I did further testing after that post and it was inconclusive. I couldn’t get a reliable difference at all.
The only changes I made was to the framework I use. The templates were changed to point all images to images.examle.com rather than example.com/assets/images and the like. It wasn’t much that needed to be changed because I set up images.example.com to just go straight to example.com/assets/images. I’m not sure if it’s a symlink or a rewrite, but PHP would not see the request the same way, so I had to parse it a bit. Nothing major. The only benefit I found was audits of my site improved because images, css, and js were being served off of cookie-less (sub)domains, but I couldn’t see any real world advantage.
Again, I think it depends on your site. I have a low traffic wordpress site loaded with pictures and other goodies. I put that behind CloudFlare and saw a big improvement, but that’s because the load time was already 5-7 seconds or something like that. On the site where I tried the poor man’s CDN, even without the changes, the site loaded in under 2s, and often faster than 1.5, so there wasn’t much room for improvement.
That’s why I recommend setting some clear benchmarks before you make any changes.