The future of Libre Computing: Crowd Funding Campaign Starts Now

268 replies [Last post]
Eemeli
Offline
Joined: 01/04/2014

You can indeed use the Guix software on ARM, however there's only instructions on how to install the GuixSD operating system on x86. I can't wait for someone to port it to ARM! Perhaps I'll take a look into it once the EOMA-68 is sitting on my desk!

Also, it would be a nice learning experience to see if Trisquel could be ported to it. It will be very convenient to swap "hard drives" by just plugging a memory card out and another back in instead of having to open a laptop with a screwdriver each time.

muhammed
Offline
Joined: 04/13/2013

Thanks buddy

lkcl
Offline
Joined: 03/29/2015

ha, well.... welcome to modular computers: with EOMA68 you can have *two* affordable computers, one for "experiment with updates", the other "stable"... and swap between the two, within seconds. cool huh? :)

muhammed
Offline
Joined: 04/13/2013

It's reassuring that you have not had any big problems with Parabola. I will back up regularly so that if something goes wrong, I can use a different computer until I figure out how to get Parabola working properly. Thanks for your reply Eemeli.

lkcl
Offline
Joined: 03/29/2015

> Sorry for all the questions -- and thanks for taking the time
> to answer ... just a few more from me

no problem, i got time :)

> For the laptop, does the LCD connect to the Computer Card via a LVDS?

EOMA68 is RGB/TTL. there's an LVDS converter IC (a TI SN75LVDS83b)
on the laptop housing "Main PCB1".

> Will your business sell different LCD screens (perhaps higher res,
> perhaps just replacement screens) in the future that we can buy and
> swap in to the current laptop body (perhaps with a case mod, perhaps not)?

higher res for this laptop? no, sorry - higher res means "more power".
more power means you're over-budget (15 watt max), more power means
"redesign of PCB1 due to increased backlight current", also a totally
different connector would be needed.... it's really not as simple
as "plug in a different screen"

the whole *LAPTOP* has to be customised based around a holistic
approach. staggering, but that's how it is.

the LCD that's been picked is one that's mass-produced, commonly
available and pretty low-cost. you'll be able to get replacements
for years.

> Will your business sell replacement keyboards, batteries, wiring,
> battery controllers(?), etc?

the customised and bulk-bought items - battery, wiring, PCBs: yes.
the PCBs will be (are) available as GPLv3+ "Open Hardware" so
that other people can make them as well for you.

keyboards: we will deliberately pick one that's commonly available on
ebay so you can order replacements easily. likewise the LCD.

> Is Parabola appropriate if I want a stable, reliable OS to use for work?

for basic stuff - internet, email, libreoffice, printing - yeah.
you won't have adobe flash (but that's not such a big loss - bye bye
adverts hmmm i'll really miss you).

> I have heard over the years that Parabola is for advanced computers

ultimately, software is pretty much the same regardless of the OS
Distribution. it's more about the window manager than it is about
the distro.

XFCE4 is pretty much the same look-and-feel as Windows 95 was, and
it's still just as easy to use.

personally i prefer KDE 3.5 (now Trinity Desktop) but it's not available
for ARM at the moment.

> and that people who use Parabola must regularly fix bugs because
> Parabola gets all the newest updates.

people who *want* the latest updates can get them. but here's the
thing: what you can do is either:

* take a copy of the microsd card OS before doing an "update"
* get 2 Computer Cards and use one for experimenting.

the 1st way is a cheaper version of the 2nd way, but basically
you can check if an update works for you, and if it doesn't
then you just go back to the older card (memory card or
computer card).

> I am fairly novice and won't have time to learn how to fix bugs

as long as you can _report_ them clearly, explaining what's
wrong, and are happy to follow instructions that people ask
you to, to help track down the problem, then you won't be
the one actually _fixing_ the bug.

remember to respect the fact that the people you will be talking
to are volunteers: don't *DEMAND* that they fix *YOUR* problem.
be patient with them (and yourself)

and, given that you can copy (and then swap) the microsd card
or just use a 2nd computer card, the problem of "argh i updated
and need it fixed right now" goes away.

btw when applying fixes / patches recommended by the developers
treat it as an entire upgrade ok! use the same procedure:
make a copy of the micro-sd card (or use a 2nd Computer Card).

