I think the Pi0/Pi0W are *part* of what is preventing Raspbian from going 64-bit now, but by no means the full reason.ejolson wrote: ↑Wed Oct 17, 2018 9:07 pmWhat makes the Pi Zero interesting for many is more the small size and low power consumption than the small price. With existing 64-bit SOC packages, where to place the memory becomes a problem unless components are mounted on both sides. With existing 64-bit SOC packages, power consumption is a problem unless all cores are disabled but one and that remaining core is clocked in the sub-gigahertz range. Would anyone be interested in a tiny Pi with one 64-bit ARMv8 core clocked at 700 Mhz, 1GB of memory and a single micro USB port at a price 5 times greater than the current Pi Zero? I think the answer to this question is why Raspbian has remained 32-bit so far.
The real issue for upgrading the Pi0/Pi0W is going to be price. On a technical basis, it is probably possible to design a single-core (using, say, the Cortex-A53 like the BCM2837), clock it at or near 1GHz (which the Pi0/Pi0W are now) and keep the TDP in check enough to permit PoP RAM. Even better would be to use some of the tricks that went into the Pi3B+, like using the PCB as a heat sink. If the RAM had to be kept at 512MB to do that, so be it. It's enough for a single-core system. The "gotcha" is...who would ask for such a chip and/or how do you pay for the design work?
I will admit that if someone gifted me with the One True Lottery ticket, I would be sorely temped to walk into Broadcom and say, "This is what I want, and here is a check to cover the full cost of reducing it to manufactiring." Then the cost of the chips would be *just* the manufacturing cost and profit to the fab and Broadcom...which should make it possible to put such a hypothetical chip onto a Pi0/Pi0W board. Anyone got a spare $10 million to $20 million burning a hole in their pocket?
As to price increases... If you take the Pi0 to 5 times current one-off price of $5, you are more expensive than an A+, and at that price, a Pi3A+ should be feasible. At 5 times more than the off-off price of a Pi0W, you've got a board that in more expensive than a Pi3B+. Such price changes would completely alter the decision making process for what board to pick. The *only* places it would make sense to use the 'neo-Pi0"/"neo-Pi0W" would be where small size is *the* critical element and power requirements *might* affect the decision. After all, for many uses, would you rather a single-core 500MB 1GHz that is very small or a slightly to quite significantly cheaper quad-core 1.4GHz (or more) 1GB board that somewhat larger?