Re: performance config help

From: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
To: Greg Smith <greg(at)2ndquadrant(dot)com>
Cc: Bob Dusek <redusek(at)gmail(dot)com>, pgsql-performance(at)postgresql(dot)org
Subject: Re: performance config help
Date: 2010-01-11 22:44:18
Message-ID: dcc563d11001111444w6251c638y2aa39f386fb0d1d8@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Mon, Jan 11, 2010 at 3:04 PM, Greg Smith <greg(at)2ndquadrant(dot)com> wrote:
> Scott Marlowe wrote:
>>
>> So, I took a break from writing and searched for some more info on the
>> 74xx series CPUs, and from reading lots of articles, including this
>> one:
>> http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3414
>> It seems apparent that the 74xx series if a great CPU, as long as
>> you're not memory bound.
>>
>
> This is why I regularly end up recommending people consider Intel's designs
> here so often, the ones that use triple-channel DDR3, instead of any of the
> AMD ones.  It's extremely easy to end up with a memory-bound workload
> nowadays, at which point all the CPU power in the world doesn't help you
> anymore.

The DDR3 Nehalem and DDR2 AMD are both actually pretty close in real
world use on 4 or more socket machines. Most benchmarks on memory
bandwidth give no huge advantage to either one or the other. They
both max out at about 25GB/s.

It's the older Xeon base 74xx chipsets without integrated memory
controllers that seem to have such horrible bandwidth because they're
not multi-channel.

For dual socket the Nehalem is pretty much the king. By the time you
get to 8 sockets AMD is still ahead. Just avoid anything older than
nehalem or istanbul.

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Greg Smith 2010-01-11 23:17:08 Re: performance config help
Previous Message Robert Haas 2010-01-11 22:15:37 Re: Choice of bitmap scan over index scan