> ... reliability is really important to me. Would Debian be safer
> in this regard? How can I use Debian while sticking with Trisquel's principles?

the OS is on the Micro SD card, you could actually try both out.
debian has more developers, but i found that the parabola team
were actually much more responsive, they surprised me by packaging
some software within two days! you can't get that kind of response
from debian!

you can however meet *most* of trisquel's principles by removing
the "nonfree" repository from /etc/apt/sources.list (or just
not adding it in the first place). however you also need to
watch out for browser default search-engines.... a stack of
other things as well... i actually don't know the full details
but i know there exists a package called "vrms" - a "Virtual
Dr Stallman" which will help you find things on debian and warn
you :)

muhammed
Offline
Joined: 04/13/2013

Thanks again Luke. I appreciate your thought process for sourcing parts.

I hope that I can make Parabola work; I will think about whether to choose Parabola or Debian (if Debian, I will install the RMS program on my own). I will need Microsoft and Adobe for work, and will use a different computer when I need their programs.

I hope that the developers of programs like Parabola (and Trisquel) gather enough of a following to run successful businesses on their work. The volunteer aspect is cool, but like you said our community can only count on that so much.

lkcl
Offline
Joined: 03/29/2015

> I will need Microsoft and Adobe for work, and will use a different computer when I need their programs.

you can use rdesktop to connect remotely if you prefer, as long as they're not too graphics-intensive programs. enable the "Remote Terminal Service" on the windows machine and you can work remotely. i used it once over a VPN - it was slow but allowed me to access the machine.

> The volunteer aspect is cool, but like you said our community can only count on that so much.

... that's why selling hardware works. with hardware it's clear that the "exchange equal to value" has to happen. which i mention here https://www.crowdsupply.com/eoma68/micro-desktop/updates/feedback-from-backers in a bit more depth, hope you don't mind everyone, i referred to the conversation here :)

muhammed
Offline
Joined: 04/13/2013

Haha, it's cool that you linked here

mzs114
Offline
Joined: 09/27/2013

Any possible teaming with Cubie(radxa, etc) project in the future?

Also, the more I read your responses the clear it becomes why you dropped supporting SATA, Ethernet interfaces and instead focus on the common ones, like the "Universal" Serial Bus(USB). When I compared what is there to what you said would be when you started the project(SATA, Eth support), I was hesitant to back the project..
It took lot of time to follow your updates wherever possible, Whew! it would have taken more to compose these!

Also, can there be a version of the micro-desktop where we can have more USB ports? Or, any other supported ports by the EOMA68?

And nice to see that EOMA68-A20 supports OS like FreeBSD(choice to user :) not tied down to just GNU/Linux.

lkcl
Offline
Joined: 03/29/2015

hiya mzs114,

tom and prince were inspired by the early work on gpl compliance with allwinner that i did back when the A10 first came out. tom (and many others) did not have the restriction of strict gpl compliance for the bootloader, nor of having to develop a standard. Tom however worked very hard internally (before he quit working for allwinner) and even harder afterwards to get gpl compliance and to create fantastic documentation. he's actually been supportive of *me* because without that persistent pushing behind the scenes (which the linux-sunxi team worked on as well) he wouldn't have been able to start his business.

now, about working with him _in future_, i'm sorry to have to tell you that when we worked with aaron seigo on the Improv, we had agreed that we would handle the CPU Card, and his team would handle the "engineering board". we had everything ready to go (despite what i learned only two weeks ago he had been telling everybody in the free software world), but he stopped communicating with us despite repeated attempts to ask what the numbers were on the campaign.

