There is only so much mindshare that can fit in a headline, but don't forget that there is also a 28k logic cell FPGA on that ARM processor.
This board has about everything you could want in a hardware hacking board.
• Normal, well supported, ARM processors for the easy stuff. USB and HDMI in case you want to interact.
• Gigabit ethernet. Most ARM boards you find are 100mbit.
• Sanely programmable array of processors for bulk processing.
• FPGA for heavy hacking. Also very handy for interfacing to devices, let it handle the low level timing sensitive stuff.
• About 40 GPIO, but since they are connected to the FPGA you can turn them into higher level interfaces. It appears that the HDMI output is built this way.
• Community momentum. There are 3400 people behind this already. 5 signed up while I was spell checking this comment.
However if I read their block diagrams correctly you can't use the FPGA and the Epiphany at the same time ( e.g. they use the FPGA to interface with the Epihpany Core ). Still if they give easy access to reprogram the FPGA they not only made an awesome Parallel computer, but also the only platform I know of to easily get into FPGA hacking.
The FPGA is the most under reported part of this project.
We didn't focus on the FPGA because the project was more about parallel programming methods, but it's an open platform so the adventurous ones can hack and modify it freely.
The FPGA is indeed doing the glue logic to the Epiphany, but that's only one of its functions; there's plenty of space left for other jobs, one of which is HDMI output. They do plan easy access to reprogram the FPGA, but since Xilinx synthesis tools are only available on x86, you'll also need a PC to make new FPGA configurations.
In the Kickstarter updates there is a link to a paper about using the FPGA to both interface to a digital radio front end and the Epiphany, so I assume that the Epiphany logic doesn't consume the whole FPGA and that they make source available for you to extend.
Most high-profile Kickstarter projects raise 20 % or more of their total pledges in the last 48 hours. If you check the pledge total trendline on Kicktraq , you'll see that it aligns pretty close with that number.
I disagree with a previous commenter. They've delivered enough proof as far as their ability to deliver. The problem is their "old" pitch was targeted towards developers & there is just not enough of these on Kickstarter to raise 750k, only yesterday did they fine-tune their selling points & replaced the project video with one that's more likely to resonate with mainstream users (Ubuntu, media center, a Rapsberry Pi on roids etc..)
I wish them nothing but the best.
It's definitely possible to raise this money. The rate of funding has increased a lot over the past day or so. Think about it like this, if they average about $4600 an hour for the next 32 hours they'll make $147200 more, which will take them over the goal. Watch the funding for an hour or two and you'll see it's definitely going to be close to making it, just need an extra push to get there.
Just to follow up on this, if you'd like a visual representation the funding pattern over the duration of the project, the graph found here will be of interest: http://canhekick.it/project/5064d31d8f647c24e5ad60d0
I set the Parallella project to update every 5 minutes instead of every hour, but the website serves the project page from a TTL-cache with a 1 hour expiration.
Added in edit: Anyone is welcome to make use of the API in their own applications. There is documentation[1] and I'm very accessible if you have unique requirements.
The site looks great. One piece of feedback: middle-clicking using Chrome a project name in the home page browses to it in the same page; the expected behavior is to be open in a separate tab. Other links (such as the ones in the footer) behave normally.
Thanks! The fix will show up the next time I deploy. Which might be a while -- my revision control discipline on this project isn't as good as it could be.
What uses would something like this have? What types of tasks are processor intensive?
Video encoding, speech/face recognition? Computation that normally a GPU would do?
I bought a 16/core but I'm not sure what I'll use it for other than to run Linux.
I really hope that someone develops a distributed computing client (for Folding/Boinc) that would run on this and take advantage of it. Than I would buy a 64/core
More and more consumer applications need high computing performance and a tighter integration of memory and computation. You already mentioned a few examples yourself.
However, I think that the most interesting thing about this parallel computing platform is that it's completely open. This will allow a community formed around it to create new, unthought of, innovative, applications with it.
I have no doubt they can deliver (the engineering samples do work) but the target audience is small and until there are multiple working applications. The concept, general purpose parallel computing, is the way forward. So if you like to learn and tinker, this is one of the few if not only way to get access a development board at a very low cost.
Backing them does not cost you anything if the pledge is not successful. If you want to have access to a cheap development board then back them. I do.
I'd like to see them succeed. The sweet spot I would love to use for this technology would be portable SDR, just like one of their app notes states. It's a pain dragging a powerful laptop into the field just to do some rote signal processing for an radio.
True, im afraid its unlikely they will make it. I would really like this Kickstarter campaign and ultimately this project to succeed, but I believe they have not delivered sufficient proof that they are capable of delivering the product they are promising.
If this post manages to stay on the front page, they may get enough traction, but still unlikely
Yes it is. However, I've followed the funding progress in the last few hours and it's really increasing fast.
I really do hope it works out. I really see a lot of potential for (clusters of) these boards for machine perception and other neural computing applications.
This board has about everything you could want in a hardware hacking board.
• Normal, well supported, ARM processors for the easy stuff. USB and HDMI in case you want to interact.
• Gigabit ethernet. Most ARM boards you find are 100mbit.
• Sanely programmable array of processors for bulk processing.
• FPGA for heavy hacking. Also very handy for interfacing to devices, let it handle the low level timing sensitive stuff.
• About 40 GPIO, but since they are connected to the FPGA you can turn them into higher level interfaces. It appears that the HDMI output is built this way.
• Community momentum. There are 3400 people behind this already. 5 signed up while I was spell checking this comment.