Purism on creating an SBC

Would like to see a completely free SBC from purism one day though I doubt it will come soon from pureOS not really working on ARM yet from my understanding. Any time soon or just false hope?

1 Like

PureOS is working on ARM64 quite nicely - the Librem5 is running it.

The question is rather, what would a Purism SBC bring to the table what e.g. a recent Raspberry Pi does not?

Cheers
nicole

8 Likes

Any completely free SBC on the market that do not cost an arm and a leg? Right now, I don’t really see anything amazing. Read a little up and good source being https://www.fsf.org/resources/hw/single-board-computers and the atmosphere seems pretty bleak.

The Rpi isn’t shocking in performance in my opinion though it is pretty good (which is the reason I own one) but… https://news.ycombinator.com/item?id=20321603

Edit: purism to the rescue?
Edit1: h-h-HKS?

1 Like

Once Purism has done the hardware and software development for the Librem 5, it doesn’t seem like it would be that hard for Purism to make a tablet and an SBC based on the i.MX 8M Plus, although I doubt that it will be cheap.

Of course, if you aren’t impressed by the RPi 4B’s performance, you probably aren’t going to be happy with the Plus either. I’m really hoping that the future RK3588 will make it possible to create competitive RYF devices, but it isn’t clear that everything will run on FOSS, especially the new Mali GPU in the RK3588.

There is already olimex working on something stronger based on NXP iMX8QuadMax :slight_smile: - though as they explain it will not be cheap (but stronger than RK3399).

Another option for a i.MX8M SBC is the Hummingboard Pulse.

What we really need is for a chip maker to release a fully FSF compliant SOC to the market, complete with all required documentation. I work at a company that would have the means to do that if it wanted to and I have even released a few new products to production myself. But I am just one Engineer there and am not very high up the management chain. During non-pandemic times, I eat lunch in a large company cafeteria at a long table of mostly IC architects, chip designers, and applications Engineers (my peers). We talk about these kinds of issues sometimes as our choice of topics as we eat and enjoy brainstorming with our colleagues while taking the lunch break. But the executives tend to eat their lunches at other tables, with their peers. Large companies need to maintain a certain status quo. It’s not quite as easy as deciding to eat lunch at a different table some day and pitching a new idea while there, although any of us could actually do that if they were brave enough to do it. I am guessing that to such an idea, some executives may see high risk, many legal questions, and at best a need to backfill many competent positions (people working on existing projects) if they bought-in to a new idea of this type and decide to fund it. Any potential rewards may appear to be highly speculative when compared to their current returns on existing investments of their resources. I don’t know how much they make personally. But I know they get a lot of stock options and bonuses every quarter if things just keep going the way they are going now. Anything that puts that situation at risk tends to be seen as a bad idea, by default, until proven otherwise.

I see a huge potential for someone to come in to the market and steal away a huge part of this market from both Google and Apple and others. Obviously, I don’t work for either of those two companies. But my employer does ship a lot of silicon. When I get my L5, I intend to take it to this table where my colleagues and I eat lunch each day (and after everyone returns from the pandemic), and to show it off. Some of my colleagues are anxiously waiting to see it already, and many of them currently have some doubts that I will ever receive it, or that it may be on-par with other phones in the market if I ever do get it. But every one of them has a firm understanding of the revolutionary changes that such a device would bring to the market and what the technical obstacles are.

I can see the potential for a day (at least I’ll keep dreaming) where perhaps myself, an IC architect, and an IC designer (people I already know well enough to pitch semi-radical ideas to) may get up enough courage to walk over to that table where the executives eat (metaphorically if not literally), show them an L5, and have a friendly discussion about how much better it could function given the right FSF compliant SOC to build upon. The executives may not understand all of the technicals, although most of them rose out of the Engineering ranks. But doubtlessly, they would like to steal-away a lot of business from Apple and Google if they could. Given the company culture where I work, this scenario or something similar isn’t completely implausible.

Some day, some SOC manufacturer somewhere needs to sit down with Todd and his team and say “give us your wishlist. If we were to build exactly what you need, exactly what would that look like?”. Then they need to go back and start building their own business plans to develop the SOC based on that information. There is a huge untapped and unrecognized market there, despite the over-abundance of surveilance capitalism that owns the market right now. It’s just much too early for most people to recognize this… especially if they don’t want to risk their current income streams with an unproven and potentially speculative market trend.

