A little bit, yeah: it's like a bit like single-pass radix sort, i.e., where the "digit" in the radix sort is selected to be large enough that one digit encompasses the whole key. Another difference is bucket management strategy is different: using a fixed set of buckets with chaining rather than counting ahead of time the buckets you'll need and laying them out linearly.
Note that locality of reference is so important that even though "single pass" radix sort does the least work, the ideal radix turns out to be 6-10 bits usually, even if that means e.g., 5-10x more instructions executed for 64-bit keys: because sparsely writing into buckets spread across memory is just really slow.
Note that locality of reference is so important that even though "single pass" radix sort does the least work, the ideal radix turns out to be 6-10 bits usually, even if that means e.g., 5-10x more instructions executed for 64-bit keys: because sparsely writing into buckets spread across memory is just really slow.