I know that the worker or per-child mpms should be more efficient than the prefork mpms, but I'm having trouble proving it . I tried out the worker mpm and ran ApacheBench against my server and I found that prefork was actually producing better numbers - even after I increased the MinSpareThreads and StartServers numbers. Is ApacheBench not a good testing utility, or are there tweaking considerations that I'm not aware of. BTW, I'm using apache 2.2.3.
depends on how you configure it. if you have many CPU's you need to have enough httpd's to keep them all busy in worker mpm