Also, plug computers like FreedomBox: http://en.wikipedia.org/wiki/Plug_computer

And single board microcontrollers like the arduino boards, make controller, basic stamp, etc.: http://en.wikipedia.org/wiki/Single-board_microcontroller

I personally would like to see a build-your-own tablet type of thing that can run android or ubuntu. Of course there already is the nexus 7 though.

There is no single word for the horrible, utterly disgusting, nasty setup that these things required. There is some sort of tokenizer for Linux that kind of makes the required upload. The GUI is BAD. And when they had serial interfaces, you had to buy a USB-RS232 from them because they did screwy shit.

Thank the gods for the Arduino. Working with that cheap microcontroller is pleasant.

I agree that by today's standards, programming via serial port is cumbersome, but it's actually a good thing that the STAMPs are still available because they are used in a _ton_ of embedded applications that are still in use and need maintenance and replacement parts.

My dad had a few of these. One I distinctly remember was a 2 part combo. To write, you put the chip in a ZIF socket (he changed this from a socketed to ZIF) in a huge metal box, and what looks like a parallel port to the computer. Then, you realize that wasnt a parallel socket, but something going to an 8bit ISA card.

To erase, you had to untape the window and put the chip in a UV box for a few hours.

I've still got it around here, somewhere...

http://en.wikipedia.org/wiki/EPROM

There was plenty of benefit to EPROMs back in the day to a developer, but long-term data retention was certainly not one of them.

The thing I wonder for things like this, though, is at what quantity I can get them, and what the reliability looks like. One of the things I like about the Pi is that it's well tested - Even if it's a bit slower, they're making 4K of them per day, and everyone's using one, helping map out the gotchas, etc.. If you go with a lesser known board, you're a lot more on your own..

The rpi is far more than a bit slower, more than the cores and clock counts imply. The pi is a slower microarch and is cache deprived too.

For somethings this doesn't matter but many of those things should probably be done on a proper microcontroller.

The odroid stuff appears to have a lot of users too, and these devices are less 'weird' than the rpi— they're more like the pandaboard and beagleboard— so there are less sharp edges to figure out.

I assume that, because in that case - they'd be fanfaring a lot more about how their newer design is being worked on. I'm seeing that they're more focusing on developing the userbase and utilities of the current board.

ie. with expansion cards, refining the software stack and such.

Something like this with SATA and Ethernet on a fast bus would be awesome

(FWIW: that apple-branded adapter is almost certainly an ASIX part -- the same chipset is sold under dozens of brands and works quite well in my experience.)

This one caught my eye. I'm not sure if the i.MX 6 quad is as capable as the exynos and there's only 1GB of RAM, but the board has SATA (there's a controller built into the SoC AFAICT) and Gigabit ethernet on an RGMII interface. Not that I really know what that means, other than I'm guessing a dedicated, gigabit-capable interface on the SoC. There's also PCIe which they're apparently going to release a breakout daughter-board for at some point.

Of course this board is not all that cheap compared to the ODROID.

EDIT: this was it, the Ubiquity Networks RouterStation Pro http://www.ubnt.com/rspro ; it was the basis of an official "MIPS Linux Starter Kit" http://store.mips-store.com/mips-linux-starter-kit.html which is now "[t]emporarily out of stock". 1 × USB2, 128MB RAM, JTAG, no video. $79 USD MSRP.

http://www.8devices.com/product/3/carambola

Runs OpenWRT

I received one from the indiegogo campaign, haven't played with it much yet though.

After that, it's just like every other rootfs. I focus on Gentoo, so other distros... I don't know, but they will have an Ubuntu image up soon.

There are no graphics until X though, so you're stuck with ssh or serial console until then.

