Well obviously, but just because the BogoMIPS numbers
aren't proportional
to the exact speed, doesn't mean they don't still mean *something* about
the relationship. The K6 system I'm typing this on gets 600 BogoMIPs and
is low-end by today's standards, and whether or not it's actually 2/15 the
speed, the Winchip is a whole bunch slower. That's all I meant.
The problem with that reasoning is that BogoMIPS aren't an absolute
scale of *anything*. There's a different "correction factor" that gets
used to scale the number depending on what kind of processor the kernel
thinks it's running on.
Given only two BogoMIPS numbers from two systems, it's not even possible
to infer which processor is faster to run the BogoMIPS loop.