Micro computers
August 9, 2013 1:02 PM   Subscribe

I have a fascination for micro computers such as the Raspbery Pi, the wide array of micro Android PCs, and low-power wall plug computers.

Can you help me find:
1. Reviews
2. A reference for various categories that would allow me to see the most powerful/featurful items currently on the market
3. Specialty retailers that can supply these in the American market from overseas, if they haven't been released here yet

Very much interested in #2 and #3. It's easy to find announcements for or reviews for disperate models, hard to compare features, performance, and price for products actually available today to me.
posted by jsturgill to Computers & Internet (14 answers total) 26 users marked this as a favorite
Adafruit has been my "goto" place for these types of things....although more on the microcontroller/electronics spectrum. (outside of going directly to Newark/Element14). The forums there are also fairly active.
posted by samsara at 1:50 PM on August 9, 2013 [1 favorite]

Adafruit and Sparkfun are popular, SeeedStudio.com and Tindie might also be worth looking into.
posted by RobotVoodooPower at 1:55 PM on August 9, 2013

Check out the chromecast dongle...they're apparently flying off the shelves...not sure how customizable they are, but hey $35.
You might want to check out engadget.com...tho they mostly are concerned with phones and computers, they always give good coverage to the microcomputer/modding/maker/3d printing/wearable/etc scene...
posted by sexyrobot at 2:18 PM on August 9, 2013

liliputing is a good source. This stuff changes so fast and it's so cheap that reviews are nice but not all that useful. I'd be looking at either the ultra cheap processors from TI or the Google TV dongles, depending on what I wanted to do. Chromecast hasn'[t been broken yet and for a few dollars more you can get something that has an active community already established.
posted by rdr at 2:44 PM on August 9, 2013

This might be a good place to start, though I could've sworn I came across a better comparison matrix recently.

This sector is kind of hard to keep track of since most people generally pick one product and stick with it for a while, so the broad perspective is rare.
posted by Standard Orange at 2:49 PM on August 9, 2013

If you want to track what's new in this area, try linuxgizmos.com. (Same editor as the defunct linuxdevices.com)

It's more for the embedded systems engineer, but this is where the tech starts before it trickles down to this new "consumer" area of devices like beagle, panda, and RPi.
posted by JoeZydeco at 3:22 PM on August 9, 2013 [1 favorite]

I check cnx-software.com a couple of times a week.

I'll ask a question I've asked myself: How much time is it worth spending finding the "best" device in a category where price-points are typically below $100, almost never above $150, and where the rate of change happens at fantastic clip?
posted by Good Brain at 5:11 PM on August 9, 2013

Response by poster: Good Brain, I'm assuming that price points are relatively stable, and new products are coming out quickly. The difference between a $75 thing released 18 months ago and a new $99 thing might be an extra core or two, much better graphics capabilities, twice the RAM... whatever.

That seems worth keeping up with, no? I mean, even if it's just by going to a store and sorting the right category by price (or RAM or by selecting a specific processor to filter by). I'm hoping to find some tools or a retailer that can make it dead-easy to see what's out there, so that it's not a time sink.
posted by jsturgill at 6:58 PM on August 9, 2013

Lightake is a Hong Kong drop-shipper. I used them once to buy a cheap Android tablet with good results. They currently have only one Android mini-PC, but it's $32 + free shipping.
posted by qxntpqbbbqxl at 12:02 AM on August 10, 2013

Just a bit of experential advice from my years of developing with these SoC devices: you can never keep up with it. Chipmakers have roadmaps that go 5+ years out from here and are always making advances. I'm about to start a design using a Beagle Black as a prototyping board, which is OMAP4, and OMAP5 boards are already trickling down into the cheap EVK space.

I honestly believe the reality is that any recent board you pick up will be powerful enough to do whatever you want, short of trying to run a desktop or play full-fledged Flash movies. A dual or quad A9 or A15 can do a *hell* of a lot.
posted by JoeZydeco at 12:40 AM on August 10, 2013

I'll throw in some orthogonality here, though it sounds like JoeZ and Good Brain started down this path:

The "most powerful" board is usually not measured in MIPS or megs. It's more about the support and documentation, the examples, the OS updates, the community around the board. Also, if you're trying to build a product, you need something with guaranteed availability for years to come. The field gets really narrow when you start looking for that.

If you find a spec-perfect board with poor support, by the time you make the whole toolchain and get the thing doing what you want, it'll already be halfway obsolete. That equation turns out differently if you have a whole development team, but as an individual, there's value in following at least some crowd.
posted by Myself at 11:46 AM on August 12, 2013

Following up:

I guess I'm reasonably clear on what you are looking for, but I don't have a great handle on why, other than some reading between the lines and inference.