But one thing is really clear to me. The ground-breaking that Purism is doing right now is absolutely required first. When an executive at a silicon manufacturing company actually holds an L5 in his own hands and taps a few things in to see that it actually works, and then his own experts tell him that they can build a SOC that will shrink the thickness by fifty percent while increasing stand-by time from four hours to thirty-six hours, the market will change from highly speculative to something worth taking a serious look at. The window between that first realization until market prolipheration of that new market defines the boundaries of the new business opportunity. Someone is going to fulfill this need sooner or later. It’s not about if, but when.

4 Likes

Look at LibreSOC. Purism seems to support the project.

1 Like

I think that the intent of the LibreSOC is good. Most of those who are capable of contributing to that work may be tied to contracts with their respective employers as I am. My employer owns anything I develop. I do get a paycheck and can’t complain about the arrangement.

Perhaps some people who have left the industry can contribute. It’s going to take a lot of work and if the Cathedral doesn’t want to reap the rewards, perhaps the Bazaar may end up doing everything (referencing the book “The Cathedral and the Bazaar”). But at some point, capitalists will step in and use financial leverage to scoop-up every part of the market they can get. It’s my opinion that some new business models may need to develop as a result. They can’t just patent, copyright, and keep secret everything that is already safely protected already under opensource agreements and legal protections. So the idea is to go with this trend instead of going against it. If you as the manufacturer own the silicon product and all rights to manufacture and distribute it, why not give away all firmware required to drive it? The little guys with small product lines that rely on foundries for all of their silicon production may be vulnerable if they have to disclose their firmware code in addition to trusting the foundry with their mask set too. Maybe it sucks to be them.

The FSF doesn’t require silicon and other hardware manufacturers to give up any of their most valued manufacturing and chip design IP. All they need to do is to publish everything a user could possibly need know to use their product including all backdoors. Many silicon manufacturers do this anyway. It looks to me like it’s only sheer greed and perhaps short-sightedness that causes a silicon manufacturer to sell their silicon products without openly publishing the firmware code required to drive that silicon. All operational codes and reference code examples should be published in the product datasheet. Some silicon manufacturers do this by default while others may be driven out of the market through fair competition if they try to continue keeping everything secret and locked up.

2 Likes

There is this project


They started to sell Olimex:

And that’s the problem. The market leader, the Pi, sets the standard and the standard for price is low. For what the Pi set out to achieve, that’s great. However it is going to make it a difficult sell to come in with a board that basically does the same as a Pi but costs, what, six, seven times as much (and probably a worse comparison at the low end).

I think we can see, theoretically, what a Purism SBC would bring to the table - the same things that other Purism products bring to the table - but to a segment of the market that is quite price-sensitive, where devices are almost cheap enough to be throw-away.

Another challenge would be the burgeoning range of options. The manufacturers of the Pi have the volume to support a range of options (about 10 models, plus RAM variants within at least the 4B model). Purism might in practice be limited to offering one option.

Soon a fully open SoC with RISC V (open hardware) will be available from NXP:

Linux-driven RISC-V core to debut on an NXP i.MX SoC

“NXP is thrilled to be a key contributor to the CORE-V Chassis project leveraging our world class i.MX platform,” stated Rob Oshana, Chairman of the Board at OpenHW Group and VP Software Engineering at NXP.

6 Likes

Wow! It’s exciting to see a few of the bigger semiconductor manufacturers producing opensource silicon SOCs. I didn’t realize that opensource work had developed so far.

well, speaking hypothetically, based on what i know of Purism, i would imagine that closed source binary blobs for boot BIOS, drivers and firmware for accelerated functions would not be tolerated. consequently, all Audio, Video, 3D GPU, cryptographic and all other functions would be entirely Libre driven with full source, giving no opportunity for DRM lockin or spying backdoors.

on top of that base, customers and engineers could develop and debug complex applications, with less time less money and less risk.
you’ve no doubt seen this article and understand its significance
http://www.h-online.com/open/news/item/Intel-and-Valve-collaborate-to-develop-open-source-graphics-drivers-1649632.html

this has been my goal since 2012. search for “towards an FSF endorseable processor”. currently LibreSOC http://libre-soc.org

you can actually get SoCs that are considered FSF Endorseable: some of Beagleboard products (OMAP3525), and some based on the Ingenic jz4775, the problem being: they’re all 10+ year old technology that was anemic even when they came out, due to simply not having accelerated (closed source) accelerated RTL Blocks in the first place. with anemic performance and capabilities nobody wants to buy them, not even FSF supporters.

