Parallella Main-board (via Adapteva & Kickstarter)
The company Adapteva is currently testing one of their last prototypes of a tiny parallel computer board they are calling Parallella, which you are surely to hear about. This mini computer has reached a speed of 90 GFLOPS using just 5 watts urging some to call this a mini supercomputer. Kickstarter supporters are on pace to get their Parallella during the summer this year.
The main principles behind the Parallella mission are to have open access and share open source information on parallel processing while offer the board at an affordable price and “democratizing the access to parallel supercomputing”. The company sees processing in parallel as the future of computing and many would agree.
At the core of the device is the energy efficient Epiphany multi-core accelerators produced by Adapteva. The “45 GHz model” (The performance of a system at that speed) has a 64-core accelerator tied to a dual-core ARM Cortex-A9, 1 GB of RAM, microSD slot, two USB 2.0 ports, 10/100/1000 Ethernet port and an HDMI port. All of this is packed in a dev board that fits in your pocket comfortably. Another model uses a 13GHz 16-core accelerator capable of 26 GFLOPS.
Concept of the Main-board Parallella connected to the Epiphany multi-core accelerator 64 core edition (via Adapteva)
The Epiphany multi-core chips are composed of an array of reduced instruction set computing (RISC) chips, whose simple design allows for faster processes. This scalable array is programmed using C or C++ and they communicate between each other using a network embedded in the chips but share the one memory architecture. The Parallella operating system will be Ubuntu 12.04 and the team’s next goal, before release, is getting Linux to boot on the device.
The board hardware schematics and the SDK will be released as open source information and Adapteva has already made their device drivers available to the public. They are doing this because their target audience is developers. The faster and more widespread this information becomes, the sooner we can all benefit from the advantages or parallel processing.
The uses for this type of devices are pretty endless. Extreme multitasking and performing all sorts of computational tasks will be much more efficient in parallel than using common linear processing. Real-time analysis of markets, registering fingerprints and recognizing faces could be done with this tiny credit card sized computer board. Encryption and code breaking, efficient multi-media storage and playback, handling multiple real-time streams, keeping track of multi-sensor robotics or
tracking objects could be done in a cost effective, energy efficient and timely manner.
The time is now to start thinking more simultaneously. The Parallella development board will be available, in its 16-core form first, this summer.
C
See more news at:
