About this project
Note: for those new to Kickstarter, If we don't reach our campaign goal you DON'T get charged.
Making parallel computing easy to use has been described as "a problem as hard as any that computer science has faced". With such a big challenge ahead, we need to make sure that every programmer has access to cheap and open parallel hardware and development tools. Inspired by great hardware communities like Raspberry Pi and Arduino, we see a critical need for a truly open, high-performance computing platform that will close the knowledge gap in parallel programing. The goal of the Parallella project is to democratize access to parallel computing. If we can pull this off, who knows what kind of breakthrough applications could arise? Maybe some of them will even change the world in some small but positive way.
Project backing ends Saturday October 27th at 6pm!!!
We are offering some new "last-chance" tier customizations for all based pledges above $99.(not mutually exclusive)
CASE(s): Get a case for your Parallella board. We will be designing and making a case for the Parallella board using design inputs from the backer community. The more people who sign up for the case, the fancier the better it will be.($20 for every board in your reward tier)
SDK-ACCESS: You get a special login account and early access to the Epiphany SDK within 4 weeks of successful funding of the the project. Everyone else will get general access to software at time of shipment in May 2013. (Add $50 to any reward tier that doesn't already get early SDK access)
ONE-MORE: Available for MINI-CLOUD and CLUSTER rewards for backers who intend to donate the kits to a school of their choice, but want a (16+2) core kit for themselves.(Add $99 (+20 for international shipping if applicable))
PLAQUE: Anyone donating the MINI-CLOUD or CLUSTER reward to a school will receive an engraved plaque with the inscription "Donated by_________" or "Donated in honor of________". (Free)
X2-BONUS: Get the Gige Switch, cat cabling, and power supply adapters with the $1500 64-CORE-X2 selection (Free)
T-Shirt: Black T-Shirt with a Parallella slogan on it. For now the default is "I speak Parallella, do you?" Don't like it? Come up with a slogan and submit as comment. We'll run a survey and you will be able to pick the T-shirt size and which slogan to print closer to the shipment date. (Add $25 to any reward tier )
Book: Receive a book "An Introduction to Parallel Programming" written by the Adapteva team. The book will be provided in PDF form and will include source code examples of parallel programs that use OpenCL, OpenMP, Erlang, and MPI. (Add $25 to any reward tier)
To add these options, just go to "Manage My Pledge" and increase your pledge by the appropriate amount.
The Parallella Computing Platform
To make parallel computing ubiquitous, developers need access to a platform that is affordable, open, and easy to use. The goal of the Parallella project is to provide such a platform! The Parallella platform will be built on the following principles:
- Open Access: Absolutely no NDAs or special access needed! All architecture and SDK documents will be published on the web as soon as the Kickstarter project is funded.
- Open Source: The Parallella platform will be based on free open source development tools and libraries. All board design files will be provided as open source once the Parallella boards are released.
- Affordability: Hardware costs and SDK costs have always been a huge barrier to entry for developers looking to develop high performance applications. Our goal is to bring the Parallella high performance computer cost below $100, making it an affordable platform for all.
The Parallella platform is based on the Epiphany multicore chips developed by Adapteva over the last 4 years and field tested since May 2011. The Epiphany chips consists of a scalable array of simple RISC processors programmable in C/C++ connected together with a fast on chip network within a single shared memory architecture.
Here is a link to the Epiphany Architecture Reference Manual
Examples of the the scalable performance that can be achieved with the Epiphany architecture is shown in the following papers (with source code!).
Parallella Computer Specifications
The following list shows the major components planned for the Parallella computer:
- Zynq-7010 Dual-core ARM A9 CPU
- Epiphany Multicore Accelerator (16 or 64 cores)
- 1GB RAM
- MicroSD Card
- USB 2.0 (two)
- Two general purpose expansion connectors
- Ethernet 10/100/1000
- HDMI connection
- Ships with Ubuntu OS
- Ships with free open source Epiphany development tools that include C compiler, multicore debugger, Eclipse IDE, OpenCL SDK/compiler, and run time libraries.
