Next I bought an Efika MX Smarttop (https://www.genesi-usa.com/store/details/11), which works ok, but doesn't always boot reliably (I boot it headless most of the time, so can never get to the bottom of why).
I wish I could jailbreak my Apple TV 3, because it's cheap and the form factor is perfect for this. But it's looking like it will be more difficult to jailbreak than previous generations.
I just want something I can run Debian or Ubuntu on and know that when I restart it it's going to come back up. Any suggestions?
For me, a build bot has to be ultra powerful. The thing I'm usually building has 40k+ files and the linker dies unless I have more than 4gig of ram. This VIA or a Raspberry PI just isn't going to cut it as a build bot.
My library upb (https://github.com/haberman/upb/wiki). It is ~5k LoC and builds in 10 seconds on x86. 4gig of RAM is entirely unnecessary. It pays to be small. :)
Used Google cache to read the blog. This is what was written in the blog:
So, now it’s time to take it to the next level – I’ve paid for the beaglebone to go into a colocation rack in Telehouse North, with a friendly colocation company called Jump Networks who were happy to help out with the experiment, and who only charge for £50 + VAT for installation and a very low monthly cost for hosting the equipment, as little as £5 per month – perfect for an ARM server.
Sorry about that, the blog is hosted on linode, who had scheduled a reboot for tonight, which turned into a 2 hour outage...
The colo I picked only charges for electricity and excess bandwidth, which is why I picked them.
It cost £50 to install, and £12 a year in hosting costs for electricity and bandwidth, but the host I'm using says they'll have to introduce a minimum charge of about £5 per month if lots of people start doing it with these micro servers.
Except it doesn't have much memory, no sata/ide interface, a relatively slow CPU and a version of the ARM that isn't the same instruction set as most Android/Linux targets
Dell makes ARM machines? Do tell. I already have headless build bots for x86 (new Mac Mini) and PowerPC (old Mac Mini), which are perfect in form factor for this. I specifically want an ARM machine to add to them.
This is about the price-range I would like to see for a ChromeOS box.
Put it in a box I can hide behind a standard monitor, let me reuse the old mouse and keyboard and I can finally throw away the Windows XP PC my in-laws are using for Chrome and Freecell. And use 1/10th of the power.
I'd rather pay more for a high-end Cortex A15-based ChromeBox. Browsing will be very slow on an ARM11 chip, especially if you use it in "desktop mode", which will make it feel even slower compared to using it in a mobile phone. Also, it better have a good GPU, otherwise it won't even support resolutions higher than 800x480 (this one might).
But I do think ChromeOS devices should be somewhere in the $200 price range (or free with contract if you want LTE and plan on using it on the go).
In 1989 I ran a Unix system with 1000 users on hardware less powerful than this, why this kind of horse power cannot run something as simple as a web browser says more about the inefficiencies inherent in current operating systems and programming methods than they do about the hardware.
"why this kind of horse power cannot run something as simple as a web browser"
Web browsers aren't simple. Every web page is a complicated nest of sizing constraints and it seems like hardly a CSS property is added without making that worse. And then IE4 had to come along and make all of these sizing constraints dynamically changeable. And then we wanted fancier font rendering, so glyphs weren't free or even cheap anymore.
And I'm just talking text & tags here, not even remotely about anything like OpenGL or video, just core browser stuff.
You can render web pages on mere dribbles of power with Links/Lynx, but it's not some sort of mere coincidence that the result is a much degraded experience, it's actually fundamental to what a web browser is nowadays. I mean, look at this: http://www.youtube.com/watch?v=ZTnIxIA5KGw And that's a simple webpage.
> as simple as a web browser says more about the inefficiencies inherent in current operating systems and programming methods than they do about the hardware.
No. This is simply not true, at least at the extreme you are suggesting.
What I have open on my quad-core 8GB RAM desktop right now:
two Eclipse sessions; One emacs server, with about 20 client windows; Firefox, with 30+ tabs; 20 or so console sessions; five PDFs of documentation; an IRC client; an image viewer and a Jabber client. All of this is spread over two big monitors, with antialiased fonts, fast scrolling, lots of undo history and all sorts of good things.
The truth is that expectations have changed. Say all you want about having 1000 users on a single core, but users today are getting a much richer environment, and capabilities that we only could have dreamed of twenty years ago. This isn't waste or inefficiency, it's using what we have.
