Well a compacF is around 5 or more times the speed of the older NewPacs.
So one compacF needs the same work generated and written to USB as about 5 NewPacs
Don't these stick miners check diff, if needed generate a new nonce and re-hash all by themselves?
Do they simply get data and send back a SHA-256 hash of it instead?
I always thought having the block generation done by the host would be way too much communication overhead and that all miners just had a small MCU on them which recreates a new block candidate.
That way, communication would only be a new block template with previous block hash and a set of transactions when needed, but not a communication round for each and every single attempt / hash.