t3 uses time tagged triggers, samm~ sends clicks that are not time tagged.
I see, are there other advantages (cpu-wise maybe)? Since these periodic clicks could also be generated with t3-based abstractions.
If anything I would expect t3 to be more cpu-efficient, but the CPU usage of samm~ is so low that probably the two approaches are interchangeable. I would be curious to see an example t3-based abstraction that does sample accurate metronome clicks. I think the main advantage of samm~ is convenience. It's nice to be able to set up a bunch of synched sample accurate metronomes, all at different speeds, without giving much thought to the underlying mechanisms. According to jkc at Cycling '74 all of the functionality of my system of sample accurate objects could be replicated with MaxMSP core objects, and I believe him. It would just be more awkward to do it the native way in many cases.
Best,
Eric