On 08/04/2011 04:36 PM, Kevin Grittner wrote:
Anyway, I'm always willing to take advantage of the benchmarking work of others, so I'm very curious about where performance topped out for you with HT enabled, and whether disk waits were part of the mix.
Hah. Well, it peaked at 2x physical cores, where it ended up being 60% faster than true cores. It started to fall after that, until I hit 64 concurrent connections and it dropped down to 36% faster. I should also note that this is with core turbo enabled and performance mode BIOS settings so it never goes into power saving mode. Without those, our results were inconsistent, with variance of up to 40% per run, on top of 40% worse performance at concurrency past 2x core count.
I tested along a scale from 1 to 64 concurrent connections at a scale of 100 so it would fit in memory. I was trying to test some new X5675s cores against our old E7450s. The scary part was that a dual X5675 ended up being 2.5x faster than a quad E7450 at 24-user concurrency. It's unreal. It's a great way to save on per-core licensing fees.
We're also on an 8.2 database. We're upgrading soon, I promise. :) -- Shaun Thomas OptionsHouse | 141 W. Jackson Blvd. | Suite 800 | Chicago IL, 60604 312-676-8870 sthomas@xxxxxxxxx ______________________________________________ See http://www.peak6.com/email_disclaimer.php for terms and conditions related to this email -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
