Comments on: WordPress Metric Comparison of 2.9.2 and 3.0

By: Milan Petrovic

Milan Petrovic — Tue, 06 Jul 2010 22:45:18 +0000

I made very detailed benchmarks of WP 3.0 few weeks ago for admin side, and I am preparing blog side (it will be done in a day or two), and WP 3.0 is slower on the server side, but with faster execution in the browser. Also uses more resources. But, all that is expected because many things are added to WP 3.0 to improve user experience, and every next version will require more resources, and I doubt that anything can be done about it.

By: Ron

Ron — Tue, 06 Jul 2010 22:13:36 +0000

Thanks for adding the extra metric. Interesting that the difference in the two default themes is negligible.

By: Corey

Corey — Tue, 06 Jul 2010 20:50:49 +0000

Great test for WordPress speed. I’m intrigued that 3.0 is slower even with the lesser amount of queries. What tools did you use to check the speed? Did you just do browser side or were there any server side speed tests?

By: Matt

Matt — Mon, 05 Jul 2010 23:40:25 +0000

For everyones enjoyment I have updated the post to include metrics from Kubrick on 3.0 and updated the comparison to be a more apples to apples comparison but have left the Twenty Ten metrics as well.

By: Demetris

Demetris — Mon, 05 Jul 2010 17:26:33 +0000

My tests show different numbers of queries.

For the home page with 2 widgets (Search and Categories), 2.9 + Kubrick needs 11 queries, while 3.0 + Twenty Ten (and no custom menu) needs 10 queries.

Twenty Ten did not start very well at the queries front. See:

http://core.trac.wordpress.org/ticket/13085

But then some fancy functionality was removed from it and now it does well.

Another significant improvement in queries was for custom menus, which were rather expensive at some point in the 3.0 development cycle. But little by little the number of queries for custom menus was reduced to a very acceptable number. (Thanks mostly to several tweaks by Ryan, if I remember correctly.)

Cheers!

By: Peter Westwood

Peter Westwood — Mon, 05 Jul 2010 06:34:48 +0000

All the time the comparison is done using two different themes you are comparing more the one variable.

For example it could just be that TwentyTen uses less queries but takes longer to render.

By: Matt

Matt — Mon, 05 Jul 2010 00:06:17 +0000

@Mark: Yeah, I was just about to update the post, those figures were actually an average over 25 runs each. And I used the “default” theme on each version. It wasn’t necessarily meant to be an apples to apples comparison. More of a default install to default install comparison. As I mentioned it wasn’t exactly comprehensive. I could have tested with Kubrick on 3.0 and posted results between Twenty Ten and Kubrick. I’ll however leave that to the curiosity of the reader to test.

If anyone wants to perform such tests, just post in the comments.

By: Mark Jaquith

Mark Jaquith — Sun, 04 Jul 2010 23:57:19 +0000

You just did one run each? You should do a bunch and grab the average. Also, is this using Kubrick on 3.0, or Twenty Ten? Should be the same theme for a fair comparison.