> A lot of people seem to think that you can't address 4GB of RAM with a > 32 bit OS, which isn't true of course. However there is an effect, > particularly under Windows, on the size of an individual process: IIRC > Windows grabs half the process space for itself while Linux only grabs > the top 25%. I don't have a reference for that or I'd include it. > However, this is probably neither here nor there for most people: not > many of us do the sort of heavy duty rendering or video editing that > needs several GB per process and they will all have considerably more > RAM installed, so a 64 bit OS will be mandatory. Well, i can't speak for others, but i am aware that you can use PAE to get above 4gig on x86. As far as processes using large amounts of memory, for myself personally - there is a reason i have 8gig of RAM and am thinking of upgrading to 16gig. ...and yes, 64bit OS is manadatory - regardless of the OS, for me anyway. > That's not true. A more correct statement is that a 64 bit version of > the same program MAY run faster than a 32 bit version on a CPU with the > same speed and core count. Simply using longer registers has almost no > effect because only a small proportion of instructions in any program > will involve 64 bit operands apart from address referencing in the 64 > bit version (of course) and so any speed increase will be entirely due > to whether the compiler optimiser takes advantage of the extra registers > in the 64 bit CPU. Some compilers may do this. Others won't. Any speed > gains here must be offset against the cost of reading the bigger 64 bit > binary into memory and the impact of the extra code size on L1 and L2 > caching. Its also possible that JIT systems like the Java JVM will look > at the number of available registers at runtime and optimise > accordingly, in which case a 32 bit JVM would run at the same speed as a > 64 bit one. Well, i don't think i ever implied that 'simply using longer registers' would automagically make code execute faster. It is a prerequisite (to what i said, and also assumed) that the compiler would be taking advantage of your CPU's features and optimizing accordingly. I've run both 32bit and 64bit OSes on my phenom II 3.2ghz quadcore, and i can tell you right now - 64bit is noticeably more performant ... >> I'd also like to point out that telling someone with some /lib32 >> issues, to install a 32bit OS is not a solution. >> > Only true if the problem is bugs in the library. Most of the 64 bit > problems appearing on this list are due to the absence of said 32 bit > libraries, typically those required by video drivers, not to the bugs in > them. that maybe true, but re-installing your operating system should NEVER be one of the first options given. One should exhaust all other options and possibilities, first., IMHO. cheerz and thanks for your insights.
