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

[brucehoult]

And, sadly, here we are more than three years later with nothing having shipped yet, and the "Orders made now will ship July 31 2019" recently changed to September 6 2019 or October 4 2019. Confidence level that will not slip further: low. Good thing I don't have anything in particular planned for my board.

Update: shipping date is now September 30 2020.

[speculatrix]

I searched this thread and couldn't come up a reference to the seemingly abandoned Intel Compute Card "standard"

it seemed like a good idea at the time
https://www.theverge.com/2019/3/22/1...mpute-cards-pc

[brucehoult]

And, sadly, here we are more than three years later with nothing having shipped yet, and the "Orders made now will ship July 31 2019" recently changed to September 6 2019 or October 4 2019. Confidence level that will not slip further: low. Good thing I don't have anything in particular planned for my board.

[mj-technology]

Why not try this? MJ Technology’s Katana Tablet is amazingly powerful & since it is Powered by openSUSE, you can take advantage of all the available hardware. Pledge for one today https://igg.me/at/mjtech-tablet

[Anyasz]

Not Alwinner, but Mali the issue. With all half-decent chips from Freescale that have a Vivante GPU and other stuff that has a half-working opensource driver that only needs more help to go full-blown support, they choose a crap chip with a Mali.

Still, their laptop is the best shot at a good chassis for a decent DYI ARM laptop.

[SystemCrasher]

sorry SystemCrasher i promised you a reply, apologies took a while. yes i mind! a lot!

NP, as you could see I could be busy on my own to degree I'm getting lost from Phoronix for a while.

now, what i *might* do is create a special (completely different) standard that's specially-suited to "low power blade servers" - increase the thickness a bit, to make room for a heatsink, even a small fan, and to require SATA and GbE, perhaps even single-lane PCIe: still have to think about that (whether PCIe is a good idea).

Small fans are noisy, inefficient and unreliable, since they need extremely high RPMs to get any noticeable airflow. So I guess powerful modules should use "usual" server fans, possibly not in module itself. Or module should allow room for "usual" server fans. These would give enough airflow through heatsinks even if fan is not in module itself.

On side note, "mobile" SoCs tend to be BGA and AFAIK it dissipates more heat over PCB itself. And while "mobile" SoC progress is exciting, its hard to tell the same about more specialized server solutions, which appear to lag behind, their prices/availability is nowhere close to being exciting. Revolution of low end happens due to Moore's law, cheap lowend things just got cool enough to handle larger tasks previously taking large expensive power hungry servers. Does not implies larger expensive ARMs on servers have bright future by default, since they struggle hard in terms of development times and woeful prices.

USB 3.1 was added to the standard because i *assumed* that by the time USB 3.0 and 3.1 come out in these ultra-low-power SoCs, the standard and its implementations will have stabilised.

Unfortunately it seems its adoption rate is below of what anticipated. Furthermore, dedicated HW tends to perform their tasks better. Just because it being optimized for their task.

i took GbE and SATA off the standard 1-2 years ago to make room for USB 3.1 on the basis that 10gb/sec "general purpose bus" is better than GbE (specialist interface) and SATA (specialist interface).

I think one should be careful about mandating things and long-term predictions. Say, PCs do not mandate most stuff like buses/connectors which are separate standards and mostly "typical" and "very good to have" kind of thing. Yet PCs got wildly popular. The goal of standards is to improve interop and compatibility. Where it needed and works well. If standard is hard to implement or pointless (like e.g. pci over usb bridge) I think it could face issues facing wide adoption. E.g. implemending PCI via usb lands rather unusual task on engineers, they're unlikely to have experience with it. Then it would probably perform well below of normal implementation and give no real benefits beyond formal compliance. But it would add costs for sure.

.... let's hope that turns out to be true, ehn?

So far it has been worse than anticipated. Though I would agree universal 10Gbps bus is a good thing to have any day. Though it would also bring more badUSB woes as well as some performance concerns. Say, if we'll compare e.g. latency, USB 1.x/2.x are hard-limited to 1ms and it actually screws up plenty of time-critical applications. For router it is bad if router introduces 1ms delay (real Ethernet hardly exceeds 200-300uS even on "slow" devices). Some ppl sweared about getting precise timings from GPS, 'coz they want it to be beyond 1ms accuracy. I.e. uart to usb bridge gives far less accurate timings compared to UART. And so on. I don't remember, did usb 3.x tackled this latency issue? Because bandwidth is one side of the problem. Another side is latency. Specialized interfaces balance it the way it works for them, but if it attempts to be generic interface it serves two difficult and often mutually exclusive challenges where one wants both huge bulk performance and superb latency.

[SystemCrasher]

[QUOTE=en97;n893687]Project just pushed past $150k. I think this is amazing but this project is very unappealing, its so under powered tech thats many years old and lacks ethernet and Wi-Fi?[quote] You see, in modern world, here and now one does not gets too much choice when it comes to what I call "crap-free computing".

