performance is great but i would imagine the paralellized version requires a significantly higher minimum amount of RAM than the non paralellized ways... he cla

performance is great but i would imagine the paralellized version requires a significantly higher minimum amount of RAM than the non paralellized ways... he claims that each more perfomant solution is less compute cost than the one before it, but in the case of paralellization its just the same amount of compute in just a shorter amount of time, right?