Mathmatically they are the same, but maybe someone else would be able to shed more light on this. I would think that less at 100 would be better because it would be more consistent but I could be wrong.
without hyper you have more mathematical power on the single share (and you should solve it faster and then complete earlier your work)
with hyper you have 2x share to work on but with half power..
think the only way to know is try 24H each and see..
I agree, keep records and compare
IMO: by using single thread with higher hash rate you 'll reach slightly better results ... just tip, I'm not an expert