As I mentioned previously, I intended to sign on the kickstarter campaign for the mini-cluster of 4 64-core boards. I didn't change my mind until very late into the campaign when it became clear to me that the 64-core version wasn't likely to happen. Since that time I've never seen any 64-core boards or eLink interconnected 16-core clusters for sale.
As far as major promises did the Parallella campaign NOT live up to (excluding the 64-core version), as far as I know it's still possible to connect multiple Parallella boards together using the Epiphany eLink interfaces. That novel interconnect was to me a key feature of the Epiphany architecture. Without it interfacing multiple Parallella boards together is forced to go through the Ethernet provided by the Zynq... which (assuming the eLink architecture performs as well as is claimed) is impediment & disappointment.
Without fully exploiting the eLink interconnect I think it's really difficult to get a full and accurate understanding of the capabilities and weaknesses of the Epiphany architecture.
Please don't take this in a personal or highly negative way. You guys have accomplished a lot and I really, really don't want to diminish that. You've also dealt with the difficulties that the project faced with a professional, straight forward, and positive demeanour, which I feel is both lacking in many other kickstarter campaigns and which speaks volumes for the quality of character of the developers working at Adaptiva.
I really want to see you guys succeed. I am still interested in buying / building a mini-cluster but I'm holding off until they can be properly interfaced through the eLink fabric. Also, I very much would like to see a next generation Epiphany chip and subsequent nextgen Parallella board, beyond the v2.0 Epiphany III Parallella board that was discussed in the forums.
As far as major promises did the Parallella campaign NOT live up to (excluding the 64-core version), as far as I know it's still possible to connect multiple Parallella boards together using the Epiphany eLink interfaces. That novel interconnect was to me a key feature of the Epiphany architecture. Without it interfacing multiple Parallella boards together is forced to go through the Ethernet provided by the Zynq... which (assuming the eLink architecture performs as well as is claimed) is impediment & disappointment.
Without fully exploiting the eLink interconnect I think it's really difficult to get a full and accurate understanding of the capabilities and weaknesses of the Epiphany architecture.
Please don't take this in a personal or highly negative way. You guys have accomplished a lot and I really, really don't want to diminish that. You've also dealt with the difficulties that the project faced with a professional, straight forward, and positive demeanour, which I feel is both lacking in many other kickstarter campaigns and which speaks volumes for the quality of character of the developers working at Adaptiva.
I really want to see you guys succeed. I am still interested in buying / building a mini-cluster but I'm holding off until they can be properly interfaced through the eLink fabric. Also, I very much would like to see a next generation Epiphany chip and subsequent nextgen Parallella board, beyond the v2.0 Epiphany III Parallella board that was discussed in the forums.