Site speed and page load times – optimization issues

According to PageSpeed Insights and GTMetrix (Vancouver), the site optimization of Designer-Light page is quite alright. Very alright for a site like this. But…

But not so good news in terms of load times. GTMetrix reported load time is 9,9 sec, Pingdom’s the same roughly 10 sec. I don’t know what the heck is going on with this server once again.

For hours I have been trying to figure out – once again – what causes the server’s Physical Memory Usage bounce through the roof with no traffic or no apparent stress whatsoever. I think this is, what causes this totally unforgivable slowness. Most of the time it’s the memory exhaustion, but some other times it’s the CPU, that’s running 100% and the site’s being unresponsive. Once again with no apparent reason.

I had this problem also earlier this year, and I thought that it was after lots of trial-and-error fixed for good, and I was quite happy with everything.
I had overall load times even faster than on some much more high-end servers.
This specific Designer’s load times were at best like 1,5 sec – and now this!

Hours and hours already wasted and no result in the sight. This is so-oo boring.

Now, lastly, Pingdom (Frankfurt) gave 4 sec load time, which of course is better than 10 sec, but here the performance grade for some reason is quite mediocre.

Story continues the next day

This morning started with the same issue in my mind.

Eventually I stumbled upon a few lines .htaccess code, which after some tests, seemingly enables me to remove one WordPress widget, and seemingly this way of doing things has some speed advantages, too. This thing is still under scrutiny.

Maybe this, along with more normal server behavior today, and a few other tweaks, shows me now quite decent, quite nice performance results in GTMetrix (Vancouver):

This I can live with. With some more content back on the page, GTMetrix in London gives decent ratings too.

So along these lines we go, somewhere around these speeds. In the lower screencap, the significant drop in YSlow score is mostly due to the fact, that GTM London* claims, that my pics aren’t on a CDN, although they are.

This is once again a  reminder to us all, that these scores are not-so-much scientific, but more or less approximations. And as you can see, there can be quite big variations for so many reasons and variables, that it’s hard to tell exactly what and why. If somebody tells otherwise, well, that’s not true. But there are certain paths and quidelines, some of which always help, and some of which usually help. Then there is a lot of wishful thinking.

And the situations and scores can change quite rapidly – usually for the worse. The “optimized equilibrium” is a very shaky equilibrium.
(Which I just noticed again! But I’m not anymore getting into that.)

Lastly: on this site, all WordPress speed related plugins are free, and the web host’s solution is far from a high-end one, quite the opposite. So, in relation to that, these scores, and my optimizations and tweaks, are a lot better than okay. If more money is put on the table, there could be better scores.

* A few hours later, this GTMetrix (London)  gave me this better score. So it changes, this one time for the better, eh?

Now the site has GPDR notice back on, which somewhat changes the speed results; the change works both ways, though, slowing and speeding the page load. Slowing, because that’s a new item for a browser to consider. Speeding, because maybe the browser doesn’t need to render everything on the page before the GPDR validation.

A few days later 1.6 secs, with some more tweaks

At this moment, Google Analytics is on, so this is pretty good. Very good. Earlier today PSI gave me 97%, which is nice. Quite satisfied now.
And I still haven’t implemented a better CDN with WebP’s and all… and still on a mediocre, budget host plan.

GTMetrix score on

And lastly with Fast Velocity Minify activated, the score gets a little better.

GTMetrix score on

Leave a Reply