I've spent a little time looking for a resource that sounds like what you are asking for, and I've spent even more time looking at and for things that might bring me in contact with such a resource, and I honestly haven't found much that is both comprehensive and up-to-date. You might look around at places like SlateDroid, XDA-Dev Forums, and ArmTVTech, to name just a few. It seems like from time to time someone makes an effort to compile a list of detailed specs. What seems to happen next is that they either loose interest, or the move on to the real work of trying to get a device to do what they want it to do.

I'll share some insights I've had that might help you in various ways:
  • For anyone older than about, say, 20, mental models about computing devices were either formed in an area where personal computing capacity was relatively rare and expensive, or strongly influenced by people with that background. I think that background can lead to a tendency to obsess about the cheap, plentiful personal computing devices available to us today.
  • The rate of progress and apparent variety in cheap ARM based devices is truly dizzying until you take a few steps back and consider a few things...
  • Almost everyone making the devices themselves are shopping for components from the same suppliers, assembling them with the same tools and labor pools, and selling them in the same places, which means that they all have largely the same influences on pricing.
  • The SoCs are the core of these devices are highly-integrated, which means that the main decisions the device makers are making is: Which SoC, followed by how much RAM, Flash (generally, as little as possible) network chip, form factor and IO capabilities (almost always shitty: USB2 and 100Mbps ethernet or single stream, 2.4GHz 802.11n at best). What's more, the combination of economic pressures and technical limits mean that almost everyone ends up making similar tradeoffs.
  • The SoCs at the core of these devices are themselves largely built with off-the-shelf building blocks (aka IP) sourced from a limited number of IP makers and built by a limited number of merchant foundries with relatively similar fabrication capabilities. Most likely it is going to be two to four ARM Cortex A9, A7 or A12 cores. Higher end devices might have A15 cores and some of those will include some lower-power A7 or A5 cores too, but that is a power consideration issue that is most significant for battery power devices. Don't over assign significance to the particular generation of the core design. For example, A12 is a successor to the A9, but it isn't necessarily "better" in any way that is going to matter to the end user.
  • Those cores are paired with a GPU. I think there are one or two ARM GPU designs that get used, in combinations of 1-4 cores, with PowerVR and Vivante making a limited appearance. Again, larger forces tend to drive the SoC specs into a few relatively narrow bands. There are other implementation details that might matter, but they aren't going to be obvious to an end user, unless they are really glaring.
  • Most of these SoCs are targeted at Android, made by companies that have very limited software chops, and sold to device makers who themselves have limited software capabilities and little or no desire to comply with the GPL, which means that the software is often shitty and opensource enthusiasts have to do a lot of reverse engineering and scavenging of SDK parts and drivers to get what they need to make and release improved software, all of which take time to sort out (if they ever do).
  • There are a few exceptions. For example the Freescale iMX6 has found its way into some cheap devices, and Freescale actually provides decent documentation and linux drivers. Plus, all the chips in the family support Gigabit Ethernet and some support SATA (though those capabilities aren't always exposed in devices using the SoCs)
  • There is a good chance that the devices with the most compelling hardware specs for the price won't give you the best available value because the software support is lagging, perhaps by 6 months or more.
As to what to do with any of this information, that comes down to what your goals are. It is a lot easier to find your way to an optimal solution if you are clear about the problem you want to solve. By way of example, if your main goal is something you can use for inexpensive fast network storage, you best bet is probably something like a ZyXEL NSA320. The CPU is relatively slow compared to something with a dual or Quad Cortex A9 or A12, and it only has 512MB of memory, but it has SATA and Gigabit Ethernet, which are the real bottlenecks, and people have done the work of getting Debian running on it. If you want to run XBMC, I don't know what to suggest, but GPU and video decode acceleration driver support is going to be at least as important as the RAM, or the type, number and top speed of the of CPU cores, or even the number and type of GPU cores.
posted by Good Brain at 2:44 PM on August 16, 2013

Response by poster: There's a lot of great knowledge shared in this thread. Thanks everyone.

The two use-cases you mention--networked file server and XBMC--are both examples of the kind of thing I might want to purchase one of these small boxes for. I don't have the engineering chops to do more interesting projects.

The ZyXel NSA320 sounds like a known-good set of hardware... if you're in the know. I'm not, and I don't think I have the time necessary to get there.

How would I have found that information out if you hadn't told me? Probably hours of trolling forum archives on sites I've never heard of and don't ordinarily visit. Possibly days, depending.

That's a huge time sink and not really a fun addition to any project I might want to put together.
posted by jsturgill at 3:54 PM on August 16, 2013

Just buy products that already do what you want and that you can return if they fall short of expectations.

Spending hours or days digging up information is par for the course if you want to use this stuff as something other than what was intended. I just finished reading a thread on the XBMC Forums, about the Little Black Box, a Linux device that ships with XBMC installed. It is based on cheap Chinese ARM "MiniPC" hardware that first debuted a year ago and it still sounds like it could be a project.
posted by Good Brain at 12:07 AM on August 19, 2013

« Older what should I give her?   |   Help me land a job as a barista. Newer »
This thread is closed to new comments.