Advanced Micro Devices Inc (AMD)

[chipguy]

It's difficult to argue about realistic "ranges" without being more specific about what your "range" means. One standard deviation? Over what domain of applications?

Well that's the rub now isn't it. The 64 bit applications that
appear will be a self-selecting sample space. As BUGGI
correctly points out why would an ISV bother with expense
of an AMD64 distribution of a program unless it provides
a performance benefit beyond the basic noise level of
benchmarking.

Pointing to N released AMD64 programs and claiming their
average, range, or variance of performance increase over
32 bit equivalents as typical of 64 bit compilation will be
fallacious. The programs that are losers or just break even
with 64 bits will not be represented in anywhere close to
realistic proportions.