ARM computer, Mini x86 computer (when we get access to ValleyView), Mini MIPS computer (Ingenic jz series), Mini {insert CPU model here} computer, mini FPGA card, mini pass-through card.

the EOMA-68 standard is *not* limited to a particular CPU - it's not even in fact limited to a CPU *at all*. take a look at this for example:
http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68/Passthrough

that's what we call a "pass-through" card. it has HDMI/DVI **INPUT**. not HDMI output from a processor. it has HDMI *IN*. that input gets converted to RGB/TTL and is "passed through" to the EOMA-68 connector.

what's the purpose of that?

well, imagine that you buy an EOMA-68-compliant LCD Monitor. it comes with a "pass through" card. it costs the same as a standard LCD monitor. it has an HDMI input. except this monitor, you can press a button on the side, pop out the pass-through card, and insert an EOMA-68 Computer Card.

voila - the monitor has instantly been transformed into an all-in-one computer!!!

how absolutely cool is that?

you could turn it into a TV by popping out the Computer Card and putting in a TV card.

you could take that same TV Card and pop it into your 7in tablet "chassis" and you have a portable TV!

are you starting to appreciate just quite how powerful this concept really is?

Re:Definitions, please?

[gman003]

As others have said, this seems very pie-in-the-sky right now. Have you yet produced *any* EOMA-68 card? As in, actually-shipping?

Sure, if you manage to pull it off, it would be great. None of the ideas, themselves, are that bad. The problem is that you're doing far more at once than I would consider possible. When you were just focusing on the mini ARM PC, it was considered risky enough. Adding all these other ideas seems... unwise. Sure, plan for them, don't do anything as part of the mini-ARM-PC that would break these ideas, but don't waste engineering or PR time on them until you've actually gotten *something* out the door.

Some of your ideas, like the HDMI passthrough, clearly require you to have significant market share to succeed. Monitors are already operating on pretty thin margins - adding the cost of a removable card, while possibly only adding a few dollars to the unit, still costs them quite a bit in the long run. Someone might do that if they're also selling the PC cards, but they're more likely to develop them as separate products (how many people, exactly, will turn a monitor into a full PC?), which eliminates the need for a removable card.

Re:Definitions, please?

[rephlex]

Future samples are approved for sale as a standalone product because "they boot", which obviously qualifies them to ship.

Isn't that the same criteria the Raspberry Pi Foundation used?

Re:Definitions, please?

[dublin]

My point was that this exact thing has been tried before, and the market clearly decided it wasn't worth the trouble and expense. This is no different.

Rhombus Tech is building EOMA-68 as a duplicate answer to a question no one has asked in years. By any reasonable standards of modern embedded hardware (I've been working with deeply embedded hardware for my own company for the past few years), EOMA-68 is poorly designed, not tightly integrated, and does not even begin to match the capabilities of many of the embedded modules that are already available at good prices from a wide variety of manufacturers. RhombusTech seems only interested in trying to create and push a "proprietary open source" form factor just to be different. BTW, the Casio/Epson/whoever PCMCIA CPU cards also redefined pinouts, too, but at least they had the good engineering sense to use different specific keying (and to get the PCMCIA consortium to register it as such) to avoid plugging them in where thy could not work, or where they could damage or be damaged by other PCMCIA equipment. Relatively cheap and available (if huge) connectors are about the only thing PCMCIA/CF has going for it in today's world: there's a lot more working against it, starting with lots of wasted packaging space (it was great 25 years ago!), compatibility issues, heat dissipation difficulties, I/O limits, etc.

that's why we also created EOMA-CF which, surpriiise, re-uses Compact Flash. however that's *really* small
No, sorry, wrong. Bzzzt. Thank you very much for playing.

As modern embedded electronics go, CF is simply enormous, rather than absolutely gigantic, as PCMCIA is. Have you actually taken a look inside a modern phone to see the scale of tight integration in these devices? Sure, you can chain yourself to PCMCIA/CF, but you can't do it without also chaining yourself to an anchor wedged firmly in the mid 1990's.

Like the earlier attempts, this attempt to dictate a "standard, open" CPU card form factor will fail - even with "Sparkly Magic Open Source GPL Sprinkles! (TM)" The market moves far too quickly for this - look how hard it is for AMD and Intel to even stick with their own CPU pinout and bus standards for more than a few years! By your logic, I should be able to buy a Core i7 plugin for my old Socket II Pentium computer, and somehow, magically, expect it to deliver all the goodness of a Microsoft Surface Pro despite the fact that everything that surrounds that processor (for input, interaction, communications, interfaces, etc.) has changed even more than the processor itself. Really, what you are proposing (and building!) makes that little sense!

That said, I think standard CPU form factors (even as pluggable modules) are a good thing, but the market has shown that only the most open and flexible ones have any chance at adoption. VIA is arguably the most successful here, and look a their success rate: MiniITX = Hit, NanoITX = meh, PicoITX = meh++, other attempts = Can Anyone Even Remember What They're Called?

