Starting from 8.3 Pega uses batch-based ID generation which improves performance and scalability. Coming to you question it still depends on pc_data_uniqueid, its just that it doesn't update the table for every id generation.
When a new node requests for a case id, for the 1st time in that node request will go to the database and batch size of 1000 (LastReservedID->LastReservedID+1000) will allocated to that node by default in Memory and LastReserved would get updated with LastReservedID+1000. For the next 999 case ID generations the request doesn't go to the database.
Similarly, other nodes also get their batches assigned and update the LastReservedID. Nodes only request database a new batch when batch is completed or node is restarted.
Case id's on node one would be 1,2,....1000 at the same time cases created on other node would be 1001,1002,,,2000. You can notice that the id's are not sequential but however this approach improves the performance.
You can change the default batch size using the DSS idGenerator/defaultBatchSize. If the value of this DSS is set to 1 then for every id generation the request goes to the database table (Also resulting in a sequential case id's).