There is no quantitive test, while there are anecdotal tests of its capacity, that is largely affected by exactly what you want to do.
In general if memory serves it takes 1GHz to do 300 playbacks or 600 bridges with media (I may have those backwards) via sip. This will not be a linear progression however. As for scalibility, that too varies with how you are going to do that, if you want really high end, freeswitch does run on
telcobridges.com's gear, which is of the caliber that most expect to see and have in a proper phone company.
Afaik there arent tests on TDM connectivity, at least nothing that is making the rounds as much as the sip tests.