ngx_pagespeed isn't as stable as I'd like but I'm running it in production with XF just fine.
Both will speed up page load times when it comes to css and js but http concat will do it with alot less server resource overhead in my opinion. So if you don't need full ngx_pagespeed feature set, http concat is good alternative.
IMO, both modules are useless. For example GZIP + CSS minification will blow http_concat, while for nginx_pagespeed, if you like to live with 3 x more server load without any significant improvements, go for it.
EDIT: My experience with ngx_pagespeed. And another interesting thread.
@motowebmaster, can you post more details please? I updated the previous post with my experiences related to ngx_pagespeed. Basically, pagespeed will pretty much tweak things for you. Why not do it yourself into templates and eliminate the server stress? I remember that @Mike Edge posted something about high server load on his multisite servers.
Some time ago, I did try modpagespeed on a previous server and experienced extreme server utilization. There was some tweaking-related conversations, and updates from the Groups that I joined on the subject, but gave it up after a while. Today, it appears that many are trying the latest stable version and still killing their servers.
This time around however, the experience is much different for me. My server is being utilized a bit more since modpagespeed was installed, but I'd call it more of a bump that a dramatic change. The theories for this aren't worth posting at the moment. This evening and tomorrow are the busiest times of the week for me, we'll see if I still have brain cells by then.
When this site was on vb3, I had years to tweak templates and develop alternatives to improve performance - but I don't think any of us will have that luxury on Xenforo for a while, we're having too much fun
Pagespeed does not significantly increase average server load as many are suggesting here. There are filters you can enable that significantly increase server load consistently. But in reality most of the filters optimize content then cache it and serve the optimized content from the cache on subsequent requests. You'll see a significant spike in server load when resources are first being called, optimized, then cached, but after a bit of time server load will level back out and you'll be saving throughput.