Disclosure: Some links on this page are monetized by the Skimlinks, Amazon, Rakuten Advertising, and eBay, affiliate programs. All prices are subject to change, and this article only reflects the prices available at time of publication.
Want to build a so-called “supercomputer,” but don’t want to spend a ridiculous amount of money? No problem. Now there’s an option that costs just $99. Sort of.
Adapteva’s Parallella is a $99 system-on-a-board that features an ARM Cortex-A9 dual-core processor, 1GB of RAM, Ethernet, HDMI, and Ubuntu Linux. Sounds kind of neat for a $99 device, but not all that super.
But here’s the nifty thing: as the name suggests, Parallella is designed for parallel computing and the heavy lifting is handled by an Epiphany multicore accelerator. So you can buy 2 units, a dozen units, or more and use them to perform some pretty serious computing tasks. Ultimate the team hopes to enable support for 1024 processor cores to work together.
Adapteva ran a crowd-funding campaign last year, and the first boards are nos whipping to backers. If you didn’t get in on original campaign, you can also now pre-order units that’ll ship in October.
Here’s a roundup of tech news from around the web.
- Adapteva’s Parallella $99 “supercomputer” module now available for purchase
After a successful Kickstarter campaign last year, the Parallela single-board computer module is now shipping. Buy one and you’ve got a simple Linux-compatible computer. Buy a whole bunch and arrange them for parallel processing and you can build your own “supercomputer.” [Adapteva] - Next-gen Asus PadFone (Infinity) may feature a Qualcomm Snapdragon 800 CPU
Asus offers a line of Padfone products that blur the lines between a phone and tablet. The basic unit is a smartphone, but you can tuck it into a tablet dock to use it as the brains of a 10 inch tablet with a bigger screen and larger battery. Online benchmark results suggest that the next model will be called the PadFone Infinity, and it could have Qualcomm’s speedy Snapdragon 800 quad-core processor. [Ringer’s Blue Men] - Ramos is launching a new tablet August 8th
Chinese device maker Ramos has earned a reputation for making cheap, halfway decent Android tablets. The company’s next model is coming in a few weeks. [Padhz] - Canary is crowd-funding a simple, versatile home security device
Canary wants to offer a small device with multiple sensors that you can set up in your house and access from a mobile device. You can use it as a baby monitor, security system, internet-connected camera for watching your pets while you’re away from home, or much more. It has a camera, mic, motion sensor, temperature sensor, air quality sensor, and more. Early backers can try to snag one for $199. [Indiegogo] - SwiftKey Cloud beta syncs your dictionary, other data across devices
SwiftKey is a keyboard app for Android that offers powerful text prediction technology to make entering words and phrases a lot faster. But if you have multiple Android devices, up until now you’ve had to train them each independently. Now you can sync your personal data across multiple devices using SwiftKey’s new beta. [Android Police] - Benchmarks show Mediatek’s upcoming octa-core chip is super-speedy
MediaTek may focus on low-cost chips. But the company’s upcoming MT6592 ARM Cortex-A7 octa-core processor is no slouch in the performance department. Benchmarks show it may not be much slower than Qualcomm’s Snapdragon 800 chip. [Tech2.hu] - Rumor: Samsung working on a 12.2 inch tablet
Samsung has a habit of releasing phones and tablets with every conceivable screen size. So maybe it’s not a question of if the company’s working on a 12.2 inch model, but rather when it will be released. [SamMobile] - Apple earnings report: Revenue up a bit, profit down a bit, iPod sales way down
Apple reported its quarterly earnings today, and year-over-year results show that revenues are up about $300 million from the same quarter last year, profits are down nearly $2 billion, and iPod sales are down 32 percent. You can probably thank growing iPhone sales for part of that, but the increased price of even entry-level iPod touch models probably doesn’t help things either. [Engadget + some informative charts at TechCrunch] - Using Miracast wireless display tech on the Google Nexus 7 tablet
Independent developers have figured out how to add support for Miracast and screensharing to the Nexus 7. This video shows a Nexus 7 acting as a secondary display for a Samsung Galaxy S3 smartphone. Theoretically it should also be able to stream content from the tablet to a TV that supports Miracast. [reddit/YouTube]
You can keep up on the latest headlines by following Liliputing on Facebook, Google+ and Twitter.
Why not use GPUs? You can get hundreds or thousands of cores already on one board, for less money than this will take.
Power consumption.
nos whipping, sounds awesome. 😉
I think these units would be good for Primecoin mining.
For interacting with the physical world, I think the Parallela will be miserably slow. The ARM architecture is hobbled by on-die busses and processes that deal with them; akin to the AMBA bus. These slow GPIO speed down to a crawl. So you might be able to calculate prime numbers or render graphics pretty fast, but trying to interact with the physical world is going to be painfully sloooow. For example, the likes of a GHz-scale ARM core is much slower than a RISC micro-controller chip clocked well below 100 MHz. Yeah, It’s that bad.
What about the FPGA currently used for HDMI output but that can IIRC reassigned to do other things ?
BOINC for Android looks like a far more interesting development in the arena of bulk computation.
https://newscenter.berkeley.edu/2013/07/22/new-app-puts-idle-smartphones-to-work-for-science/
The Arm CPU running Linux on the Parallella isn’t the important part, it just loads up the real parallel computing processor that lives on an entirely different chip and manages the system. The high speed node interconnects are between the Parallella chips on the stack of boards. The thing is a supercharged DSP engine intended to do massive computation. Problem is each individual node is pretty pathetic, more like the DSP/SPU nodes on a Cell processor than what most people think of as CPUs. So if your job can fit into the severe resource constraints on each node it is going to be a good idea, if not it will disappoint so read the specs carefully.