So, on my test server (300 MHz K6-2):
First number is with php parsing off, second with it on. Each is an average of three runs with 100 accesses at 10 concurrent with the following command line:
ab -n 100 -c 10 testmachine/test/test.htm
100 byte file: 162k/sec 74k/sec
500 byte file: 320k/sec 215k/sec
1k file: 460k/sec 328k/sec
4k file: 617k/sec 700k/sec
16k file: 789k/sec 640k/sec
This is a file with NO php code actually in it, just a bunch of text inside the standard HTML/HEAD/TITLE/BODY tags.
This test was run across a 10BaseT network, that was relatively unloaded (i.e. on Friday, hehe)
I currently have the following set to be handled by php on my server:
.php .php3 .html .js .htm
.js is nice because it lets me embed php into my java scripts, which is handy for customizing pop-ups and custom folding menu trees in java.