[wplug] Need suggestions

Brian Medley bpmedley at 4321.tv
Sun Oct 21 14:38:47 EDT 2001


On Sun, Oct 21, 2001 at 01:50:12PM -0400, Edward Walter wrote:

> Personally, I take most of the benchmarks I encounter with a grain of
> salt... there are just to many variables which can cause benchmark
> performance to diverge from what you encounter in the real world.  That
> said, benchmarks do have value in that they can help to establish a

yes. I agree with you.

> One other thing to note is that the results on both of these benches are
> normalized against CPU speed... just to even the playing field.

However, does it really even the playing field?  It appears that their
normalization process is: (HITS/S) / MHZ.  Doesn't this assume a linear
relationship between processors?  Wouldn't this imply that the way
instructions are handled on the particular processor didn't change *at
all*?

It's been some time since I read about it, but I thought that intel
"tweaked the core" and added instructions imbetween the PII and PIII lines.
So, assuming this, it is conceivable that one instruction will take a
different amount of time on the PIII.  I would think that this would skew
the linear relationship.

-- 
~'`^`'~=-.,__,.-=~'`^`'~=-.,__,.-=~'`^`'~=-., \|/  (___)  \|/ _,.-=~'`^`
                          Brian Medley         @~./'O o`\.~@
"Knowledge is Power" brian.medley at verizon.net /__( \___/ )__\  *PPPFFBT!*
  -- Francis Bacon                               `\__`U_/'
 _,.-=~'`^`'~=-.,__,.-=~'`^`'~=-.,__,.-=~'`^`'~= <____|'  ^^`'~=-.,__,.-=
~`'^`'~=-.,__,.-=~'`^`'~=-.,__,.-=~'`^`'~=-.,__,.-==--^'~=-.,__,.-=~'`^`



More information about the wplug mailing list