>: >I have been goin over a bunch of old posts, and have found very little
>: >favorable posts about it.
>I certainly posted positively some time ago. the K5 is completely
>problem-free, and fully P5-compatible, even on esoteric features
>like 4M pages, RDTSC, etc.
>kernel compile times vary widely depending on what .config you choose.
>for mine, a P5/133 did a make in 7.5 minutes, 7.0 minutes with a K5/133.
>identical hardware, which was a Supermicro STE/512, 32M EDO, 3G WD EIDE
>using busmastering. kernel 2.0.20 or so.
>I have to admit I'm jaded by my current dual P6/200 though ;)
>: But as far as my experience goes, the AMD K5 is superior in every
>: way to a similarly clocked Intel 80502.
>it's _definitely_ not in FP performance. there it's up to 40% slower.
>and that's not on P5-scheduled Quake or some asinine magazine benchmark...
I think this depends on which model of K5 you get. Apparently the
newer ones cheat on their advertised speed; the K5PR133 runs at
either 100 or 110mhz internally. The earlier K5s (75,90,100)
actually ran at the advertised clock speed and would, in my warhorse
test of doing quantum chemistry modelling, complete the task about 5%
faster than the similar Intel product.
I think AMD made a mistake when they started ignoring the clockspeed
in favor of this bogus pentium rating. Though I suppose that 90% of
the target market won't care much about the speed of floating point
and will be more impressed to see K5/PR133 instead of K5/110 labelled
on the chip.