Performance testing is one of the most complex and hard types of testing. It’s really hard to make fair tests which fit everyone. There are a lot of things which can impact the measurement result. That’s why I don’t believe any particular test result which I found on the Internet.
If you are a software or hardware vendor, it’s a big temptation to test only your strengths. Even if you want to get fair results, you can miss something during the testing or don’t have enough results to see a performance impact. If you test your application with only one core/process but it’ll have two cores/processes on a production environment your tests could not show application performance on production.
Web servers performance could not be measured in general. E.g.: what will affect Apache2 vs Nginx performance results? There are a lot of things could affect performance like: static or dynamic content, what app server do you use (mod_wsgi or uwsgi), what does your app do (e.g. I don’t care on Wordpress performance results if I develop some custom apps with Django).
That’s why I believe that everybody should do own performance testing for their apps in a production-like environment. Any existing tests results found on the web should not be treated as 100% of the truth.