Even the well-engineered schemes (RhombusTech's engineering is joke like, at best) that have significant industry backing (I'm thinking Qseven, here) have a hard time getting the required traction, and the pace of innovation and technology advancement makes it really hard to have a standard that's still relevant by the time it's developed and in production. Likewise, something will someday replace the hoary PC-104, but with modifications, it still clings to life as a contender. (And you could easily argue that the Arduino shield implementation is a non-standard "standard" of sorts, although it is not well-engineered, either...) Saying you have a fallback to Cardbus ("we have a plan, here, if it all goes to hell in a hand-basket: we use Cardbus", from your own comment below), is simply admission that you have no bloody idea how to do this right in the first place. CardBus is only slightly less antiquated than PCMCIA, and was itself sup

Does Realtek RTD1186 have a FPU

[jonsmirl]

Not clear if Realtek RTD1186 has a FPU. It is possible to have those graphics co-processors and not have a FPU on the core. Missing FPU messes up games. This chip may be purpose built to implement something like a Roku or that Mele STB which makes it uninteresting except for those specific purposes.

Re:Does Realtek RTD1186 have a FPU

[archshade]

Maybe I'm missing something but why are you focusing specifically on that chip? Some of the chips (e.g. A20) definitely have an FPU unit. If you need an FPU (and who doesn't), get a card (or cards) with an appropriately specified CPU core.

Re:Does Realtek RTD1186 have a FPU

[jonsmirl]

The A10 chips definitely have a FPU, I have one on my desk.
Unclear if the Realtek RTD1186 has an FPU. Many MIPS cores don't have one.

A pipe dream...

[n7ytd]

And every CPU that goes into the products will be pre-vetted for full GPL compliance, with software releases even before the product goes out the door. That's what we've promised to do: to provide Free Software developers with the opportunity to be involved with mass-volume product development every step of the way.

If "full GPL compliance" is a goal of the project, then it's doomed to mediocrity. Real chip vendors are not going to share their secret sauce, either because they can't due to patent/IP agreements or because they don't see a reason to risk handing the crown jewels to their competition. It just ain't gonna happen.

Re:A pipe dream...

[lkcl]

If "full GPL compliance" is a goal of the project, then it's doomed to mediocrity. Real chip vendors are not going to share their secret sauce, either because they can't due to patent/IP agreements or because they don't see a reason to risk handing the crown jewels to their competition. It just ain't gonna happen.

then we will not talk to them. they can fuck right off. we only need one or two companies to cooperate: that's the beauty of it. we don't need *every* chip vendor to cooperate with us, we just need *one* chip vendor to cooperate with us. when the other companies see just how much volume we're shipping through our clients they'll want a slice of the action, and we will remind them that we will NOT expose our distributors to massive liability of primary and secondary Copyright Infringment Lawsuits.

i'm staggered beyond belief that huge companies like Amazon aren't aware of the fact that they're risking being sued to the bedrock with a secondary Copyright Infringment Lawsuit. they should be banning these GPL violating products *outright*! but they're being hoodwinked... and unfortunately for them, in the eyes of the law, that's no excuse.

we *are* aware of the GPL, and the implications of Copyright Infringment, so we simply cannot and will not expose the distributors to that liability - end of story.

basically, your comments fail to recognise that the SoC vendors who "want to keep things secret" are in most cases now operating illegally, due to their criminal infringment of Copyright Law. many of them, like AMLogic, have *already* lost their rights to distribute the Linux Kernel Source code due to their GPLv2 violations of two years ago. for a SoC vendor to do that is COMPLETELY insane!! especially given that AMLogic is now owned by a USA-based company.

but in the cases where these SoC vendors *are* operating within the law yet are keeping things proprietary (through the "System Library" GPL exemption clause), there what we will do is put some funds towards reverse-engineering their hardware. ironic that we will use the money gained from the sale of their own products to do that, but it is, long-term in their own interests.

i don't know if you're aware of this, but in the case of 3D GPUs, the actual 3D GPU vendors *want* the free software community to reverse-engineer their hardware! the reason is this: the sole reason why they cannot publish information about their own GPUs is because of the risk of a patent war. i don't know if you've seen that talk given 6 months ago about this, but the situation between NVidia, ATI and so on, because they are mature products, they've come to an uneasy truce on their various patent portfolios. the so-called "embedded" GPU companies, they're new at this, and they are nervous as hell. *but*, they know the advantages that free software brings! google the story about the Intel GPU team getting together with the Valve/Steam developers: one of them said "it was the most productive work meeting they had EVER had", and it's because *BOTH* teams could read each others' source code... without having to go to their respective Directors and get NDA clearance, which would apart from anything have taken MONTHS.

so there is a lot more going on here than it first seems, ok?

Re:Price?

[lkcl]

not a chance. that mis-printed and mis-read story is annoying. i actually said "the Bill of Materials for a 7in tablet is reported by the SoC vendor to be around $15". by the time you add in all the other components (e.g. 1gb of RAM not 256mb) you actually get to around $30 worth. so the sale price is going to be another 50% on top of that, then you will need to take into account tax, shipping, customs tax, customs tax on shipping, VAT, customs tax on tax on VAT, packaging, power supply etc. etc.