- Dimensions are 3.4'' x 2.1''
Once completed, the 64-core version of the Parallella computer would deliver over 90 GFLOPS of performance and would have the the horse power comparable to a theoretical 45 GHz CPU [64 CPU cores * 700MHz] on a board the size of a credit card while consuming only 5 Watts under typical work loads. For certain applications, this would provide raw performance than a high end server costing thousands of dollars and consuming 400W. For a better indication of true performance of the Epiphany-IV and Epiphany-III processors in standard benchmarks go to http://www.coremark.org and check out our scores or read our blog post here.
The Team Behind Parallella
The Parallella project is being launched by Adapteva, a semiconductor startup company founded in 2008. The core development team consists of Andreas Olofsson, Roman Trogan, and Yaniv Sapir, each with between 10 and 20 years of industry experience. The team has a strong reputation of executing on aggressive goals on a shoestring budget. Our latest Epiphany-IV processor was designed in a leading edge 28nm process and started sampling in July, demonstrating 50 GFLOPS/Watt. To put this in perspective, consider that the Epiphany energy efficiency specs are within striking distance of the 2018 goals set by DARPA for the high profile Exascale supercomputing project.
Our passion is to design hardware platforms that are powerful and easy to use. To us there is no bigger satisfaction than seeing someone use our platform for a purpose that we could never have imagined.
Some of the testimonials we received for our work on the Epiphany architecture:
"In the course of my travels around the world I have been fortunate enough to meet some truly great engineers. However, it's rare that I am completely blown away by someone on the engineering front. At least, this was true until I was introduced to Andreas Olofsson, president and architect of Adapteva Inc. As far as I am concerned, Andreas is "an engineer's engineer." Clive Maxfield, EETimes
"Adapteva, with its laser focus on floating point performance and with no allegiance to either the x86 instruction set or graphics support, is able to squeeze a lot more performance per watt out of its design." Michael Feldman, HPCWire
“Visual computing can enable gesture-based gaming, advanced user interfaces, augmented reality, and even improved health and safety. Visual processing, however, requires many more flops than voice processing. Adapteva’s architecture can deliver the performance required for visual computing.” Linley Gwennap, Industry Analyst
"Andreas and his team have managed a rather fantastic feat: to develop a genuinely useful microprocessor with many task-parallel cores, operating within the power-budget of smart-phones and embedded devices. Accomplishing all this with a team of only a few full-time employees may seem unbelievable, but I’ve seen how he operates his company and the magic is simple: Focus on the simplest solution to any given task, hire a consultant when you need expert advice, and stick to all schedules!" Magnus Snorrason
The Epiphany story has been well documented over the last year in the press. The complete set of Epiphany online articles can be found here.
Where we are today
We feel great about the status of our Epiphany chips, software development tools, and prototyping boards. They have been thoroughly tested by early access partners for the last year and we are already starting to see complete applications being developed around our platform by others.
Most of the impressive work done on the Epiphany platform is either published as academic papers or hidden deep inside R&D labs, so we also created our own demos for you to see what the platform can do. The following video shows how our Epiphany-IV based EMEK4 prototype handily beats an x86 based processor on a key mathematical kernel while consuming a fraction of the power.
Here is a strong testimonials from on of our software development partners, Embecosm.
Why are we asking for your help?
Since our first Epiphany chips came out a year ago we have had a ton of interest from R&D labs, universities, and private technology enthusiasts (the early adopter crowd) but getting large corporations to buy into parallel computing has proven incredibly challenging. This phenomenon has been seen by a long list of parallel computing companies over the last twenty years that have so far failed to reach ubiquitous general purpose adoption. (See list) Based on the struggles of these companies and our own experiences in selling the Epiphany, we now understand that there is no way a single company can do it alone. The only way to create a sustainable parallel computing platform is through a broad grass roots movement.
