Third benchmark is used to measure the performance on the actual website or blog side visible to all visitors. Since WP 3.0 has a new theme, test was expanded here to include both Kubrick and Twenty Ten themes under WP 3.0, and only Kubrick under WP 2.9.2.
To check the results from first part of the benchmark, go the this post. There you have test setup and results for memory usage, page generation speed and executed SQL queries. Second part of the admin test is in this post.
Tests results presented here are average values from 20 measurements. Also, only active plugin in all tests was GD Press Tools Pro. Adding more plugins have no effect on the overall difference in terms of memory use or execution time. Showing more tests that have the same result in the end is not needed.
Test 6: PHP Used memory on site
Similar to the admin side, WP 3.0 uses more memory on the site side. Difference is about 1.6MB, and it’s not much in this normal setup with no plugins active (expect for GD Press Tools).
Adding plugins to the mix will have as a result more or less the same difference in used memory, or slight increase to 2MB of more memory. But even with more than 20 plugins on site side WP will still use much less memory than admin side. Still, 32MB for 20 or more plugins is not enough.
Also, used memory with Kubrick and 2010 themes in WP 3.0 is almost the same, with new theme needed a bit more memory, considering new features added in it, this difference is not important.
Test 7: SQL queries executed on site
Code optimizations in WP 3.0 were beneficial on site, and WP 3.0 will run less SQL queries for same result from WP 2.9.2. This is only couple of queries, but from server point of view, very important.
With less queries, SQL execution time is lower, but that didn’t help WP 3.0 in the next test, with overall time needed to make a page increased.
Test 8: Page rendering server side on site
With both themes WP 3.0 is slower than WP 2.9.2 by 10-15%, in all the tests. This difference is just about the same as on the admin side. And it’s to be expected, more code is added to each execution (more memory is needed), and PHP needs more time to run all that.
Due to some added features new Twenty Ten themes is slower than Kubirck. Surprising result is on the single post page where WP 3.0 with Twenty Ten was faster than with Kubrick. Getting post data seams to be optimized better in WP 3.0 and that shows here.
More tests: Client Side?
Progress has it’s price. And for WordPress 3.0 that is more resources needed and slower execution. With todays available servers and faster Internet, this is not a big factor, considering what you will get after you upgrade to WP 3.0. So, not to point only negative side of the benchmark, I will write conclusive article in the next few days that will put this values in better context and real evolution of WordPress in it’s current release.
Bottom line: upgrade to WordPress 3.0 as soon as possible (as soon as you are sure that all plugins you use are compatible).
3 thoughts on “WordPress Benchmark: 3.0 vs 2.9.2, Part 3”
Well, these issues do come into play a lot if your WP installation is complex and or very busy. But I still prefer WP to most other platforms…
I guess we will be spending more time trying to cache those php files and reducing database calls.
And thanks for running this test, its highly useful to me and should be to Matt and the WP development crew.
Exactly, the WP core is getting bigger (and slower) with each version, and there should be a line somewhere so that WP doesn’t became bloat-ware like so many other great platforms and programs did. Maybe is a good idea to make some of the features optional and that can be added later if needed. For instance, I don’t need export, import, links or some other things and I don’t need to install them every time. Such level of optimization is not so easy to do, but it will be very useful and welcomed by most WP users.
MillaN: Isn’t that just what they did. Ie export and import is available as links. But you have to download as plugins the first time you use them.
I just installed/upgraded and am moving several blogs into one using the builtin network capability. I guess that possibly points towards bloatware… but I’m quite happy since it’s my main use (to have several blogs operate from one install).
Comments are closed.