Let assume we want more or less trusted system. Trust is not what marketing bullshit from Wintel tells us. Security got nothing to do with "secure" boot either, unless system's owner gets REAL root key of trust HW enforces in his/her hands. Not this fake "owner" key from UEFI, which is a misnomer. So those who has got eyes and working brain could see what MS and Intel are up to. Sure, not everyone is smart and knowledgeable to degree required to understand these schemes.

When ME firmware capable of reinstalling OS and hijacking system in arbitrary ways gets signed by its own non-removable owner key, it seems Intel has decided to be the only real owner of their hardware. Everyone else is a guest who could be kicked out of their HW at Intel's will. Not to mention Intel could always afford gross break-in, bypassing "security" schemes. Its like buying house where previous owner retains keys and even decides when you could go in and out. Furthermore, owner instals locks the way ensuring you can't even change them without totally destroying whole house. Sweet, isnt it?

So, there're no reasons to trust most modern x86 systems, especially Intel based and quite many other ICs running firmwares and so on. Furthermore, my previous experience suggests when someone gets secrets they aggressively refuse to show general public, it often happens to be something nasty.

I hope I'm wrong but I think this is going to be a flop. Less than 2000 people backed this so far, RaspberryPi had over 150k orders at launch.

Most humans are just a bunch of stupid consumers who do not really care about their freedoms and rights. Furthermore, computer-like devices could be WAY TOO COMPLICATED for their little brains. So they prefer to buy marketing BS.

Much less ppl could afford second thought, disregard loud marketing crap and dig deeper into subject, eventually figuring out some grim details. When they do, we could figure out Mark Zuckerberg puts tape on his laptop camera and calls those who "trusts him", ahem, "dumb fucks!" (qutation!).

It is really depends if uncloaked details are good or bad. Say, those digging into Wintel/UEFI and x86 are inevitable doomed for plenty of unpleasant discoveries on what these wintel guys are really up to. Actually, x86 HW easily scores as Untrusted Computing, Corrupted Computing and Treacherous computing. One could really expect foul play from Intel based devices and recent AMD things aren't much better either.

As for wi-fi: only few wireless ICs would do it crap-free, clear and predictable ways. Only few ICs publish datasheets (though more datasheets are floating around "unofficially"). Most wi-fi ICs would require shady, non-free blob-only firmwares. Needless to day one can't really trust such device and since its networking, it is quite bad when you can't trust network "card" to do its job right. At very least it could be (ab)used for aggressive privacy intrusion. In some cases it is probably possible to mount attacks using stuff like badUSB or unfiltered DMA accesses. While these aren't easiest things to conduct, we live in post-Snowden world where Equation Group readily patches HDD firmwares to get unprecedented level of malware persistence. So we have to consider what hardware and firmware could actually do and if this going to be good or bad for actual system owner. That's how one could evaluate if its okay to trust particular system.

When it comes to Pi, Broadcom shown multiple times they're fairly uncooperative and proprietary company, overly inclined on vendor locks, failing to get idea what the "open market" is and so on. Just buying Broadcom SoCs is awfully difficult adventure. OTOH I could buy e.g. just three Allwinner SoCs easily if I would ever need or want that. Somehow, openness is complicated and multifold. Broadcom is closed and proprietary in nearly any level of ecosystem I could imagine, from IC markets to boot sequence. It maybe not a big deal for some shortsighted consumers, but once one wants to go beyond that and e.g. do some more serious application with custom devices, such companies quickly prove to be PITA to deal with on nearly all imaginable levels. So its really nice china companies like Allwinner and so on are putting some pressure, changing markets and so on. When I implement custom device, it is really handy to be able to select from like 20 boards and modules from various vendors. Broadcom wants to squeeze all profit single handed. So we get like 2-3 options. Which isn't too much, to say the least. As the result it boils down to either you're doing devices the way Broadcom thinks you have to, or you're totally out of luck. Which sounds like a trouble on the way. Chinese companies only care about selling their stuff and everything that helps to sell more is good for them. So they're far more flexible and dynamic overall. Though they're quite bad at documentation and when it comes to software support like e.g. kernel.

[starshipeleven]

Project just pushed past $150k. I think this is amazing but this project is very unappealing, its so under powered tech thats many years old and lacks ethernet and Wi-Fi?

I hope I'm wrong but I think this is going to be a flop. Less than 2000 people backed this so far, RaspberryPi had over 150k orders at launch.

No wait, 30k in like 3 days? There must be some heavy hitter out there.

And yes, this project is using a turd card that removes all useful connectivity from the SoC.

[en97]

Project just pushed past $150k. I think this is amazing but this project is very unappealing, its so under powered tech thats many years old and lacks ethernet and Wi-Fi?

I hope I'm wrong but I think this is going to be a flop. Less than 2000 people backed this so far, RaspberryPi had over 150k orders at launch.

[ssam]

Last few days for people to get there pledges in. Currently at 75%