Together we can launch a full on attack on the daunting problem of converting the software industry to parallel over the next few years after 70 years dominated by serial programming.
What will your pledges be used for?
The following picture shows our most recent Epiphany prototype platform based on an off-the-shelf open source reference board. The picture also highlights some of the work needed to reduce the size and cost of the platform to fit the size and cost goals of the Parallella project.
Epiphany Chip Cost Reduction Steps:
- Chip foundry retooling changes will reduce silicon costs to a few dollars per chip. These changes are very expensive and account for a large part of the funds needed to produce a low cost Parallella computer. Our 16-core Epiphany chips have been in the field for over a year and have been tested thoroughly by many hardcore developers. Chip product retooling for the sake of cost reduction is usually referred to as a "full mask tapeout" and should be considered a low risk part of the project. Our current low volume chip manufacturing flow only yields 50 dies per wafer. By creating full production mask sets for our chips we will be able to yield 1000's of dies per wafer.
- To reach the aggressive $99 price point we will also need to set up high volume chip testing and assembly flows.
Partners have already been lined up for both cost reduction steps discussed.
Parallella Computer Development Work:
- All major IC components have already been selected for the Parallella board, but cost minimization will continue.
- We will be engaging with an experienced external board product design team to complete the design and layout of the Parallella boards.
- We will work with internal and external resources to seamlessly integrate the Epiphany coprocessor drivers and development tools with the Ubuntu distribution currently running on the reference platform.
- Buying in bulk significantly reduces the cost of the platform. Without the large batch build enabled by this project, the cost of the Parallella boards would be many times higher.
Except for the Epiphany multiprocessor chips, the Parallella computer is a fairly standard ARM based low cost single board computer, giving us confidence that we will be able to meet our size and cost constraints.
We don't have time to wait for the rest of the industry to come around to the fact that parallel computing is the only path forward and that we need to act now. We hope you will join us in our mission to change the way computers are built. We could put 1,000 cores on a single chip in two years. Are you ready for that?
What will come out of it? We don't know but we do know that the following applications are DESPERATE for more efficient processing and are stalling today because bigger companies aren't serving their needs.
Consumer: Small energy efficient computer media box console emulator movie rendering
Imaging: face detection/recognition, finger print matching, object tracking, stereo vision, gesture recognition, remote sensing, video-analytics, manufacturing inspection, augmented-overlay
Communication: video conferencing, network monitoring, deep packet inspection, software defined networking,
Automotive: autonomous driving, driver assist, fog penetration, glare reduction, holographic heads up display, intersection traffic monitor,
High Performance Computing: real-time internet stream analytics, real-time market analytics, portable in the field supercomputing, soft encryption engine, code breaker, data logger, in the field seismology processing
Medical: portable ultrasound, dna sequencing,
Robotics: robotics brain, space electronics, robotics sensor unit, multi sensor inertial navigation
Speech: real time speech recognition, realistic speech synthesis, real time translation, speaker verification
Unmanned Aerial Vehicles: synthetic aperture radar, hyperspectral imaging, IR imaging, smart stream compression, large focal array sensor imaging, autonomous flight,
Wireless Communication: GNU radio, cognitive radio, small cell base stations
Parallella and Education:
We just announced an effort to hold Parallel programming competitions to be held next year. The rules were sent out to four different Universities and the feedback was very positive!
First Attempt at Competition Rules Found Here:
First Competition Proposal Found Here:
Donations To Schools:
We think a MINI-CLOUD or CLUSTER would be a really great teaching tool for high-schools and universities. One of the challenge for schools these days seems to be that they never have enough money for things like this.
If you are donating your kit, send a comment to "claim" your school. This will let others get a sense of where these kits will end up and might foster local competitions and collaboration.
