Quickest way I can get a 64 core Parallela board?

[joeish80829]

I would like to buy a board and any extra accessories it takes to give it all the functionality of my current PC which is like any other 2 year old PC. My current pc has an ATI HD4650 video card, not sure what the on board video of a parallela board is like but if someone could give me an idea that would be great. The board would need to be able to run Ubuntu have a mouse, keyboard, display and webcam attached. I would like to connect my speakers via a 1/8 inch cable so I would need to know what to get to do that too. Basically I would like to have a 90 gflop pc using a Parallela board I can expand with more boards in the future so I can use it for my computer vision projects and get mind blowing coputing power.. I would like to get this as soon as possible so any advice on that would be great e.g. do I need to donate asap etc..I would also like to know how much the Parallela 64 will cost. Thanks in advance for any help on this.


Edit. I just saw here:

https://www.kickstarter.com/projects/ad ... yone/posts

This:

Pledge $199 or more

903 backers Limited (64633 left of 65536)

64-CORE: You get everything in the SUPPORTER reward and a 64-core Epiphany-IV based Parallella board (2 ARM cores and 64 Epiphany cores) if we reach our stretch goal of $3M. If we don't reach our stretch goal but meet our minimum funding goal, you will receive a pair of Epiphany III boards (2 ARM cores and 16 Epiphany cores). Both boards come loaded with the development software needed to create your own innovative projects. (International order should add $20 to the pledge amount).

I would like to be 100% that I could get the 64 core board though be cause I assume the 16 core would be 1/4 the computing power. From the above link I think, and correct me if I am wrong, the 16 cores would be $99 and the 64 core would be $199 (a better deal). I'm so into super fast computing I would like to start building cheap Paralella based PC's for my customers so I can fund me obtaining alot of boards to build my own supercomputer. That being my end goal.

[9600]

[joeish80829]

Thanks for getting back to me so soon:), I'm thinking of going the GPU accelerated route for now because it would take 60 Paralella 16 boards to equal the processing power of 1 $150 Geforce gtx 750 video card(1.5 TFLOPS). Is there and reason why I should maybe choose Parallella anyways, with my end goal being High Speed Real time Computer Vision and Prototyping CV applications. I'd really like to make the best decision and I know Paralella far surpasses Rasberry PI.

[9600]

When evaluating performance you need to do more than simply compare FLOP ratings and you should also consider the architecture, tools and, of course, application.

Since GPUs are single instruction, multiple data (SIMD), this means that it is much harder to keep all the cores busy, than it is with a multiple instruction, multiple data (MIMD) architecture such as Epiphany. So a GPU may win hands down on theoretical performance, but is your application suited to the architecture and able to achieve anything like this? If it is suited to use with a GPU, great, but I suspect that there are many more applications that are not, than that are.

Other considerations include energy consumption/efficiency (big GPUs tend to be very power hungry), availability of documentation (comprehensive Epiphany docs are available without NDA) and openness of the development tools (the Epiphany and COPRTHR OpenCL SDKs are fully open source, others may not be).

Regards,

Andrew

[shodruk]

I think Parallella is a minimum configuration of the Epiphany supercomputer.
It is not a supercomputer in itself, but it becomes real supercomputer by clustering many Epiphany chip with grid topology.
We can't buy real Epiphany supercomputer, but we are able to learn parallel/concurrent programming on Epiphany with $99 Parallella.

http://www.adapteva.com/introduction/

My micro supercomputer image

viewtopic.php?f=11&t=171&start=60#p6598

viewtopic.php?f=11&t=171&start=70#p6788

[Transcendental]

[Transcendental]