PHP: The Version Dropdown Is a Speed Control

Whenever two visitors see different HTML — a logged-in view, a basket, a page of search results — PHP is building that response at request time, usually assembling it from rows in a MySQL database. The 8.x line runs materially faster than the versions PHP's dated reputation was formed on.

Concurrency is the part people forget. Each simultaneous uncached request occupies one worker for its whole duration, so a slow query does not just make one page slow; it holds a worker that the next visitor needed. Cache hit ratio and worker count are the same conversation.

Without a bytecode cache the interpreter re-parses and re-compiles your source on every single request. OPcache holds the compiled form in memory, so a warm process skips straight to execution. This is why the first request after a deploy is slower than the ones behind it, and why benchmarking a cold cache tells you very little.

Warm it before you measure. Request the page once, discard that reading, then take your three runs. A benchmark that includes the compile step is measuring your deployment, not your application.

Clone to staging, switch the version there, and exercise the paths that break first: checkout, forms, anything using a deprecated extension. Read the error log rather than the front page, because deprecation notices land there long before something visibly fails.

Once it is clean, switch live in the panel and re-measure. Google's good threshold for server response time is 800 ms; an uncached WordPress page on healthy hosting typically sits in the 200 to 500 ms band, and sustained full seconds means the problem is a plugin or a query rather than the runtime.