Web browsers are also far from simple. High resolution graphics, interactive sites, multiple format support, dynamic content loading, antialiased fonts and all these other things do add up. Compare that to what a user was doing on a tiny slice of a machine 20 years ago. I'd call it progress.
You can still run a lot of users on "modest" hardware. Dave Richards posts regularly about his terminal server work for Largo, Florida. 250 users running GNOME, firefox, libreoffice, etc. on one server with 64GB of RAM may not be up to 1989 standards, but it isn't shabby.
Agreed, just that image of the VIA board takes 730k. That's over 11 Apple IIs and most people have multiple tabs each with multiple images, streaming video players, iTunes, etc. running.
Heck, just a unicode font that includes every character in the basic plane takes 22Meg! Did that guys 1989 unix box display Chinese, Japanese and Korean?
Now go read the hardware compatibility list: http://www.menuetos.net/hwc.txt. It only apparently supports 4 network cards, 2 audio cards and a handful of video cards, and all of them are very old tech.
It's essentially a niche operating system, and isn't going to do 1/10th of the stuff that one written with "bloatware" is going to be capable of. I mean, look at what it has for a browser: http://www.menuetos.net/098b3.png
Oh, I'm sure this will run lynx with no problem at all! ;) It appears to me, though, nowadays most people prefer nice graphic interfaces, free-style layouts, visual and audio effects.
The whole multimedia thing, you know.
Also, this Android box costs and weighs a fraction of that wonderful machine you used to serve 1000 of users.
So I'd still call it a progress worth 25 years of technology evolution.
People should read your comment a few times to let it sink in.
But, iRobot, why should we care about efficiencies? Energy grows on trees.
I predict we are going to see many, many more ARM devices. And Apple is going to lead the way.
This is going to bother some people who like to write bloated, inefficient code. Their focus is on multi-core and concurrency. But to do simple things (retrieving text, images, audio, or video via http, sending/receiving email, etc), one does not need that much power.
The rise of ARM will create opportunity for a different set of programmers who are more efficiency-conscious.
You should always be wary if any binary blobs are required (often the case with video drivers). They will limit what you can do in terms of upgrading the kernel and hence the rest of the OS.
You're entirely right, but Wondermedia/VIA actually have a moderately good history with their GPL source for ARM system-on-a-chips.
Although most end vendors of devices using the WM8xxx line of chips are GPL violators (direct-from-China business models where noone cares), Wondermedia themselves have been supplying GPL source via Harald Welte of Gpl-Violations (formerly also employed by VIA):
100% blob-free to my knowledge. Unfortunately they haven't released anything for the WM87xx series (apc.io) yet, but these devices are brand new so here's hoping.
If you're interested you can dig up some more details, and the follow from-scratch open kernel development efforts, at
My vision for these boards is not to display graphics, but to send/receive data from the internet and a local network. I'm not going to use a Pi to watch movies or some other resource-intensive task. I'd just as soon use the power of GPU that comes with Pi for some other task besides video.
What would be interesting is if one of these boards would be designed such that, by pure coincidence, it could fit into an Apple form factor (cheap/old/maybe used). I've read that in, e.g., Vietnam, people with soldering irons do all sorts of hackish things to iPhones.
Or maybe a market for curved edge form factor casings develops. Maybe it already exists. But I never saw any Apple-like form factors in the mini-ITX offerings.
Whenever you see something with a cool form factor, it seems it's always a proprietary package, hermetically sealed, not easy to tinker with.
The Raspberry Pi starts up by loading an opaque blob into the GPU, and that then runs the CPU under supervision. ie the CPU is a servant and controlled by the GPU, not the other way around as is normal in the PC world.
So irrespective of your video intentions, you are still at the mercy of a blob.
But unless your are a real FOSS zealot (say > 500milli Stallmans) you are at the mercy of the bios, the disk controller firmware and the CPU microcode on any machine
In reality being able to run whatever user programs you want in whatever combination you want without artificial is what FOSS is all about.
The "user programs" you want to run are determined by the operating system. And the operating system is determined by blobs, controllers, BIOS and similar gunk.