Re:Does Realtek RTD1186 have a FPU

[lkcl]

It would be difficult to be OpenGL compliant without an FPU, as the OpenGL support libraries will have to run on the CPU and need to manipulate floating point numbers. Obviously this can be done, but it would be a little tricky. There are also rumours of a working Android port, which also would be tricky without an FPU.

i've learned from hunting around in one of the firmware packs for an RTD1186 HTDV product that the GPU is a PowerVR SGX 531.

Re:Does Realtek RTD1186 have a FPU

[Narishma]

Not necessarily. OpenGL ES has different profiles for floating-point and fixed-point arithmetic.

Re:the elinux.org link is pretty informative

[lkcl]

As the sort of person who is interested in up-gradable hardware, this is actually very exciting. I realize this may go against the spirit of a "throw away" consumer culture, but its a fun hobby.

this is precisely what we're setting out to show the mass-volume appliance industry, that there are people who *want* to buy product that is upgradeable and eco-conscious. strategically, what we're counting on is the fact that the "chassis" and "cpu cards" will be made separately in such high volume that the overhead of EOMA-68 (which is currently about $6 on the BOM in the 10k volumes range) will come down significantly and, also, that people will begin to recognise the value of the upgradeable approach and will actively seek out products and be prepared to pay the extra overhead, precisely because they DON'T have to shell out for an entire new product - just one or other half that they want, and they can even share the CPU Card between products, thus reducing the cost *overall* of buying multiple products.

i don't know if you've seen the cost of a motorola atrix lapdock: you can get them for around $70 [get the revision 1 don't buy the revision 2 they're locked to a specific USB device id!]. that gives you some idea of the extra cost of the CPU, RAM, NAND flash etc. which would normally go into a full "one PCB" style laptop. if you can share that cost across multiple products by only buying one really great CPU Card and three low-cost "Chassis" products, think how much money you saved rather than having to buy 3 products with 3 CPUs and 3 sets of RAM that you're never going to use all at once.

food for thought, huh?

Wake me when I can BUY one

[jockm]

I am not saying the product is without merit. I do like the form factor, but right now I have two Raspberry Pis, a BeagleBone, and a pcDuino on my desk (for use in various client projects). Those are just three of the various hobbyist and industrial small ARM based systems out there.

Right now the EOMA-68 is more or less vaporware. Wake me when I can buy one, then we can talk...

Re:the elinux.org link is pretty informative

[riondluz]

are you the same lkcl that provides so much useful information regarding rtmpdump?
If so, thanks, and is there any corellation between that and this endeavor?

Re:Does Realtek RTD1186 have a FPU

[lkcl]

You probably know that means "DANGER! DO NOT TOUCH THIS WITH A 10-FOOT POLE!"

i do, but the price is *very* compelling. quotes i'm seeing are around $3.80 which is *half* that even of the Allwinner A10... and it's got PCI-Express, Gigabit Ethernet, SATA and USB-3. incredible. so, i can't turn the opportunity down.

what i'll do once a lot of money comes in is put some of that towards full-time payment of someone to do the reverse-engineering of powervr.

Re:the elinux.org link is pretty informative

[lkcl]

are you the same lkcl that provides so much useful information regarding rtmpdump?

yup

If so, thanks, and is there any corellation between that and this endeavor?

no problem. the only correlation is that the ethics that i've settled on which drive me to do things like rtmpdump you *know* that i will apply those exact same ethics to this new venture. that means that when i say "all products will be GPL compliant", i ABSOLUTELY MEAN IT. when i say "i want free software developers to be involved and to benefit from this synergy with china mass-volume factories", i ABSOLUTELY MEAN IT.

Re:the elinux.org link is pretty informative

[lkcl]

But it's better yet -- by swapping the one CPU card (which includes some storage for boot-up and some user data, with arbitrary additional storage in each chassis e.g. for movie collections and such) not only you save on buying three CPUs (and that every time you upgrade one device's CPU card, you benefit threefold), you also get "syncing" of user data without depending on the cloud -- when you slot your CPU card, the data is instantly there because you brought it with you!

now you're getting it. the cloud's a fad. this is hardware. it's *your* hardware, and it's *your* data.

but yes: typically a media centre chassis would have terabyte storage, which, obviously, you'd not have on a tablet, but that's ok: that's the way it should be.

the bit that's going to be interesting is how the OS reconfigures to cope with the differences. that's why i'm interested to work with the KDE Team, and also why they're excited about the possibilities here. KDE Plasma Active's underlying core is designed to dynamically completely reconfigure the applications - right down to the size of the menus and what's *on* the menus - depending on the capabilities of the device (screen size and so on).

that's *really* fascinating and a perfect match. whoops, i woke up and found my screen has changed - err should i reboot? no, damnit! should i terminate the app and restart it? no, damnit! should i run a completely different app, one that's designed for the small (or big) screen size? no!!

KDE Plasma Active is about the only OS that even remotely has the capability to reconfigure right now in this way. everything else is like hard-configured for a particular device size. it's gonna be.... interesting, to say the least :)