m.roth@5-cent.us wrote:
Longer answer: every single move, down at the machine/assembly level, can move twice as many bits as on a 32-bit system. That will show up as a very serious speed increase in your software.
actually, the pentiums have had a 64bit physical memory bus since the first 60Mhz version, and all L1/L2 cache <=> physical memory operations are 64bits. the CPUs have all optimized things like REP MOVSB to move by 64bit chunks whenever possible.
The main performance advantage of x86_64 vs i686 is in the additional general purpose registers, this allows the compiler (or assembler programmer) to minimize the number of load/store instructions to implement a given sequence of operations.