we'd guessed that they were low, but we had no idea how low. so we had to think ahead, *WITHOUT* any input from aaron, because he'd stopped replying to our repeated and urgent requests for clarification and progress reports, what numbers he could potentially have got. i asked tom's team, as an emergency measure, for help, and their factory went to all the trouble of quoting for an extremely low production run of only 500 units (which is a lot of time and a very small run for very little profit). he did it as a favour because we'd helped him get his business started.

when aaron finally got in touch, he was pissed off and blamed *us* that he had only managed to get 250 preorders (1/10th of the planned 2500). what he hadn't told us was that he was running out of money, so couldn't buy the 500 units, let alone the 2500 planned on which the *pricing* of the improv had been based. and because he'd set the pricing of the improv so low (based on a 2500 production run) there *was* no profit based on the unit cost of the 500 orders. and i certainly couldn't go back to tom's factory and ask them to redo the quote for as little as 250 units: they were too big a factory for that kind of production run.

so he destroyed not only the reputation of his own project, but also destroyed our reputation with a China State-Sponsored Factor (and lost the entire Free Software community a massive funding opportunity in the process), destroyed our reputation with Tom *and* adversely affected Tom's reputation with his main factory.

and i hear only 2 weeks ago that aaron's been telling everybody that *we* were the ones who failed to communicate and were trying to "hide failures" and "weren't ready". total and absolute utter bullshit but of course it would have helped "placate" his team and his friends... temporarily, until they learn in other ways that he operates in an unethical way.

anyway. leaving that aside....

the micro-desktop has only 2 USB ports because to add more would require an on-board USB Hub, plus an uprated Power IC (which is already at 15W). that would potentially involve heat-sinks so it is all extra cost and complexity that i'd like to avoid. you can always use a powered USB hub.

one thing that would be nice which i'll consider in the future is an RS232 port, but they're quite big so have to be evaluated to make sure it can fit without changing the size of the box. in the meantime a USB-UART on the 20-pin header does the job for early boot debugging.

"other ports by EOMA68"? EOMA68 won't ever change, it's final. that's really important. standards should never be changed (they should only get faster in a future-compatible way, like both USB2 and MicroSD do).

if you mean "will the micro-desktop bring out any other ports that are on EOMA68"? everything's brought out already or used up already: there's a 20-pin expansion header inside - that's where "everything else" is brought out that's not already on external connectors http://rhombus-tech.net/community_ideas/micro_desktop/

the A20 has pretty amazing OS support, it's because it's been around for a while, and the fact that the A20 has both SATA and RGMII (GbE) makes it really popular. there simply isn't any other low-cost SoC out there with as much support for free software with the same extensive level of interfaces, it's really very unusual.

let's hope the R40 due out soon has been designed with the same set of interfaces.

mzs114
Offline
Joined: 09/27/2013

I see, but when I intend to use the micro desktop as an end user, it lacks the audio/mic ports usually found on such desktop devices.
I compared to what the laptop was offering and thought maybe due to the size constraints the ports are lacking, hence my query.

Next, I am thinking of possible uses a typical user here in India could use this desktop for, we have regular power cuts, some use UPS to protect our work, but the cost is high and many average Indians shy away from such devices, a low power power device can be boon to many here.

On a side note, a year ago I mailed the IIT-B group to consider Olimex hardware for the next version of their product. They acknowledged and were positive of this.

http://laptop.fossee.in/

Please contact them if you can help them in any way. Their group mail ID:
laptop[@]fossee[.]in.

