Re: Feature Request --- was: PostgreSQL Performance Tuning

On Fri, 4 May 2007, Michael Stone wrote:

>> P4 2.4GHz    107ms
>> Xeon 3GHz    100ms
>> Opteron 275    65ms
>> Athlon X2 4600    61ms
> PIII 1GHz    265ms
> Opteron 250    39ms
> something seems inconsistent here.

I don't see what you mean.  The PIII results are exactly what I'd expect,
and I wouldn't be surprised that your Opteron 250 has significantly faster
memory than my two AMD samples (which are slower than average in this
regard) such that it runs this particular task better.

Regardless, the requirement here, as Josh put it, was to get a way to
grade the CPUs on approximately a 1-5 scale.  In that context, there isn't
a need for an exact value.  The above has 3 major generations of
processors involved, and they sort out appropriately into groups; that's
all that needs to happen here.

> You misunderstand the purpose of bogomips; they have no absolute meaning, and
> a comparison between different type of cpus is not possible.

As if I don't know what the bogo stands for, ha!  I brought that up
because someone suggested testing CPU speed using some sort of idle loop.
That's exactly what bogomips does.  My point was that something that
simple can give dramatically less useful results for predicting PostgreSQL
performance than what you can find out running a real query.

--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD