The EOMA68 Upgradeable ARM Board Faces Another Setback: HDMI Connectors Don't Fit

[wizard69]

I backed the original crowdfunding project. After a few of the delays I contacted the maintainer Luke Kenneth Casson Leighton and offered to donate my pledge to the project without getting anything, since they were cash-strapped.

I love the idea, but I think even if he ships the completed project it's effectively a dead end. I hope I'm wrong.

The project was dead even before it launched as the idea simply isn’t competitive. The idea sucks for a laptop for a number of reasons:

1. PCMCIA sucks for passive and active cooling so realizing full performance out of the chip will be difficult. This especially if the SoC is laptop class.
2. The requirement for daughter card means excessive costs with no real benefit.
3. The card is narrow so even if you put some I/O on the exposed edge, you are extremely limited physically.
4. Modern SoC have many I/O ports built in requiring a minimal of support to interface to the real world. This makes it extremely silly to go off board to reach the real world interfaces. Even more importantly those real world interfaces are evolving rapidly. Connecting to a daughter card means you are constrained to the legacy ports built into that card. Either that or you replace both cards at the same time.
5. Connectors imply reliability problems.
6. If you want to market to other uses beyond laptops the board must be able to run stand alone. This comes back to the I/O problem.

I could go on but 6 issues should be enough to explain why so many thought that this was a boondoggle. Now all of that said I’d love to see the industry adopt a decent board / module standard for laptops. That standard would need to address cooling, either passive or active and some of these other issues. Well defined basic requirement ports should be established and the space for optional ports.

The defined set of ports might be a power port (ideally allowing wide range voltage say 12 to 28 volts), two USB ports, and HDMI port and maybe a USB-C port. The optional port area might be 10 x 30 mm. Ideally the board would bolt to a frame / heat sink providing passive cooling that is also part of the standard and which bolts to or snaps into the laptop chassis.

The obvious thing here is that you get a known number of ports and have an expansion area the vendors can define. If the standard is well done the board can easily go into a desktop chassis that supports active cooling. Likewise it can be embedded into another product. So you have flexibility for mass production supported by many use cases. You could even support internal expansion, especially SSD ports.

[tuxd3v]

What about that one that Google/Motorola ditched but the driver code still exists for in the kernel as Greybus? That was all about modular interchangeable parts.

The Unipro specification?
I believe that GreyBUS, was the application Layer, L1-L4 of the OSI model..
I believe it had a big Overhead.. but was created to be modular.

Don' t know what happened with that Ideas/Project( Google brought what we know as project ara, to create a modular phone.. ).
Initially I tough it would be a good Idea,
But then Google also found out, that "selling parts", would prevent them from selling phones

I think( not sure ), people wanted "modularity for free", but with it, comes complexity/bigger amount of processing power/big latency, which defeated the initial idea..

[blackiwid]

I would absolutely blame the management team

I don't, of course a company like LG or Samsung can make just other demands and the company deliver them stuff make very very sure that they don't deliver shit because loosing them as customer costs many millions if not billions of dollars. While a hand full of people in a small startup are a small fish to them, they do the absolute minimum to comply to the contract and probably not even care to much to fulfill that fully. Because the chances of a law suit are small and even if they have to pay them something back it's peanuts.

It's like complaining that a small new game development team should be as good as blizzard... you just can't expect that.

But I think it's not as dark as the article portraits it:

1. the investment sum of the individual bakers are very small not like 600-1000 euro for a big phone or tablet but a much smaller amount.
2. the power of the hardware doesn't really matter than much, cause the audience doesn't want a strong multimedia pc but more or less a writing machine.
3. the longer it takes the better the driver get's since a few months all winner drivers are pretty descent before that they were utter crap.
4. I still see huge value in such devices, somebody could create a laptop case with a mechanical keyboard that has a lifetime of 10+ x years... you basically can't build a OEM 1 part Notebook with a mechanical keyboard because it would be waste after 3-5 years the system would be obsolete and you can't fix it.

Heck I see it even with my non-mechanical X220 keyboard, I thought the newer more chiclet style keyboards can't be that bad, but they are imho.so I would love to have a more modern cpu for that device, or even better would be able to switch the case against a identical with fullhd which does not exist for X220.

To get that more modular is a huge gain... Yes the hardware is no huge thing, but cheap enough that it does not matter and could work for some mobile requirements even in the first iteration. If nothing else then it's a great prototype like this cars that remember us that driving could be very different but then never get into series, but they are further on that front.

[tuxd3v]

Yep and again it doesn’t address current realities. It is far cheaper to take the common ports from the SoC and apply them to some edge connectors. So at a minimal a couple of USB ports, and HDMI port, a power port (maybe a USB-C port) and maybe a network port. In other words enough basic I/O to be functional standalone.