A professor by name Kannan Moudgalya is the lead of the project, a FLOSS supporter.
http://www.che.iitb.ac.in/online/node/52

lkcl
Offline
Joined: 03/29/2015

power budget and simplicity: the schematic for the micro-desktop fits onto a single page. the moment you add a built-in USB hub it's added 40 extra components and requires about 1 sq in of board space (which isn't available). adding a built-in USB Audio IC, another 50 components, and another 1 sq.in of board space.... it's a totally different project in other words.

you can always get a CM108AH USB-Audio dongle, like the one that thinkpenguin sells.

if you want to run off-grid, you can. the power chip in the micro-desktop runs on anything from 7 to 21 volts, so even if power is battery-operated e.g. 12v and starts to run down, the computer will keep operating even as the battery drops to 9v or below.

you'll have to find a 12v LCD from somewhere with a VGA connector, those are quite common despite what everybody says "the whole world is going to HDMI"... they're not :)

i'll be in touch with professor moudgalya, thank you.

lkcl
Offline
Joined: 03/29/2015

managed to get in touch with him, thank you!

tonlee
Offline
Joined: 09/08/2014

http://laptop.fossee.in/ If the notebook gets a port for the computer card and the notebook sells in big number, that could, if price is right, be beneficial for the computer card.

Andromium is making a notebook format for phones. https://twitter.com/getsuperbook. Maybe it can get hacked for a computer card.

lkcl
Offline
Joined: 03/29/2015

the fossee laptop needs a complete redesign - as in, it's a designed-for-obsolescence standard intel ODM product - i've been through the process of assessing whether laptop shells are suitable for conversion to EOMA68: they aren't. apart from anything the entire design mindset is not suited to "right to repair", components have a massive knock-on effect... you're truly, truly better off just starting from scratch.

now, the superbook *might* seem like it could also be converted, but you don't have to. it has a USB port. it uses a DisplayLink USB-to-VGA IC. DisplayLink has supported ARM (reverse word-order of x86) since around 2010 when i worked with bernie to alert him to the bugs in the linux kernel drivers.

bottom line: you can just plug in a superbook directly into an EOMA68 computer card. incidentally it would be fun to run dual-screens, plug in an HDMI LCD as well.

lkcl
Offline
Joined: 03/29/2015

btw we'll be adding the "passthrough card" to the list, update soon but here's a bit about it at the moment
http://lists.phcomp.co.uk/pipermail/arm-netbook/2016-July/011204.html
http://rhombus-tech.net/community_ideas/passthrough_card/

Eemeli
Offline
Joined: 01/04/2014

Congratulations on reaching 40k!!!

lkcl
Offline
Joined: 03/29/2015

thanks :) still have to get another 87 computer cards before we reach the first critical threshold (MOQ 250 computer cards)...

hack and hack
Offline
Joined: 04/02/2015

lkcl, I was just thinking about something: Why don't you try to make a short but dynamic video about the whole thing?

You know, the fast happy music kind, with fun animations and bullet points. I know I sound sarcastic but seriously, it's a format that packs a punch, and it's easy to share.

I have yet to read the whole thing, but to me, this is what it's about, and what's the coolest things about it:
* credit card-sized computer. Super awesomely practical.
* printing laptop parts (like one frame default color, another one with a brighter color, and the last one with like way too much colors with "uh..." written on the frame, just for the fun effect.
* 100% libre. This one is hard to illustrate in a simple AND appealing way. Portraying the non-free constraints is probably the easiest part (with short real-life examples).
* To not make the video too long or too hard to digest, I'd add all the other points (the ecological aspect) in a fast-paced rythm for a few seconds. And then point to the funding page of course, and maybe a contact adress (or this forum's thread).

The lack of current CPU and RAM super capacity shouldn't be hidden, but shouldn't be put forward too soon though, I suppose. Same for the connectivity.

Anyway, just food for thought. A punchy video is worth a thousand words. Beats my short attention span any day.

