FAQ

What's the goal, again?

  • To create a synergy between the ultra-low-cost Factories and SoC vendors of China with their expertise in Hardware, and Software (Libre) Developers with their expertise in GNU/Linux and other OSes, with a view to leveraging the combination to create affordable and desirable mass-volume products that are GPL-compliant before they hit the Retail Hypermarket shelves;

  • For those products to be modular, versatile and open, so that they can be upgraded without the environmental waste of throwing away an entire device; for Retailers, Factories and users to be able to keep up with the rapid and increasing pace of technological development;

  • For anyone to be able to use the products for their original purpose as well as for Educational purposes, Research, Engineering and more.

How will this goal be achieved?

Very carefully, in small steps, having learned from the experiences of the OpenMoko and OpenPandora projects.

  1. Produce very simple EOMA-68-compliant CPU modules which can act as stand-alone computers in their own right (powered via USB-OTG) so that Software (Libre) Developers have something to start working on.

  2. Start designing IO boards.

  3. Software (Libre) Developers help develop the software to run on the products.

  4. Products go to market.

  5. Profit.

  6. Use profits to repeat the process, to the benefit of all parties, including the Software (Libre) Developers.

And... a CIC? really? But those are for Social Clubs!

The rules for CICs are "to not make a loss", which makes sense for any business. There is no limit on the profitability of a CIC: it's just that, at the end of each Financial Year, the profits have to be allocated to a charitable cause, or they have to have been ploughed back into the business. A Community Interest Company simply does makes more sense in the context of the goals of bringing Software (Libre) Developers together into this exciting technological area that has previously been dominated by vertical market sales strategies.

Why is the price of the Allwinner A10 EOMA-68 Card $15?

It damn well isn't! We are getting a massive amount of misunderstandings about this. We have reported that based on estimates from the Reference Board supplied by the Manufacturer of the SoC that the MATERIALS COST is APPROACHING $15 in MASS VOLUME quantities of 100,000 units.

That is excluding a case, power supply (which as the unit can be powered by USB-OTG is not needed), packaging, tax, customs duty, shipping and, most importantly, a profit margin.

Any company has to make a profit, and a CIC is no different. Charities and Not-for-Profit Foundations can get away with not making a profit, but Rhombus Tech is not a Charity.

Profits made will be used to fund Free Software Developers, as well as future CPU Cards and the creation of Reference Design Products: Laptops, Routers and so on, all of which will be done in an Open fashion.

What is EOMA?

It stands for "Embedded Open Modular Architecture". The concept of modular architecture isn't new: many companies have divided out CPUs into separate PCBs or modules, but it just hasn't been done recently, not on a mass-volume scale and not on a user-controllable basis. See the elinux.org EOMA page for more information.

Why re-use PCMCIA??

It's legacy - nobody makes PCMCIA cards any more: it's all changed to the PCIe-based "PCI express" aka "ExpressCard" thing. However, it turns out that Satellite TV "Conditional Access Modules" are in PCMCIA form-factor, meaning that the connectors, housings and assemblies are all still mass-produced. So there's less risk of having someone destroy their CPU card if they force-break the mechanical barriers (see specification for details) but the pricing on parts is still good in mass-volume quantities.

What's so special about the interfaces on EOMA-68?

The interfaces that have been picked happen to have been around for at least a decade, and the number of pins, including 16 pins of GPIO and including enough GND pins to separate each of the high-speed signals, by a jammy coincidence comes to exactly 68 pins.

  • RGB/TTL: 28 pins
  • USB2: 2 pins
  • I2C: 2 pins
  • 10/100 Ethernet: 4 pins
  • SATA-II: 4 pins
  • GPIO: 16 pins
  • 5V Power: 2 pins @ 0.5A per pin

The total comes to 58 pins, and there are 5 groups of GND pins to separate each group. Grand total: 68 pins. jammy or what? More information is available here.

Whoa, wait, PCMCIA is 100ohms approximately!

Yes, we know. It's not all bad. By a coincidence, SATA-II is 100 ohms and USB-2 is 90 ohms. We think that's close enough. Absolute absolute last resort: both SATA-II and USB-2 can be ramped down in speed. This would be a bugger, but at least product would work. Other than that: yes it is possible to adjust impedance through careful placement of tracks and ground planes etc.

Bottom line: we'll just have to pick the right PCMCIA connector supplier, that's all.

Why is the first CPU that QiMod picked a China-based one?

Cost and features - pure and simple. Consumers do not care about Software Freedom - they just don't. Only Software (Libre) Developers care about Software Freedom. However, GPL Compliance is very very important to QiMod, because we do not wish to be liable for GPL violations, and we do not wish our mass-volume Retail Hypermarket Clients to be liable for GPL Violations, either. So, QiMod has spent the past two years negotiating with SoC vendors to find one that has the three critical factors of: 1) Cost 2) features 3) GPL Compliance. Amazingly, it was a China-based Fabless Semiconductor Company that first met the requirements. Yes we are looking for more.

Which CPUs have you analysed so far, and why were they rejected?

We have analysed dozens of CPUs. With the exception of the Allwinner A10, none of them really fulfil all of the criteria. This section turned out to be so large that it was moved to its own page: Evaluated CPUs.

So what FSF Hardware-Endorseable options are there?

  • The Ingenic MIPS jz4760 (700mhz) - $USD 7 in mass-volume
  • The 600mhz ARM Cortex A8 OMAP 3503 ($19, 1k volumes)
  • The 720mhz ARM Cortex A8 ?AM3357 - ($14, 1k volumes and $5 in 100k)

Sadly, none of these CPUs however fulfil the mass-volume criteria of being able to do 3D Graphics or 1080p video. Some of them can do 720p, but that is not enough for commercial mass-volume purposes: it really does have to be 1080p now. 4 years ago, 720p was acceptable: now it isn't.

These CPUs are listed on the Evaluated CPUs page.

Is this an "Open Hardware" Project i.e. can I get the full schematics?

This is a misleading question: here's some clarification. The EOMA-68 initiative is an "Open Specification". That means that anyone can create either CPU cards or motherboards that conform to it. Thus, it is possible for anyone to create an "Open Hardware" compliant CPU card or motherboard.

Rhombus Tech has chosen to work with a small, dynamic factory in China that loved the idea of the "we'll do the software if you do the hardware" deal. It would be rather a different proposition for us to then ask them to release the full schematics.

Also in development is a 8mm-high (Type III) EOMA-68 CPU card with a AMD 64-bit x86 APU with Dual-core CPU, integrated Radeon 3D Graphics, with full Free Software support.

Bari also has an initiative to turn the Beaglebone or any other ARM SOC or AMD Fusion APU into an EOMA-68 CPU card, if enough people show interest in this happening. Given that the Beaglebone (and other systems like it such as the IMX53QSB, Origen, Pandaboard etc.) schematics are available under an Open Hardware License, the Beaglebone EOMA-68 CPU Card will be "Open Hardware".

Additionally, given that the Leaflabs Maple is an "Open Hardware" Project, there exists the possibility for the creation of EOMA-68-compliant Motherboards based around the adaption of Leaflabs Maple Boards.

Summary of the above: it'll happen. (update: 10jan12 - sooner than anticipated! schematics being developed here).