Using the chip itself (package), to broadcast/receive the data as light. Using open-air, not fiber-optics. (Since this would not be broadcasting feet and miles, just inches.)
This, being a open-source project.
I have thought about this for years, yet I have never found a good reason to "realize" the idea, until now.
For the record, I do not design chips beyond software, however, my ability is personally irrelevant to this concept.
This is my thoughts/observations...
1: Many chips now have few external "supporting hardware dependencies". (Unlike the past, where capacitors, resistors, RF-filters, and various other "physical components" were needed.)
2: The chips/dies are getting smaller, but packages are not. (This due in part to the large number of communication wires and thermal-dissipation surface needed for many designs.)
3: Smaller dies in larger packages, adds slight delays over traditional "wires/traces", which can cause various timing issues or signal-losses, as well as collecting RF-noise and undesired capacitance. (Since a wire along plastic/glass/insulation is essentially a capacitor.)
4: Mounting chips limits heat dissipation, with one side constantly resting on an insulation glass PCB surface, and sitting on traces contained within the insulation PCB layers.
My original idea (not original itself, just saying it was my first thought), was to simply use a LED source, attached to a chip's edge. The connection being the data-lines themselves. (This was on a tiny Atmel-Atmega chip.) Though it worked great, the idea sat dormant for years again.
What I am thinking now, is that the LED component, or LASER-LED, could be built right inside of the chip. This would not "demand", "Fiber-optics", though isolated fiber-optic lines would create better isolation.
With the ability to easily isolate and create specific light frequencies with the new LEDs, and sensors, I think it is time for this idea to be explored again.
In the advent of bitcoins, for ASIC mining, I think this concept "realized", would seriously help advance bitcoin mining, and have a mutual relationship with the rest of the advancements in technology, as a whole. It is about time we gave chips an upgrade. Going smaller is not much of an advancement.
The reason it would really help mining, would be the following...
1: You would only need two terminals for the chip. Positive and negative, for power.
2: You would need few micro-connections to the die. Only the two power, and two for the emitter/detector.
3: There would be almost no RF noise from the data, as it is not broadcast over wires.
4: There would be a great chance to split dies into smaller pieces, using internal optics, to assist in a wider thermal-footprint.
5: The chip could have thermal dissipation on both sides of the chip, with a direct internal "ground" leaf.
6: The heat-sink itself, could be the dies RF-shielding, and also the light-channel guide.
7: You have greater freedom of where the individual components of any "unit", as a whole, could be positioned. (As opposed to being limited to the PCB itself.)
8: More distantly, using RF-power, you could eliminate all external wire-connections entirely. (Using that empty space in the chip itself as the location of the wire-coil that capture RF-noise for power.)
9: The light itself can be the "timing", and the "data I/O", and even power, on smaller levels. (Solar.)
You could instantly feed every chip, all at the same time, the new data to be processed. You would not have to individually communicate or regulate communications. Same for the "results", which only one chip will broadcast, and each other chip can instantly prepare for new data, before being told. (Since they can openly see the others results. Unless they are all isolated with fiber-optic lines.)
So... any thoughts... questions?
Which individual chip "now", has the least demand for external supporting hardware. The block-erupter chips seems to have few external physical components required. Even if the external connections are limited, this still has a "first-step" use. (Stripping all the data-lines is the major thing. Heck, PCIe would be the size of one LED, if you removed the data-lines and redundant power-lines. See where I am headed with this now...)
)