Using an Intel Xeon 678X (HP Z4 G6i) system running Fedora 44, I ran a few PHP benchmarks to satisfy the Phoronix Premium supporter request. After all, Phoronix Premium is the leading way to help the site continue operations.
PHP 7.4 was the oldest release tested since older (unsupported) versions were seg faulting when built with GCC 16. All of the PHP releases were the latest code in each of the Git branches. All of the PHP releases built the same from the same compiler toolchain on Fedora 44.
It's a pretty straight-forward comparison using a handful of PHP benchmarks.
They were command line driven to avoid the overhead of the web server performance, etc.
With the widely-used PHPbench, the performance climbed slowly over the course of the PHP 8.x releases. Of course, the numbers would be much more impressive if able to include earlier PHP7 and PHP5 releases too, but that was prevented by the modern toolchain having issues with the old code. PHP 8.6 Git was regressing relative to PHP 8.5 but hopefully that will be closed by the stable PHP 8.6 release later in 2026.
Not much to see out of PHP's integrated micro benchmarks.
Within our own self-test, PHP 8.5+ is looking good and there was not any PHP 8.6 regression like seein in PHPbench. PHP 8.6 with the JIT enabled helped out slightly here. The numbers are much more fascinating compared to earlier PHP7 and PHP5 versions that I may do as a follow-up article on an older platform.
PHP 8.6 with/without JIT was looking nice in these benchmarks. PHP 8.6 performance will be revisited when the official release is near.
Aside from the PHPbench regression on PHP 8.6 Git, the PHP 8.6 performance was looking nice and more incremental gains over PHP 8.5.