Oh, and I'd definitely avoid a preachy tone (which I spontaneously associate with free software unfortunately) or lengthy explanations. Preach in a fast, light and fun way, that's the way to go IMHO. Then come the details.

lkcl
Offline
Joined: 03/29/2015

hiya hack, yeah we find it takes too long to explain the whole thing in one go, we need someone to "ask questions of" if you know what i mean. we tried the "explaining" video it just didn't work, and it's because there's nobody else in the world actually doing this!

but, also, i don't mind that we won't be doing 2,000 or 20,000 straight away, it would be too much: we need to go gently and make sure everybody is happy.

hack and hack
Offline
Joined: 04/02/2015

I was thinking of something really short, you know, 30 seconds or something.
I see your point though. Maybe for the second batch :)

lkcl
Offline
Joined: 03/29/2015

https://www.crowdsupply.com/eoma68/micro-desktop/updates/hope-2016

ha! :) just when i said "nehhhh" someone videod live at the hope2016 confrecne, let me know what you think

hack and hack
Offline
Joined: 04/02/2015

I love it, it does the job just right. It made me really want one.

Also, that's some impressive TMNT Donatello color scheme you got on that laptop :)

lkcl
Offline
Joined: 03/29/2015

's'not fair i was told i had to make some boooring silver colour scheme... so in rebellion i picked robotic metallic silver, and it's turned out prett y awesome because the light catches the parts in different ways (the parts are printed at right-angles to each other), it's a bit like those souped-up cars with two-tone metallic paint which changes depending on how the light catches it...

hack and hack
Offline
Joined: 04/02/2015

I personally like a matte finish more. But that's one of the nicest things about it, it's infinitely customizable. I mean I was just thinking of engraving stuff on the cover lid (yeah, like that fruit thing, which could be switched by a nicely designed banana silhouette).

But since I have better taste than that, I could put the banner of House Trisquel, or Debian, or my face engraved in it (with a stupid grin), or all of these of course.

And instead of a whole lightning system for the logo, it could simply be some reflective paper or cloth.
Anyway, fun times in perspective.
I really don't need yet another machine, but I totally want to see an 8 to 16GiB of RAM and a monster CPU one day, completely free.
personalized looks is an awesome bonus.

Btw, you probably investigated this since there are parts already made of wood, but would it be a good idea if I try to make it completely out of wood (treated against fire hazard of course)? Or at least with more wooden parts?

Last question: is there enough space in the compartment for an HDD or an SSD? can it be solidly held in there? That would be nice.

lkcl
Offline
Joined: 03/29/2015

embroidery panels are also 1.5mm thick (the plastic ones), you might get away with using them on the LCD lid, but not the base. a custom PCB at some point with SMT LEDs and components on it... there's room for another power cable at the other side...

i did think of paper mache with resin for the casework, problem is casting them into molds... getting them out... yeah, it'd be a big project.

the compartment i *think* is 8mm x 60mm x 90mm so one of those really short (embedded-style 2.5in SATA SSDs) ones would fit, remember you have to get a USB-to-SATA converter as well.

hack and hack
Offline
Joined: 04/02/2015

I've also seen another machine with nearly complete wooden casing. The beast looks super heavy though. You could cut onions on it or something. It actually looks a bit off.

Depending on the thickness of the lid, it's probably possible to modify the plans so the whatever I want to be engraved would simply be printed. But I have no idea if it would look good or not. Maybe sticking a plastic shape from a contrasting color in there would.

If it fits and is well immobilized, adding a HDD/SSD to the mix would really be cool. Depending on the cables and all (and the LCD pad), do you think there is room for me to modify the parts to print without too much problems?

EDIT: Maybe it would need some breathing space, or at least a metal plate at the bottom.

lkcl
Offline
Joined: 03/29/2015

laser-engraving on the 1.5mm plywood would not be a problem, and would be kinda cool. we have a few ideas of replacing the panels with embroidery plastic grids and so on: only recommended for the LCD panels though!

you *might* be able to fit an embedded-style SSD into the internal compartment. no there's no room to modify the casework to make it bigger than the 8mm x 60mm x 90mm (or thereabouts) that it already is. the compartment lid could have aluminium put into it instead. the base plywood takes up 1/2 the base, it's quite big (160 x 190 or something like that). splitting that in half is not a good idea, the design critically depends on that being a single piece, under compression.

a_slacker_here
Offline
Joined: 06/29/2013

That video was not porn but it got me a boner.

