$750k seems really ambitious for a kickstarter project with such a niche audience. Is that realistic for 29 days, considering you need to sell 7500 units of the $99 pledge amount? At the end of day 7, Leapmotion had 15,000 applications for a free leap motion device and SDK, and that is a device with a much larger audience.

Do you have any investor lined up that would be willing to maybe match a Kickstarter total pledge amount of something realistic like $375k?

Have you considered approaching a fund like In-Q-Tel? This seems like the kind of project they would fund, since I imagine a lot of the best parallel computing work is being done in government-funded agencies and labs. I also imagine the government is probably the biggest employer of people working on parallel processing devices. With that in mind, getting a device like this into the hands of many, allows more people to get hands on exposure to parallel computing.

Overall, it feels like the funding strategy needs to be diversified, because I imagine it will be difficult to get $750k all from one source, with the exception of a VC fund whose thesis aligns with your goals.

Lastly, it feels like a project like this would be a bit too soon. Many developers who are playing with hardware have been playing around with the arduino for a few years, some are now graduating to the Raspberry Pi, which offers clear benefits over the Arduino because you can run tons of stuff simply not possible on the Arduino. However as the Raspberry Pi just came out, I imagine that most developers are still trying to get their hands on something like it and still don't feel the pain of trying to solve problems with it, that could only be solved with something like the parallella.

As a hobbyist, besides exploring parallel computing for its own sake, what other kinds of problems can I explore/solve with the parallella which simply wouldn't be possible on the raspberry pi? Sell a dream and possibilities here. I'm personally not familiar with what would only be possible on a parallella and I might feel more interested in this project if I know why I'd want it (besides learning pp for its own sake).

[0] http://arstechnica.com/information-technology/2012/09/99-ras...

My biggest concern is Step 2. Will the Epiphany chips be easily obtained in small quantities? I've been following the A13-OLinuXino development, an improvement (LQFP processor, open board design etc.) and a geat idea, but the Allwinner A13 chip isn't available through normal distributor channels.

874 Backers

$97,486 pledged of $750,000 goal

27 days to go

If they can raise 100k in 3 days, and have 9 such cycles to go... sounds feasible, even if not easy.

They use a Zenboard (Xilinx Zync, ARM+FPGA) as a base platform. My first reaction; they claim this being 'open source'. Nothing about the ARM processor or even the core inside the FPGA is open source. What they will deliver is the toolchain and the documentation, but no IP or RTL code for the cores. Another Fauxpen source project... using it merely as a buzzword to get people involved; comparable to the Beagleboard to get usecases and branding out.

When hardware is called "open source", they need to look at how Milkymist does it. PCB design files are offered, but also the RTL verilog is available for the CPU (in fact the whole SoC).

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.

So we won't be able to fabricate our own derivative silicon. But we will have all open source drivers and tools. We won't have chips full of DSPs that we can't use, or GPUs that work a little bit through some driver that the silicon vendor had to get to MVP for a single version of Linux and can abandon in a year. Sounds good to me.

[1] http://adafruit.com/products/791 [2] http://propeller.wikispaces.com/Programming+in+C+-+Catalina

Well... technically neither can this :)

Although they do appear to have got it working really neatly with an ARM-based host board. Which is a step forward.

Not necessarily Google, they are famous for not investing in other people to do engineering, but it seems like a modest amount of money to get something that should be fairly widely applicable.

Of course if it is widely applicable and this investment gets to company what it needs to take off, well the folks who gave them the money aren't really going to benefit in a leveraged way. (No equity)

Perhaps they were hoping for a 'raspberry pi' like response (which would be hundreds of thousands of units) and be able to do a sort of stealth funding round kinda thing. No idea of course, but it would be a sweet result if it worked out for them right?

It's an interesting concept. If it's cheap enough, it's definitely useful as a DSP. But the custom instruction set is going to keep it away from general usage. They seem to be hoping to get their IP into a cellphone, but that's a seriously uphill battle.

I backed $99. I'm very excited.

Turns out their product isn't remotely similar: up to 64 cores at 800mhz, plus a dual-core ARM CPU. I wonder why they are doing a kickstarter instead of harvesting their dollars from server appliances.

The CISC vs. RISC days are long over and the battle between the two architectures is a bit silly at this point since the gap between the instruction set and the underlying implementation has gotten quite dramatic. Claims that RISC chips are inherently more efficient may have been true in 1995, but I don't see this holding water today.

I think their marketing is a bit off on this one. Does everyone actually need parallel computing on this scale? I don't think so. Those who do need it likely already have CUDA running on GPGPUs. Not that I'm implying stifle innovation but the crowd they're marketing to seems to be way off base.

GPUs are different to CPUs in many ways, not least of which is that they are very difficult to debug, do not support recursion (this might have changed?), need special,non-standard data structures (streams). Multi-core CPUs suffer none of these limitations and developers can use non-proprietary standard build tools to develop software for them.