[0:00] PHP 8.5 was just released last week and
[0:03] naturally you are as curious as I am if
[0:06] it's faster than previous versions of
[0:08] PHP. My colleague Tim set out to test
[0:12] this with a benchmark for PHP 8.4 and
[0:15] comparing a Symfony Laravel and a
[0:17] WordPress demo applications against
[0:19] previous versions of PHP.
[0:22] Can you expect a big performance
[0:24] increase like from PHP 5 to 7 or is
[0:28] there no noticeable increase? In this
[0:30] video, I'm going to discuss our
[0:32] findings. Mo, I am Benjamin and my team
[0:36] and I work on PHP performance 24/7 for
[0:39] the last 10 years, helping thousands of
[0:41] developers along the way. Let's start
[0:43] discussing the benchmark results that
[0:45] Tim collected. You find all of them in
[0:47] our blog post. The link is in the
[0:49] description. First, the setup is always
[0:52] important. Uh the blog post lists
[0:54] everything that went into this
[0:56] benchmark. So, first you will find all
[0:58] the code that we use for infrastructure
[1:00] provisioning and the demo applications.
[1:03] We used a Hetsa machine with eight
[1:05] dedicated virtual cores. We are using uh
[1:08] the Debian 13 Trixie release uh for a
[1:12] database MySQL or SQLite depending on
[1:14] the demo. And the uh one important
[1:17] factor is for PHP we used um the suri um
[1:22] packages built for Debian um including
[1:26] PHP 8.5 release candidate number three
[1:30] for the PHP setup we didn't change a lot
[1:34] uh the JIT is not enabled and uh PHP FPM
[1:37] is configured with a static pool of 17
[1:40] workers.
[1:42] The for the project versions we have
[1:44] Laravel version 12, Symphfony 7.3,
[1:47] WordPress 6.8 and we used Va for um uh
[1:51] doing performing the benchmark testing.
[1:53] As for our methodology, what we ran here
[1:56] is two different kinds of benchmarks. So
[1:59] we tested with a fixed request per
[2:02] second number so that we can focus on
[2:04] the response time of PHP um with this
[2:07] fixed number of requests. And the second
[2:10] test we did is with a fixed level of
[2:12] concurrency to focus on the sort of
[2:16] throughput. How many requests per second
[2:19] did we achieve by running this fixed
[2:21] level of concurrency. Uh both tests run
[2:24] for about 60 seconds or for 60 seconds
[2:27] and then we compile these numbers and uh
[2:30] generate charts from them.
[2:33] We try to make the test setup as simple
[2:36] as possible because we usually expect
[2:38] that you don't reconfigure PHP too um
[2:43] broadly. So what we didn't want to test
[2:46] is um PHP versions that are
[2:49] self-compiled or um use very complex
[2:53] setup of the demo applications or PHP
[2:55] itself because from experience
[2:58] um 95% of you will be running PHP on um
[3:02] more vanilla uh configurations the
[3:05] default configurations.
[3:07] So um we wanted to replicate this kind
[3:10] of running PHP. Let's go through the
[3:12] results one by one. We are starting with
[3:15] Symfony and we are looking at the test
[3:18] scenario where we run 100 requests per
[3:21] second a fixed number of requests every
[3:23] second and then we are looking at uh the
[3:26] response times that we see from the uh
[3:28] demo application. So for a symfony
[3:31] application we can see across versions
[3:34] PHP 8.2 2 to 8.5 that the response times
[3:39] are slightly slower for PHP 8.5 compared
[3:43] to 8.2 or 8.3 for example in the mean.
[3:47] Uh also like super slightly slower in
[3:51] the median 95 percentile and 99
[3:54] percentile.
