If it's a simple select * from table, the data goes straight to the client. If the SQL has some intermediate steps for data processing, such as select count(*) from table, UGA will be used. In dedicated config, UGA is part of PGA.
Of course the data blocks will be scanned into SGA (buffer cache specifically), unless it's a direct path read. Your original question is whether the data will be in PGA. I already answered that. But no matter whether it's select *... or select count(*) ..., the table data must go to the buffer cache first (except for parallel read, and a few other special cases of direct path read).