It doesn’t prevent theft. You might get lucky and have your data with you (assuming on card storage) but the rest of the device is gone.

You are describing a SBC format, I think..
The idea was something like "fire and forget..", insert the card, and all is connected..

The rest of the hardware, could be stolen( or not if it had a physical locker ), but your data, your settings configs , all were with you, providing that you only use the internal storage, which at some point would not be possible( 8 GB of internal flash, or so.. )

I really don’t see PCMCIA giving you flexibility here. It actually reduces flexibility, forces the production of an expensive mother board and physically constrains I/O to the wrong edge of the card. Consider how flexibly ATX was, the well defined panel layouts allowed that standard to adapt for years as mother boards became higher in component density and function. PCMCIA puts a box around the high density chip that ends up making the platform inflexible and expensive.

Now I don’t want to dismiss the importance of user expansion, but here they would have been better off using PCMCIA as the expansion slot especially if they could have emulated the original pin outs. Even as an expansion solution PCMCIA is a bit dated so this is only an alternate way to look at the use of the standard. A new technology expansion solution would have made more sense.

You have a lot of valid points here..
I think, the Idea was to reuse a already available standard, without the massive amount of money/Investment/Risk needed otherwise..

Creating a new Standard, and forcing people to use it, costs a lot of money, at least 10 years of forcing it into the market, and you don't really know if it will be ever accepted by the public..

polarathene, spoke above about GreyBus or Unipro, as a solution for modularity, that was an example, but it failed even with Google behind..

I kinda like the PCMCIA idea, as it is a "atomic piece", easy and fast, to insert/remove.
It also has its downsides, since you concentrate everything in one place, it turns more difficult to build a motherboard with flash/ram/hdmi, and so on, on it..
The thermal design, is also limited..

[discordian]

The problem with this Modules, is that you need always, or almost always a base carrier board, that has the same complexity has the module, majority of times even more..

And why would that be the case?

1. Either everything is on board, then the carrier only has the physical endpoints (Phy's/magnetics)
2. Or you can extend the board via PCIe (in some cases I2c or other busses)

SMARC covers alot usecases already, a Laptop using SMARC could be done today, with a choice between ARM and x86 boards.
I fail to see why EOMA68 has any advantage.

Socs typically have "too much" interfaces you rarely use at once, you shift the problem from the carrier to the module. Which is a rather bad idea,
do you really thing some modular backplane exposing the connectors on the CPU Module will fly on a laptop?
You want your Connectors being incorporated in the body around 3 sides.

They wanted to do all( or almost all) in the PCMCIA card sized board..
Also if you go for launch, you can simply get the card out, and be with it in your pocket, preventing theft for example..
You can switch between computers easily also..

Same as with SMARC, but I doubt you have any clue about that. See for ex. https://www.tq-group.com/en/products...ture/tqmxe39s/ (Intel Atom) or https://www.congatec.com/en/products...onga-smx8.html (ARM)

SMARC doesn't give you the flexibility PCMCIA cards give..

No it gives you more, if the EOMA68 card is supposed to carry a fixed amount of connectors (as far as I understand) and you would not need different CPU module configurations (=connectors) for different appliances.

[tuxd3v]

And why would that be the case?

1. Either everything is on board, then the carrier only has the physical endpoints (Phy's/magnetics)
2. Or you can extend the board via PCIe (in some cases I2c or other busses)

SMARC covers alot usecases already, a Laptop using SMARC could be done today, with a choice between ARM and x86 boards.
I fail to see why EOMA68 has any advantage.
(..)
No it gives you more, if the EOMA68 card is supposed to carry a fixed amount of connectors (as far as I understand) and you would not need different CPU module configurations (=connectors) for different appliances.

I see that you are looking into a different type of device..

You are looking into a device that:
You can configure, with some effort, but then its there with a module, until you go ahead and have the trouble to substitute it by another one..

There are Already Solutions using that idea, with SMARC/COM Express/others..
They are very good Ideas, but are not for the same target application that EOMA68 is..

You can change SMARC/COM/etc when you want to, but not with same simplicity and easy as PCMCIA.. they are different!!

SMARC/COM/etc could work nice in a Laptops too( in fact there are already solutions for that in the market...),
But even tough they are upgradable, they follow the Same concept as Ram Memory Dimms.. you need to "pull some dirty tricks", not the same concept.

But its an alternative concept, harder to change, and made not to change frequently as PCMCIA..

[discordian]

I see that you are looking into a different type of device..

You are looking into a device that:
You can configure, with some effort, but then its there with a module, until you go ahead and have the trouble to substitute it by another one..

