Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Mailing Lists -> Oracle-L -> Re: CPU speed comparison

Re: CPU speed comparison

From: Mark Brinsmead <>
Date: 2005-12-24 01:31:03

You need to consider *much* more than just CPU speed. Cache size, system (mainboard) architecture,
and dozens of other factors and components also matter.

That said, you can probably find the numbers you need for *rough* comparisons at

I presume you'll be interested in multi-processor systems -- so you'll probably want to look at
SPECrate benchmarks.

Sriram Kumar wrote:

> Hi Gurus,
> Firstly Wish you all Merry Christmas.
> We have a banking application and I am in the process of doing a
> CPU Capacity prediction of the DB server based on the Ratio Modelling
> Technique. I have arrived at the CPU requirements for particular
> platform(Say Itanium2) and now I want to arrive at the CPU numbers for
> the other Server platforms(PA-RISC, IBM PowerV/VI, Intel
> Xeon, Ultrasparc IV/III).
> Say for a given volume set I arrive at a 2 CPU Itanium2, Now how do I
> arrive at CPU numbers for the other platforms. Any inputs/pointers
> would very much be appriciated.
> I looked at
> and feel that
> we can use this script for the said comparison provided the underying
> assumption is correct. The scripts basic assumption is using LIO's to
> determine the CPU usage . Run this script in various platforms and
> use the CPU usage figure across platforms to arrive at the
> comparitive numbers. Does this approach sound logical?
> In this effect, I remembered rading Cary's papaer "Why You Should
> Focus on LIOs Instead of PIOs" and I found the following comment from
> the author
> The actual /use /of Oracle block content is the number one consumer of
> CPU capacity on any reasonably well optimized Oracle system. Your
> average LIO latency will vary, depending primarily upon your CPU speed
> and the total number of machine instructions required to parse the
> content of an Oracle block
> Thanks and Regards
> Sriram Kumar
Received on Sat Dec 24 2005 - 01:31:03 CST

Original text of this message