I am an idiot.
Just as this post proves that ASICs and FPGAs exist for all these algorithms.
I've worked with FPGAs. They can be reprogrammed in about 1 minute to a completely different 'system'.
Parallel processing as in running the same 'code' on one chip. Depending on the complexity. 1000s.
FPGAs are actually better because of the reprogrammable part. ASICs are designated hardware like a video card is designated to output video to any display.
I haven't worked with FPGAs for 18 years. But they were extremely powerful back then. The Altera software costs an arm and a leg.
I'm not an expert by any means. I could not implement a full replica of the MOS 6502 on an FPGA, because it's impossible. Look up the undocumented instructions and how coders used them to do impossible things on everything that used that CPU.
Opencores website has existed forever. They had open source implementations of anything for FPGAs.
I wrote a small synthesizer. It used the UART code, simple 1 bit DAC code, coded my own oscillators (based on the C64 SID spec sheet).
The analog filter had to be implemented on a CPU (on an FPGA). I dropped the CPU. Good CPU cores need IP licenses.
Serial type of processing on an FPGA is possible, but eats up all the resources fast.
I even experimented with the VGA core. Had VGA output to a screen working in 1 hour. Needed few resistors on the GPIO pins. That's it.
So as long as there is code that can mine using that 'new' algo, anyone with a cluster of inexpensive FPGA dev boards is way ahead of the game.
Here is the link:
https://opencores.org/One member of the community is developing a user friendly FPGA miner (which includes Keccak, among other algorithms). There was also a working prototype of this fpga mining using solar energy.