Kernel sources come from the BSP download on hardkernel's site. Should be a tarball in there named kernel_4412.tar.gz

HD decoding on these boards is performed using a digital signal processor (DSP) integrated into the SOC, the CPU (even with GPU support) is too slow for 1080P h264 Content. This is not specific to Android, for example TI supports use of the DSP of the Pandaboard on Linux as well [2] (it is still not very stable, at least on the OMAP 4430 version of the Pandaboard).

[1] http://airlied.livejournal.com/76383.html

[2] https://launchpad.net/~tiomap-dev/+archive/release

Edit: formatting

I know this works on pandaboard and some others, but those boards are about 150$ range if i remember correctly.

There is an unofficial XBMC port for A1X devices which supports the HW Video Decoder of the SOC [2]. I have an A10 device at work, but unfortunately won't be in the office to play around with it for some time.

[1] http://linux-sunxi.org/Mali400

[2] http://linux-sunxi.org/XBMC

According to what XBMC devs say - it doesnt work at this point, and there is some online drama going on ;-)

Although i would be super happy to see progress on this.

One of the links you gave me: "Allwinner officially only supports the Mali GPU driver on the Android platform only." :(

I should add that with DMA and checksum offloading, even a (relatively) slow CPU can handle 1Gbps data transmission rate easily.

brigade: thanks for the board tips!

I'm a designer who also programs, so don't fully understand why this is cool. Can someone explain?

Just six years ago, a computer this fast would probably have cost in excess of $1000 and have been at least a minitower drawing 100 watts.

But we knew we were in for the ride when we bought rPi's.

This is 1.7 GHz, quad core, 2048MB RAM.

If anything, it's more efficient per watt than RPi (and per dollar).

Like anything, you've got to pick the right tool for the job.

[1]: http://www.linuxfromscratch.org

Power consumption is the issue though, even 3W is going to tank our battery life. I've looked around, but can anyone come up with something that is better in terms of a 'processing power to power consumption' ratio?

EDIT: Also, did anyone else notice the full credit-card number on the picture...? I hope they fix that. :(

http://www.8devices.com/product/3/carambola

Stop being snarky for the sake of snark, please. Stick to the facts.

It's like the AMLogic-M3 f16ref. There are a dozen different flavors of these and the roms are interchangeable but there's no community around them to get continuous integration and good builds of XBMC setup. There should be an effort to get a distro that will generate nightly builds to target these devices with Linux/Android and optionally XBMC.

I keep rehashing this and everytime I do I say I should start a blog or set of resources for this... I need to follow through but my other projects are more immediately exciting :S

Nope there aren't, not at this price. Even the pandaboard ES is more expensive (and TI is discontinuing OMAP :-(( ). Sure there are quite a lot Allwinner and Rockchip based boards, but these don't compare to a Samsung Exynos. Not yet.

http://www.phytec.com/pdf/OMAP5_product_brief.pdf

I think things are going to spice up for the AMLogic-M{3,6} devices very soon from j1nx.nl.

Given this is infinitely cheaper than using general purpose cores, why wouldn't you just build yourself an army of cortex m3/m4's and use them for video encoding now.

ARM doesn't really need to do anything here to make this happen. Since they give you all the chip related stuff necessary to build your own, just go and do it. If you want 64 m3's, go make it.

It's also not clear whether it's a SIMD solution. If that's the case, they are basically reinventing GPUs.

It is not SIMD. Each core is totally independent of the others. Basically the cores are structured in a grid, and all cores have 32KB in-core static RAM, but can also access the memory of all the other cores (at the cost of a few cycles delay), or main memory.

You could certainly use them as much like a SIMD if you wanted to, but then you'd likely get better performance out of a GPU.

We'll see when they deliver on their kickstarter project what type of performance people get out of it in reality...

Applied Micro also intends to introduce a 128 core SoC in 2014:

http://www.apm.com/products/x-gene

Also, keyframes are not known from the start in many encodings. A keyframe is inserted when the difference between a frame and the next one is big; this can only be detected when encoding, or with a multi-pass encoding.

Video encoding is CPU-bound, not I/O bound.