"I always considered that size and speed were important for amateurs. But in the modern context, there's only three parameters that matter: power, power, and power."

--Chuck Moore

https://youtu.be/0PclgBd6_Zs?t=389

I genuinely wonder if this topology won't make a come back in the coming years.

To me GA144 looks less like a state-of-the-art CPU or DSP replacement and more like an FPGA replacement—and that would actually be lovely to see, given how inefficient and expensive FPGAs are nowadays compared to doing the same thing with a custom chip (if you only overlook the design cost for the latter). The “computer” marketing makes little sense to me.

Unfortunately, GA don’t seem to have gotten the memo about cheap entry-level dev tools that the microcontroller world has been circulating since 2010 or so. At the prices they charge for their devboards, you wouldn’t really get one just to play with, even if the chips themselves are actually somewhat cheap compared to getting the same amount of compute on an FPGA.

(It’s $500 per devboard and $20 per chip, with each chip capable of 2e9 16-bit multiplies/sec. Somebody[1] is selling a breakout board with only the chip for a much more reasonable price of $35, but you’ll need to figure out how to wire up the thing yourself.)

Charles Moore is known for aggressively patenting his hardware, though. (Well, he and every other modern chip designer.) So we might not get to see anybody else do this stuff for a long, long time. Adapteva’s Epiphany/Parallella design used a broadly similar idea with an explicit grid interconnect and also aimed for a piece of the FPGA/ASIC pie, but they had much beefier, synchronous cores with separate message routers attached to them.

[1] https://schmartboard.com/schmartboard-ez-qfn-88-pins-0-4mm-p...