If you are considering donating your hardware to a school for educational purpose, check with them about donation receipts for tax purposes.(we are not accountants and can't advise regarding those matters:-)
Risks and challenges
The Parallella carries risks like any major engineering project. Any design flaw can set product releases back months and part shortages can push out delivery times significantly. We know as well as anyone that it only takes a single bug to make a product unusable.
The Adapteva team has a lot of experience in delivering complicated projects on time within specifications and our development risk is reduced by having long standing relationships in place with first rate suppliers and manufacturers.
We are very confident that we can deliver what have proposed! Still, we want to make it clear that we can't guarantee that we will be able to deliver. If you do make a pledge to the Parallella project you are doing it knowing that there is a risk that something will go wrong. The one thing we can guarantee is that we will put our heart and soul into this project (like we did for 4 years when we built the Epiphany processor chips) and that we will be honest and transparent about our progress throughout the Parallella project.Learn about accountability on Kickstarter
2 ARM-A9 cores and 16 Epiphany cores.
We will be offering the Epiphany-IV based 66 (2+64) core version Parallella boards as soon as we reach our stretch funding goal of $3M. The reward will be available for those who pledge more than $199. The estimated delivery time for the 64 core boards would be May, 2013.
The Parallella project is not a board, it's intended to be a long term computing project and community dedicated to advancing parallel computing. The current $99 board aren't considered supercomputers by 2012 standards, but a cluster of 10 Parallella boards would have been considered a supercomputer 10 years ago. Our goal is to put a bona-fida supercomputer in the hands of everyone as soon as possible but the first Parallella board is just the first step. Once we have a strong community in place, work will being on PCIe boards containing multiple 1024-core chips with 2048 GFLOPS of double precision performance per chip. At that point, there should be no question that the Parallella would qualify as a true supercomputing platform.
The current board configuration only supports up to 1GB of SDRAM. This is a limit our current host ARM CPU. If the Parallella project gets funded, there will be more boards coming that have significantly more RAM.
The Parallella board uses a dual core 32 bit A9 ARM CPU that currently supports Ubuntu 11.10 and bare bone Linux. Our plan is to move to Ubuntu LTS 12.04 as soon as possible. It may be possible to support Windows through Wine or a virtual machine going forward, but we haven't checked those options yet. It may not be practical due to the memory limitation of the board.
Not initially, but it may be considered in the future.
The Epiphany chips are much smaller high end CPUs and GPUs. The 64-core Epiphany chip only occupies 10mm^2, about 1/30th the size of modern GPUs and CPUs. If we would scale up our chips to the same die size, the Epiphany chips would win in terms of raw performance. Still, that's not really the point. For a 5 Watt power envelop, it's energy efficiency that matters.
We think you should buy both! The Raspberry Pi has a much bigger eco-system at the moment and is a great starting point. Still, the Parallella board have some distinct advantage:
1.) 10-50x more performance than the Raspberry Pi
2.) An accelerator that can be programmed in OpenCL/ C/ C++
3.) Open specs/documents
4.) Gigabit ethernet
5.) More flexible and powerful GPIO
We have received a lot of negative feedback regarding this number so we want to explain the meaning and motivation. A single number can never characterize the performance of an architecture. The only thing that really matters is how many seconds and how many joules YOUR application consumes on a specific platform.
Still, we think multiplying the core frequency(700MHz) times the number of cores (64) is as good a metric as any. As a comparison point, the theoretical peak GFLOPS number often quoted for GPUs is really only reachable if you have an application with significant data parallelism and limited branching. Other numbers used in the past by processors include: peak GFLOPS, MIPS, Dhrystone scores, CoreMark scores, SPEC scores, Linpack scores, etc. Taken by themselves, datasheet specs mean very little. We have published all of our data and manuals and we hope it's clear what our architecture can do. If not, let us know how we can convince you.
Yes, the Parallella prototypes have been extensively tested with Ubuntu 12.04. The Ubuntu O/S runs on the dual-core ARM A9 CPU on the board.
Support this project
- (30 days)