Message-ID: <4D823CB9.9090906@emolecules.com>
Date: Thu, 17 Mar 2011 09:54:17 -0700
From: Craig James <craig_james@emolecules.com>
User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US;
	rv:1.9.2.15) Gecko/20110303 Thunderbird/3.1.9
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: Xeon twice the performance of opteron
References: <AAC806D7-56CD-4BC3-A449-83771F818A4B@torgo.978.org>
	<AANLkTimeWk=QKbjWPLBnDQUAqf2WXabJ80wTmeEby=Nt@mail.gmail.com>
In-Reply-To: <AANLkTimeWk=QKbjWPLBnDQUAqf2WXabJ80wTmeEby=Nt@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 3/17/11 9:42 AM, J Sisson wrote:
> On Thu, Mar 17, 2011 at 10:13 AM, Jeff<threshar@torgo.978.org>  wrote:
>> hey folks,
>>
>> Running into some odd performance issues between a few of our db boxes.
> We've noticed similar results both in OLTP and data warehousing conditions here.
>
> Opteron machines just seem to lag behind *especially* in data
> warehousing.  Smaller
> cache for sorting/etc... is what I'd always chalked it up to, but I'm
> open to other theories
> if they exist.
We had a similar result with a different CPU-intensive open-source package, and discovered that if we compiled it on the Opteron it ran almost twice as fast as binaries compiled on Intel hardware. We thought we could compile once, run everywhere, but it's not true.  It must have been some specific optimization difference between Intel and AMD that the gcc compiler knows about.  I don't know if that's the case here, but it's a thought.

Craig