Seriously speaking, you were phenomenal explaining it's possibilities, I really find this project a revolution and I'm sad because I'm not in condition on helping the project economically, I will try to get the word out in order to help though.

My most honest "good luck".

SuperTramp83

I am a translator!

Offline
Joined: 10/31/2014

omg, after seeing the video, I must say:

genius!

janbaggerudlarsen
Offline
Joined: 07/25/2016

I am very excited about EOMA68 and I am very happy to be able to back the project. Finally something that is Libre for real!
The Hope video was brilliant!

lkcl
Offline
Joined: 03/29/2015

that's what we did about 150-200 times at the conference, there's a better (but really tired) version coming later

lkcl
Offline
Joined: 03/29/2015
lkcl
Offline
Joined: 03/29/2015

yay we crossed $50,000!

Chris

I am a member!

Offline
Joined: 04/23/2011

Try two for hitting slashdot:

http://m.slashdot.org/submission/6157963

If you have a facebook/google/slashdot account login and up-vote it. The campaign isn't there, but could be. Still another $100 k needed.

lkcl
Offline
Joined: 03/29/2015
hack and hack
Offline
Joined: 04/02/2015

Quick question about the NAND/storage:
I've read above that there's a driver developed for it.
I was reading this thread (https://trisquel.info/en/forum/libre-storage-device), and I was wondering if the EOMA68 (with its included storage) qualifies as a libre storage device?

Also, why isn't Trisquel one of the suggested OS?
Not trying to troll or anything, just curious about the thought process.

onpon4
Offline
Joined: 05/30/2012

> Also, why isn't Trisquel one of the suggested OS?

Not possible. Trisquel doesn't support ARM.

Chris

I am a member!

Offline
Joined: 04/23/2011

onpon4 is correct. There is no ARM port of Trisquel as Ubuntu discontinued its ARM build back around 8.04 I believe. That makes it a tad bit difficult. It would probably make sense to do a Debian-based version of Trisquel as there are Debian ports for ARM.

The way NAND flash on these devices work is such that there is not a controller with proprietary firmware (this is how a SSD works, but really all modern drives including the spinning disk kind have proprietary firmware in there controllers). Thus with the on-board raw nand there is no firmware. This means extra steps are ideal to prevent killing the nand. There are what I'll refer to as wear levelling file systems to compensate even though they are rarely used in practice. This is in part probably why microsd/USB flash drives/and internal nand flash has a particularly high failure rate. Many SSD drives also have a high failure rate. However there is also something called SLC nand flash. Unlike MLC nand flash it doesn't have some of the problems that MLC nand flash has. It's also much more expensive and used in industrial environments where things can't fail. We're going to try and get SLC nand. We're hoping this combined with a wear levelling file system will increase the longevity of the on-board storage. If it ends up not being adequate enough for a laptop (and even if it is) we'll be investigating a USB to SATA adapter (or similar) so that you could still stick in a traditional hard drive. Many of the USB to SATA chips are garbage so we've got some work ahead of us. There supposedly are some decent ones though...

lkcl
Offline
Joined: 03/29/2015

argh there was a guy who did some research on the mailing list about USB-SSD drives a while ago, he found some reaaally good ones that had dual USB and SATA interfaces, those he said were superb.

one of the main problems with USB flash drive firmware is they do cacheing of the first few sectors because it's ASSUMED that you're gonna be doing a FAT filesystem, i mean who would want to do something other than a FAT filesystem on a removable USB stick, right? so you write to the beginning of the drive, nooo problem... by the time you get to the end it's often dead within a few writes.

... yah.... so anyway this guy did a really comprehensive review - i'll remember his name soon, began with "g"... it was about 3 years ago on the arm-netbook mailing list...

hack and hack
Offline
Joined: 04/02/2015

Thanks guys.

I don't understand much of what a lot of writes on disk involves exactly (changing location of files, things like that?), but looking for it on the web, I found this (https://askleo.com/can_a_usb_thumbdrive_wear_out/):
The best use of USB thumb drives and other flash memory-based devices is simply copy-to and copy-from. By that I mean copy the information to the thumbdrive to store it, copy it from the thumbdrive to a local hard disk to use it, and then copy it back to the thumb drive to store it.
Never run disk-intensive applications directly against files stored on the thumb drive.
If you copy to and from even ten times daily, you’re looking at three years of usage for the low end of the flash memory lifespan. (Yes, I know that’s not exact. In fact, it’s way more complex than that; factoring in things like the type of file system, FAT or NTFS, the efficiency of the device driver, and even the circuitry on the specific flash memory device – but it’s a good order of magnitude.)
You may also note that your application speeds up when you copy your database to the hard disk for use. While reading flash memory is typically quite fast, writing is not.

So the best strategy as a user for longevity would be to copy to and from the NAND, but not use applications that do this often (can't figure out examples. maybe intensive saves from a video editing software?)?

but since there's a possibility to have external memory cards, I suppose it's possible to use the 8GiB of NAND for anything that's not meant to move often.
Supposedly, reads wouldn't affect its longevity as much as copying to and from.

JadedCtrl
Offline
Joined: 08/11/2014

Maybe some of the OS-- parts of it that are hardly moved, but often read. (I assume that copying to and from is what kills NAND, like you said, and not reading from.)

hack and hack
Offline
Joined: 04/02/2015

If it's possible to separate parts of the OS that get written often on the SD-card, that would be really cool.
Of course, most of the 8Gib would remain unused, which would be a waste of real-estate. But at least, it would possibly keep it usable for a really long time.

I don't know for example how torrents sharing (downloads-uploads) would affect its longevity.

As for the SD-card, or a thumb drive, I have yet to see one fail. But of course, I never used one daily (for hours), for years.

JadedCtrl
Offline
Joined: 08/11/2014

It's possible-- they come with Parabola. Just modify the fstab.

JadedCtrl
Offline
Joined: 08/11/2014

It's possible-- they come with Parabola. Just modify the fstab.

hack and hack
Offline
Joined: 04/02/2015

I'll check it out, thanks.

onpon4
Offline
Joined: 05/30/2012

You can run an OS on flash memory, but you will have to replace it eventually; it's not about "failure rate" with flash memory so much as failure time, because all flash memory has a definite number of writes that can be performed on it. Intelligent wear leveling tries to maximize this number, but it is literally impossible to use the same flash memory forever, unlike a hard drive which does tend to fail after some time but theoretically can last forever in the right conditions.

Anyway, because of this, I'm planning on using a microSD card for the OS rather than the NAND, regardless of how reliable it is. I'll probably just use the NAND for a backup OS, or for extra storage. It's much easier to replace a microSD card than to replace the NAND. :) Besides, you can get more storage for the OS this way.

hack and hack
Offline
Joined: 04/02/2015

Exactly, I'd rather wear out a "disposable" and replaceable part rather than the part physically unmovable.

Also, the OS most will most likely occupy a small fraction of the available space. And considering the OS on the NAND as a backup OS is a really nice idea.
But thinking of the remaining space as extra storage kinda worries me, unless it's data that isn't meant to move, but only be accessed. Still, 10000 writes minimum is a whole lot, and if I end up using the NAND as much as I use the SD-card on my phone, I won't have to worry about it.
But one thing is for sure, the less the NAND is used, the better.

Wait, what if the NAND part wears out and becomes unuseable? Is an SD-card still useable?

onpon4
Offline
Joined: 05/30/2012

> Wait, what if the NAND part wears out and becomes unuseable? Is an SD-card still useable?

Yeah, based on what lkcl has said, support for booting from a microSD card is in the boot ROM. The NAND doesn't have to be used at all.

hack and hack
Offline
Joined: 04/02/2015

Cool, so the only downside being that it's not replaceable, then using it minimally should be the best.

and worst case, it's like there's never been any NAND. No big deal :)

Thanks for the info!