[3:56] This is not statistically
[3:59] uh significant. So um it is within the
[4:02] margin of error. um the numbers could
[4:05] statistically be the same. So there's
[4:08] really no difference in the performance
[4:10] between these versions if we um perform
[4:13] this kind of test where we run a fixed
[4:15] number of requests um against the
[4:17] Symfony application every second. For
[4:19] the second test for Symfony uh running
[4:22] 15 concurrent users to find out how many
[4:24] requests is the application being able
[4:27] to process in a fixed time frame. we see
[4:30] there um across all PHP versions it's
[4:33] around 500 requests and uh again it's
[4:36] within the margin of error so um it's
[4:39] not clear which version here is faster
[4:41] or not they all seem to be equally fast
[4:44] same for the Laravel case uh first for
[4:47] the test with 100 requests per second uh
[4:51] testing the response time we can see uh
[4:54] in the mean median 95 and 99 percentile
[4:58] are roughly the same there seems to be
[5:00] one outlier for PHP 8.3 in the 99th
[5:03] percentile but this is also not very big
[5:07] and um selfcorrected itself in the next
[5:10] so in PHP 8.3 so in general again here I
[5:14] would say the performance for PHP 8.5 is
[5:17] the same than the previous versions 82
[5:20] to 8.4 for for Laravel case with the
[5:24] scenario of 15 concurrent users making
[5:27] as many requests as possible in 60
[5:29] seconds. We can also see here around 850
[5:33] requests for this Laravel demo
[5:35] application and it's the same uh
[5:37] regardless of the PHP version. So again
[5:40] it doesn't seem to be that PHP 8.5 is
[5:43] generally faster than previous versions.
[5:46] As a third case, we've took WordPress um
[5:49] specifically also because it's possible
[5:52] to run WordPress still with PHP 7.4. So
[5:56] here we can test is there a performance
[5:58] difference between PHP 7.4 and all the
[6:02] PHP 8 versions including 8.5. And for
[6:05] the test with 100 requests per second,
[6:08] we can see the performance is not really
[6:12] um very different between these
[6:14] versions. um there is within the margin
[6:18] of error no real um improvement between
[6:21] all these versions.
[6:24] So we can conclude that PHP 8.5 is also
[6:28] not faster for WordPress applications
[6:31] compared to previous versions. For a
[6:33] test with 15 concurrent users with
[6:35] WordPress, we this is the only case
[6:37] where we can see a small difference that
[6:41] there um is outside of the margin of
[6:43] error. So we can see for PHP 8.3 4 and 5
[6:48] the number of requests that we can uh
[6:51] achieve in a fixed time frame is a
[6:54] little bit larger than for PHP 7.4. So
[6:57] um there is this difference between PHP
[7:00] 8 and PHP 7. But again PHP 8.5 is not um
[7:06] be able to handle more requests per uh
[7:10] time frame than PHP 8.4 or 8.3.
[7:14] What do these results mean in general?
[7:16] Uh so we found out PHP 8.5 is not faster
[7:21] than PHP 8.4 or previous versions than
[7:24] PHP. So uh the result that we already
[7:28] saw last year when we benchmark PHP 8.4
[7:31] is still true. PHP is at a very high
[7:35] level of performance for a dynamics
[7:37] language and achieving general
[7:40] improvements is really hard at that
[7:42] point. Uh to achieve changes you
[7:46] actually need to look at the application
[7:48] and runtime level of PHP. What you can
[7:52] obviously do is check the performance of
[7:55] your PHP application in uh specifically
[7:58] are there SQL queries that are slow? Are
[8:01] you using APIs the wrong way? Maybe
[8:03] there are more modern APIs in PHP now
[8:06] that you that weren't available
[8:08] previously.
[8:10] Uh we published a blog post on
[8:12] improvements in PHP 8.5 that will
[8:15] significantly improve the performance of
[8:18] specific use cases. But uh you need to
[8:21] find if you are actually using PHP and
[8:24] using these APIs and are benefiting uh
[8:27] from them. One thing I want to note here
[8:31] is for example the persistent storage of
[8:34] curl state across um requests. Uh this
[8:38] uh will allow applications that talk to
[8:41] third party APIs to um improve the
[8:44] performance quite significantly with
[8:46] 8.5.
[8:48] You can find a blog post in the
[8:49] description uh on this um on on the
[8:53] specific PHP 8.5 performance changes.
[8:57] A second change that is possible is uh
[9:00] to rearchitect the way you are using
[9:02] PHP. You can for example use franken PHP
[9:05] or any other application server in PHP
[9:08] to um save time bootstrapping PHP and
[9:12] reuse state across requests. You can um
[9:16] move processing out to the background
[9:18] jobs using message cues and other um
[9:21] patterns that allow you to reduce the
[9:24] time spent in the web server. So what is
[9:27] really possible to make a difference in
[9:30] application performance is in your
[9:33] hands. You are um driving this change by
[9:37] building your application in a way that
[9:39] ends up being faster for users. If you
[9:41] like my content on PHP performance, you
[9:44] can uh continue with this uh video on
[9:47] PHP 8.5 changes in performance or you
[9:50] can follow this channel or subscribe to
[9:52] our newsletter link is in the
[9:54] description. Bye.