If for example the Pi blob is such that Linux kernel 3.7 can't run then you are SOL. Or if on a device like this their blob only works with Android 2.3 then you can't run a different version. Or maybe you can't run one of the BSDs.
The video/GPU is especially relevant in the Pi case because it controls what the CPU can do.
Yes that's true - I was thinking of people who complain they don't have details for some deep detail of the GPU shader cores to write an opensource driver for their NVidia card.
What's more exciting than this computer itself is the potential it represents. If you read the actual product page [apc.io], a lot of it is devoted to just how small this new Neo-ITX form factor is.
For comparison's sake:
Mass-market paperback - 19.8cm x 13cm
Neo-ITX - 17cm x 8.5cm
IPhone 4 - 11.5 cm x 5.86 cm
2.5inch SSD - 10 cm x 6.99 cm
Raspberry Pi - 8.6cm x 5.4cm
So the Neo-ITX is a third bigger than an IPhone 4.[1] And the perfect size to nestle a normal SSD right on top.
Not very far down the road, we will be looking at a complete, very functional computer (harddrive, wireless internet, etc.) in a package about the size of a small book. All for 15 watts.
And what I love best: it will be so cheap and versatile. Because it will use modular, already popular hardware.
Remember the good old days of the desktop:
Swap out the HDD for a bigger one. Replace it easily if it fails. Swap out the whole motherboard if that fails. Keep using the same damn case and power adaptor for ten years.
And best: put in random, new PCI express cards that expand the capabilities of the computer, using it in ways the original designers hadn't foreseen. As happened in the past with:
Ethernet
Wireless Networking
Exploding GPU power
[1] I initially used a deck of cards for comparison. But I couldn't visualize exactly how big a deck of cards is. But an iPhone...
So many comments and no-one could manage any praise?
It matches the real-life cost of the Rpi, probably will have general availability around the same time, and has double the memory and a little onboard NAND.
It may not be perfect, but remember this sector is very much in its infancy, and a little more competition has to be a good thing for improvements in future devices all round.
Finally, a cheap ATX-compatible ARM board. Give me PCIe and DIMM slots and I'll be salivating. Equip it with something like a dual-core Cortex chip and I'll be throwing wads of cash in all directions.
Why not just buy any number of Atom or Fusion motherboards out there? Cost is in the same realm as what you're asking for and they already have those slots. Power consumption isn't much more than the ARMs, but there's actual product you can buy today.
VIA: if you're reading this, your address help@apc.io is busted :-(
The original email I sent was:
Hi,
I'm really excited to see the $49 Android PC; it looks like a perfect
platform for a number of projects I've been thinking of.
However I'm worried - most manufacturers offering low-price Android
devices fail to comply with the terms of the GPL.
Will you be releasing the source code to any GPL'd components used by
your system? And will you allow customers to install their own OSs on
the device or will it be restricted to your own build of Android?
Thanks for your time; I look forwards to your reply.
Same price bracket, faster Cortex A8 based processor, and (most importantly) marketed as "hackable" with available source & tech docs, and a community working on porting other OSes to it.
Even if you don't want to hack on it yourself, this means you're more likely to find interesting uses and software updates for it down the line.
IMHO these are the aspects VIA should be aiming to compete on as well, so it's not stuck with a crummy vendorware version of Android.
The Melee boards are available to ship now from Aliexpress vendors, so you'd have it in your hands sooner than the apc.io (July predicted ship date) or the Pi (still filling backorders.)
It's more complicated than that, though. You can't just download ICS and drop it on the board. There are drivers to port and middleware integration to do. Someone has to do that. Google did it for OMAP4 (the Galaxy Nexus), but that's it. This isn't an OMAP4 board, and $50 a unit doesn't pay for a lot of software integration work.
This is where the ARM ecosystem tends to fall down. The PC World is built on compatibility. No one can ship a board if it doesn't run Windows (or even DOS, frankly). Graphics cards have VESA and VGA fallback modes so that you can bootstrap a driver installation. And as a result the Linux community can leverage this to provide pretty great support for new hardware, even if it starts out as a fallback or partial implementation for a few versions. None of that exists in the SoC world.
So if VIA or NVIDIA or Samsung or Qualcomm want ICS to run on their chips, they need to do the work. So far they have not. Nor have their customers been willing or able to.