i didn’t wait. i got fed up. so i put in an NLnet Grant request and was amazed when they accepted it, put in some more and they accepted those too: you can help out if you like, there is a lot to do. don’t wait for permission: just do it.

funny that. i sent pretty much exactly that, about 90 minutes ago :slight_smile:

as Libre contributors and independent developers we own the copyright on what we do. we choose to release it under the LGPLv3. however it should be easy to convince the company that you work for that allowing you to do that is in their best interests. it is pretty standard practice for contributors to Libre projects to personally own thr copyright, and Jeremy Allison made the case pretty clear.

it always stuns me that manufacturers try to sell “information about product” rather than “product”. make it easy to use for gid’s sake!

it may surprise you to learn that the SoC is not the major cost. Allwinner A64 plus PMIC: $4. 8GB eMMC: around $5. 2GB DDR3 RAM: around $7. PCB: $2. Connectors: appx $2. Discretes: around $2 when you start to add high current inductors and capacitors.

Having a $15 SoC rather than a $4 SoC increases the BOM by say 25%? which honestly isn’t a lot. Basically the RAM and eMMC is the larger cost, particularly when you crank up to larger sizes.

2 Likes

Yes, I would expect the retail price of the silicon to be one of the least critical factors in profitability of the final product. One day, I was ordering some nuts and bolts to support a product demo board when I realized that the nuts and bolts cost me (actually my employer) about the same as the retail price of the silicon chip that the board was made to showcase. I laughed at the irony, although it wasn’t an SOC either.

1 Like

Ikcl,
Perhaps we should speak offline sometimes, if that is possible. I am not sure how to arrange that on an anonymous forum. I can share more information in private which might be your situation also. I have an NDA with my employer. But I can probably discuss in-generalities, projects that they’re not interested in… yet. I’ve actually been pursuing lines of communication, that reach pretty high in the company in a few instances where people know me and they are bouncing ideas off of a few executives. I haven’t yet had enough specific information to put together a compelling value proposition based on what I know they’re looking for. But with a feasible plan, it won’t be difficult to shortcut to the right audience. There is always money available to fund the right projects after the likely ROI becomes evident.

1 Like

well that’s quite interesting as well because we are just in the process of setting up a commercial version of libresoc, with a holding company etc etc the general idea being, where customisation is needed and turning it into silicon, that’s handled as might be expected, and where appropriate they liase with the libre team, funding and/or contracting them to provide BSPs, public docunentation etc.

the reason for the split between the two is because the libre side acts like an outsourced R&D which, being funded through a Foundation, is tax deductible! and of course, collaboration is always better, and we’ve found that by not being under NDA we can go out anywhere on the internet and say, “hey got an interesting problem anyone got any clues?” ask in the right way on comp.arch you get a firehose of intelligent responses.

imagine asking for authorisation from the company you work for to do something like that. contact me in 3 months when the legal team’s signed off on the idea (and charged more money than you’d save)

1 Like

i mean, “a company, any company” not “the actual company you work for” :slight_smile:

Personally, I have a hard time buying anything Not made in America. Especially now after my L5. And, considering how libre it is, I have no intentions on buying another RPi. Unless I come across a used one maybe.*
Currently, the only thing that comes close is a Beagleboard.
Also, Purism could have different form factors from the RPi, while still having the PINs and hat compatibility. It could support things like an m.2 card. That alone, for root and boot, seals the deal. It’s painful trying to use my Pis because of the mSD (and USB2) speeds.

For me, I see no use for the Mini or similarly sized machines. In the past I have had a fascination with building a Mini-ITX machine but that’s as much for fun than anything else. Micros would be great to have here and there for IoT or used in elegant cases for things like portable gaming consoles and music pllayers.

Sorry if that was a bit long winded and/or didn’t quite hit the mark.

Mainly, I just want a RYF made in America every type of device. :wink:
I might even buy multiple L Micros because I could probably use a few for my night job. Just keep offering the financing option and the blows to my wallet aren’t so bad. By the tine I have one thing paid off, I’ll order another one! XD

  • I actually found an original RPi at a yard sale recently. Set it up on the stereo system to relay MPD, Icecast and other music streams.

I started a thread about a Librem Micro a few weeks back.
https://forums.puri.sm/t/the-librem-micro-a-sbc-whod-be-interested/11109