Thomas Kern wrote:
Since Mhz and MIPS are such misused values, I prefer to run the same
program on old and new engines to compare the performance change. I use
an old FORTRAN (no flames please) program that computes pi to 5000
places. A boss once needed something to see if the vendor really did
upgrade our processor, just after our Performance and Capacity Planning
group was disbanded. So on an idle system, this just gets into memory
and runs with a high Total/Virtual CPU ratio. On my z890 IFL, 20
iterations takes about 8.571 sec virtual, 8.578 sec total and 8.944 sec
elapsed time.
Find a favorite program that is repeatable and keep using that to test
the speed of your engines. But all numbers must be taken as relative and
with a big grain of salt.
CPU power is only ever part of the story; if it's all that matters,
you'd all be using Xeons or Opterons.
Do a proper benchmark, one that reflects what you want to do. Even if
your current workload is constrained by CPU, doubling the speed of the
CPU or doubling the number may well do nothing than find the next
bottleneck.
In a queue at the theatre, everyone might be lining up to have their
tickets checked at the door and thereafter being shown quickly to their
seats. If there's a big queue at the door, getting more ticket checkers
won't help if you can't also show people to their seats more quickly.
Improving one component of a balanced system just makes the system
unbalanced.
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
Please do not reply off-list
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
