> Right, I was mainly thinking in terms of speedups from not having to switch tasks and the gains from specializing in one kind of computation.
Switching tasks is going to be a relatively small amount of time over a 24 hr period. Given that one person can only hash 0.67 hashes per day assuming full undivided attention I'm really not sure what you could count on saving.
The biggest advantage adding more people would do would make it actually feasible to calculate continuously as I doubt any human has the focus/patience for doing mundane number crunching for 24 hours without breaks.
I totally get that this is all hypothetical-for-fun math, obviously nobody is going to actually try to mine bitcoin by hand. But that 0.67 hashes per day is terribly low, a 50% speed improvement would only barely put you over 1h/d. Not to mention we normally compare in h/s not h/d so it's actually 7.75e-6 H/s.
And I wasn't interested so much in the absolute throughput, but how the figure would change in terms of hashes per second per human, so duplicating a full ALU module wouldn't be relevant; the concern is how much an additional human improves the rate after adjusting for the fact that you added one more human-resource.
Edit: I'm new to queuing theory, but I may just grind through the numbers there. Could do it as a "Show HN" if there is a tool (preferably as a web app) where you can model a process and tweak the numbers like in Factorio.