Joe Pfeiffer <pfeif...@cs.nmsu.edu> writes: > shawn wilson <ag4ve...@gmail.com> writes: > >> On Tue, Aug 28, 2012 at 1:34 AM, Bob Proulx <b...@proulx.com> wrote: >>> Stephen Powell wrote: >>>> By the way, there's something I don't understand. A 32-bit processor can >>>> only access 4G of "real" (extended) memory, right? So why are there >>>> motherboards available for 32-bit processors that support installing >>>> more than 4G of RAM? What good is memory that the processor can't address? >>> >>> With PAE (physical address extensions) the processor *can* address >>> more than 4G of ram. A single process is still limited to 32-bits >>> which usually works out effectively to 3G of ram but the operating >>> system can make use of more than this. It can be used for filesystem >>> buffer cache and for multiple 3G programs. A machine with 6G of ram >>> for example could run two 3G program at the same time and hold them >>> both in memory without swapping. Or run one 3G program and still have >>> 3G for the system to use in filesystem buffer cache. With PAE having >>> more than 4G of memory is quite useful. >>> >>> Using PAE does have a small performance impact. It slows things down >>> by 2%-3% in my use cases. But the increase in ram for buffers usually >>> more than makes up for the differences. >>> >> >> iirc, pae is only 48 bits too. > > "only" meaning 256 terabytes in this case... I'll be very surprised to > ever see a 32 bit processor that can make effective use of that much > memory.
Sorry to follow up my own post -- I hit "send" too fast. My response was given the assumption that PAE gave a 48 bit physical address space; of course it doesn't. It gives 36 bits -- 64 GB, which isn't nearly so outlandish. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1b7gsjndia....@pfeifferfamily.net
