if the databases don't connect, the performance will not really be affected; if you calculate however a system with (say) 45,000 processes, then this will have a drastic effect as the size of the matrix that is built affects the computation time overproportionally.
However, you typically should not combine two different ecoinvent unit process system models, rather just use one, since the modeling in each system model (cut-off, APOS, etc.) is different and thus a combination inconsistent.
Typically, it is just RAM, you need 8 GB for an ecoinvent system. And of course a somewhat modern CPU helps.