Canonical and others in the Linux community are working on making a universal kernel for ARM SoC's, and they hope it will work on most of them by 2014. That might help with upgrades and compatibility in the Android world, too, especially since they are planning to merge the Android kernel back into the main Linux kernel.
I agree with your comment entirely (esp. the PC/ARM bit.)
Also, the WM8750 is an ARM11 not a Cortex A8 or better, so it's a good bit slower than most phones running ICS. I don't think its GPU credentials are cutting edge either.
So in terms of raw hardware specs it'll come in alongside the flagship phones released with Android 1.6 and 2.1. ICS may not be a good candidate even if/when the development is done.
(Which is insane given that it's still objectively powerful hardware, but even more powerful hardware breeds equivalently consuming software.)
I hope they also have linux distributions for it other than android. If it works with arduino, these could lead to very inexpensive analog/digital controllers.
It depends on what you intend to do with it. My mom's computer is due for an upgrade and she would probably be happy with a beefier version of this machine. Swapping the motherboard should be trivial enough and the PC would continue to look familiar, at least.
She's used to Ubuntu and she couldn't care less about the ISA the machine is running. Moving to Android could prove an interesting experience, but I suspect machines like this will have outstanding support for other Linuxes as well and I'd assume Via has a lot to gain by cooperating.
By now, most likely, someone from Microsoft will have called to offer some incentive if Via favors Win8 over Android on the platform. We'll see what happens.
I'm willing to overlook the geek appeal for the low price. I can see myself building a small NAS or media center with this.
I wonder how difficult it could be to install an ARM version of Linux on it. I'm not that interested in the Android aspect of it, besides maybe using it as a cheap tablet replacement (they'd need to upgrade the Android version though.)
The main issue of (amost) ALL these boards is the code. They always use a large part of proprietary bobs, which makes using them a pain.
Even the Raspberry pi has the issue. In particular, if you've a recent GPU you're often doomed. Want video accel? Nope. Proper video support? Update to more recent libraries? Nope again.
Because having both a DVI and a HDMI port is redundant. HDMI is backwards compatible with DVI so you can convert DVI <-> HDMI easily, but not VGA which is analog.
Everyone has different levels for what "expensive" means. You can get new HDMI monitors for well under $150, and new DVI monitors for well under $100, and DVI<->HDMI adaptors are under $10.
Whether "it's expensive" depends entirely on the market. There are markets (developing countries, education) where tripling the total cost of the setup, compared to using a monitor you already have, is significant.
That's not true if you use a pre-certified wifi module. Of course such modules cost half as much as the board, largely because they eliminate the cost of intentional radiator testing. But I agree, USB wifi is so cheap these days it makes no sense to have it built-in.
For some unknown reason all those sensor terminals, atms and other smart devices are powered by windows.
Windows is an awful choice since it requires a huge pricey box with fans, since it shows its dialog messages over the interface, catches viruses and yet they stick to it.
It makes me sad thinking about how much money do they waste on it and how MS is able by get their cut while making everyone lives and products worse by using their BS power.
Windows has been available on embedded systems from almost the very beginning. Windows CE was not originally for tablets, it was for ATMs and the like.
It uses solid state storage and the RAM, graphics system, etc. is bundled into the single System-on-a-chip. Plug in the power then hook it up to keyboard/mouse/monitor and you are up and running.
OK, fair enough. I don't think the lack of a case is a huge deal (zip tie it inside a tiny cardboard box, whatever) but the requirement for a special power supply is significant. In comparison, a raspberry pi can be booted up and used with parts that a typical geek has lying around (micro-usb charger, usb periphs, monitor).
Edit: it looks like it comes with a power adapter, so it's effectively a fully functional computer out of the box (just as much as any system without kb/monitor).
First I bought a SheevaPlug (http://en.wikipedia.org/wiki/SheevaPlug) but that fried itself before long.
Next I bought an Efika MX Smarttop (https://www.genesi-usa.com/store/details/11), which works ok, but doesn't always boot reliably (I boot it headless most of the time, so can never get to the bottom of why).
I wish I could jailbreak my Apple TV 3, because it's cheap and the form factor is perfect for this. But it's looking like it will be more difficult to jailbreak than previous generations.
I just want something I can run Debian or Ubuntu on and know that when I restart it it's going to come back up. Any suggestions?