AFAIK there is no way to measure the load-time of a single field within the load of multiple fields. You could of course load each field alone and measure than the load-times of each table - but this isn't mandatory the same, whereby I could imagine that it could give some valuable hints.
I assume that string-fields will need more time as purely numerical ones especially depending on the lengths of the string-fieldvalues (a number is always a number with 8 byte - whereby by loading them as a dual-value they might be much larger).
Another important point is probably how many distinct values are within the field - more distinct values means that more values needs to be stored and also the pointer will increase.
Further it was mentioned that this load is a resident load - in this case it could be that no fieldvalues needs to be newly stored and that just the pointer may change.