There are Already Solutions using that idea, with SMARC/COM Express/others..
They are very good Ideas, but are not for the same target application that EOMA68 is..

You mean the EOMA68 where people print replacement parts with a 3d Printer, but are unable to open the chassis or a trapdoor and change a module with a PCIe-like connector?

You can change SMARC/COM/etc when you want to, but not with same simplicity and easy as PCMCIA.. they are different!!

There are more similarities than differences. Heck, just develop a Metal casing for SMARC and your argument is completely moot...

SMARC/COM/etc could work nice in a Laptops too( in fact there are already solutions for that in the market...),
But even tough they are upgradable, they follow the Same concept as Ram Memory Dimms.. you need to "pull some dirty tricks", not the same concept.

Not that I know of. The idea behind it is that Arm/Opensource Devices typically are lowerend throughout. Id like to have a chassis with a good monitor, good keyboard and be able to use an ARM (or RISC-V) module.
(If I can go completely crazy: Imagine having a TV with slots for 4 such cards, and you could plugin "PCs" or console boards there.)

But its an alternative concept, harder to change, and made not to change frequently as PCMCIA..

I am not sure the advantages of unplugging the CPU card daily is something alot people are wishing for. Even less if they have peripherals hanging physically on that card (really.. what were they thinking).

[tuxd3v]

You mean the EOMA68 where people print replacement parts with a 3d Printer, but are unable to open the chassis or a trapdoor and change a module with a PCIe-like connector?

I don't know if I understood you..
I am talking about the board of the article..

There are more similarities than differences. Heck, just develop a Metal casing for SMARC and your argument is completely moot...

I understood your point,
But what you want is not the same concept as the EOMA68..

Not that I know of. The idea behind it is that Arm/Opensource Devices typically are lowerend throughout. Id like to have a chassis with a good monitor, good keyboard and be able to use an ARM (or RISC-V) module.
(If I can go completely crazy: Imagine having a TV with slots for 4 such cards, and you could plugin "PCs" or console boards there.)

Just search for them..
there are MXM graphic cards( ex: for laptops.. )
There are carriers and modules for small computers( more for embedded.. )

But I think a complete up-gradable solution doesn't exist, at least that I know of in ARM..
I know that there are solutions like that, for MIPS32r5, but for All-in-one, or small mini-ITX like format..

You were the one speaking about SMARC.. of curse you should have some knowledge about..?!

I am not sure the advantages of unplugging the CPU card daily is something alot people are wishing for. Even less if they have peripherals hanging physically on that card (really.. what were they thinking).

That will depend, on what your needs are..
I for instance cold like the idea, of unplug it for launch, and get it with me..
If you shutdown the computer, there are no peripherals hanging..

It all depends, what you like/need..

[Michael_S]

I think the important innovations here that LKCL was trying to reach were:

1. The complete hardware device spec would be open under a copyleft license, so any person or company could produce and sell the devices. Raspberry Pi, Pine64, Orange Pi, and other SoCs do not - as far as I know - do that.

2. The focus on reusable components. Your laptop chassis or mini-desktop chassis could last 15 or more years, and you just upgrade your tiny EOMA68 device. That's better for you financially and also cuts down on e-waste.

It may be a colossal failure, but I think both ideas are worth pursuing. Big corporations will never pursue either idea. An open spec allows other companies to compete more directly, which cuts margins. Apple, Google, Dell, Samsung, etc... would rather sell you a $700 laptop every four years instead of a $500 chassis once and then a $200 card every four years.

Which is exactly why this project is beyond stupid. It effectively is 1980’s thinking and dismisses the reality of SoC technology.

The fact of the matter is that any viable solution like this needs to have the majority majority of the ports the user might want on the module. Frankly you would want the common ports coming out of the SoC available to power the device and support connecting to a user.

I'm far from a hardware expert, but my work laptop connects to a docking station that connects to two full HD monitors and four USB 2.0 ports through a single USB type C connector. And it charges through the same connector, too.

So it seems to me that something the size of a PCMCIA card could have all the connectivity required through a single USB type C connection. An external gadget could convert to HDMI and all of the other necessary connector types.

[tuxd3v]

I think the important innovations here that LKCL was trying to reach were:
So it seems to me that something the size of a PCMCIA card could have all the connectivity required through a single USB type C connection. An external gadget could convert to HDMI and all of the other necessary connector types.

I believe that today is easier to go USB-C than in the past..
USB-C is now a standard used by the market..
Probably if the project had get tremendous impetus, we would be now seeing the second iteration of it, with USB-C, and probably a Allwinner H6 on it, clocked @1.8Ghz, which would be a big jump in possibilities, with a low power CPU

Dreaming is free, or almost