That's strange. When SGA_TARGET is 512M, what are the sizes of the buffer cache, shared pool, etc. Have you tried manually setting the cache sizes and not using SGA_TARGET? On our server the cpu_count was 64 and that required a 256M buffer pool (64 x 4M granule size). The total sga was around 500M. When I set cpu_count to 8, I could have a 350M SGA. Actually Oracle would let me set it lower but that size fit out reguirements. I was using the old manual sizing (db_cache_size, shared_pool_size, etc). These were small dev/test databases. Keith ----- Original Message ----- From: Mandal, Ashoke To: Bradd Piontek Cc: Keith Moore ; oracle-l Sent: Thursday, October 16, 2008 12:41 PM Subject: RE: Memory issue with Oracle 10g database on Sun Server T5140 with Solaris 10 Brad, I tried setting the cpu_count to 8 and then SGA_TARGET of 256M didn't work and had to use SGA_TARGET as 512M. Thanks, Ashoke