modularity

[adexmont]

hello, it's about 6 month tht i'm thinking about an hardware solution but i was not able to find a commercial solution.
all i see is moving in SoC direction in the meanwhile i'm thinking on modularity tht is the opposite side.

My first question is : how far is a board tht don't need arm but can run with the epyphany core "alone"

I'm uploading a fantasy picture tho give u a view of my idea and i'd like to know if i'm too much in the fantasy world or if one commercial product like this one is possible.

Tnx for ur time

[9600]

[adexmont]

the final product will be a wide rangeproduct from domotic central unit control to home - mobile personal computer to smatphone.

The idea is tht if i need more i can add another board no matter if it's a cpu board a ram board or a peripherial board, so u can build what u need and don't buy a system tht half of the time it's not fully used or it lack something.

I'm focused on epiphany because it's a simple structure processor that do calculation (not like intel amd or arm tht include core controller and other block in one cpu) and it's the only one tht i found in the reality and not like a simulation, then in an fpga chip u can choose how the cpu do the computation.
In other words with the same hardware if i install a different software - firmware i can change the rule of the machine.

Iwas looking in the opencores.org site but i'm not able to retrive any usefull information.

This was my thought

[shodruk]

How did you make that picture?

[9600]

[adexmont]

[9600]

[Gravis]

[Richard Dunn]

I wish to congratulate you on your concept for open architecture based on the hexagonal form factor. Unfortunately for you I own the technology. The US rights were granted Sept. 23. 2013; no prior art not obvious, US 8,542,492. However I took it to a much deeper depth than you have envisioned here. This patent covers hexagonal prism enclosures/modules designed with a sliding fit on rails opening up into rooms for repair, maintenance or replacement so that modules abut, including pretty much full spectrum access. the USPTO also gave me maximum circuit density possible based on minimum propagation path length which are media dependent and nesting of components.

This was designed to supersede the current pizza box - cabinet - aisle paradine. I posted that statement on the Linkedin Multi-Core and Parallel Processing Group run by the gentleman tied to the Kilometer Square Swiss radio space telescope. It has stood for three weeks without challenge. The gentleman is also a member of the IBM HPC Group.

The design obviates the universal calculated time problem as the hexagon face is divided into equilateral triangles with equal length propagation paths. Module height and diagonals as to paths are also so co-ordinated as well as through the circuitry as necessary for processing. As the modules are abutting standard anyplace in the module array, so are the legs for uniform time standard signal propagation throughout the system with fail over and battle hardening for parallel abutting paths of abutting hexagonal walls. This is designed to minimize MTBF.

Your people do not understand why this would have a large impact on sales. I cannot imagine a reason except maybe one module would house 6-to maybe 180 plus cards pending communication rates. Kind of a cluster. Now without a discount, that would work out to at least a grand per cluster at 6. So the communication is a problem. So communication speeds and though put are kind of important to me. The idea of offloading the communication processing is also important. i see this as a point to point network with look-ahead.

Without knowing forum rules there is a post re a network controller on another thread that this is related to. You see I have some people who want me to put up a model for VC which I am not jumping up and down over. There are others in the que.

Richard Dunn

You may not want to look at the patent, 48 drawings on this one and 92 pages total.



This was designed for really big iron, average sales are $100-250M with the Chinese Milky Way-2 coming in at $390M. Your people either do not believe in the possibility of these numbers. will th modules are designed When These are assembled they look like what I refer t

[adexmont]

thanks for reply Richard, i read the patent in a fast way , beside the low quality of design i found in that pdf, i'm still thinking that my idea is a lot forward than the patent .

I'm still able to patent my idea if i have money to do that and mainly if i beleave in patent.

Copyright on a system like this is like thinking that someone can repatent the pitagora theorem :S .

i just shot a pair of solid on blender to concretize my idea and to resolve a pair of geometric solution tht i'm not able to solve by mind.

I already thought about a connection system different to the patented one, i think my junction are more veratile than P (patent) and much more simple on production step (my connection system is just one connector like ram slot with a spine that lock (in the center of the square connector)

maybe i "read" too fast but i don't see any capability to overlay a secondary plane of board.

Obviously the main problem i can see is the limit of the BUS structure ... but we are talking about a cpu designed to do that or not ?
Parallel programming and potentially infinite scalability.