50 replies to this topic

[RidinHighSpeeds]

 page-speed.gif   34.64KB   36 downloads

Today I installed and ran the Google Page Speed project on http://community.invisionpower.com and the score was 41/100. I also ran this tool against VBSeo's forums and they received a score of 94/100. I absolutely love IPB but I was always concerned as to why VBSeo users always achieve great results and their pages tend to load faster.. If the IPB skin could achieve 90% or better ranking with Google's Page Speed test, I'm certain that would have some type of impact with visitors..?

[Benny.]

i just run this and figures dont look good

40/100 for front page with ipcontent

67/100 for forum/board

[Cyrem]

why VBSeo users always achieve great results and their pages tend to load faster

Because VB looks like it's from 2002

But generally, I use firebug for testing load speeds. Basically the less requests the page makes, the faster it will load(provided those requests are not large in kb's). The more images/scripts you have the slower the page will load etc... But other things such as your server performance, distance you are from the server and your own internet connection also changes connection & loading times.

But this comment caught my eye:

Don't get too hung up on a score. Use this tool as a reference and for testing. Typically getting hung up on a score (which is just a number) will just cause you more heart burn then you need.

My forum got a 73/100.

And when I ran it on this forum it got 59/100.

VBSeo got a 72/100 also.

As you can see, the results cannot be relied upon.

[Marcher Technologies]

Because VB looks like it's from 2002

But generally, I use firebug for testing load speeds. Basically the less requests the page makes, the faster it will load(provided those requests are not large in kb's). The more images/scripts you have the slower the page will load etc... But other things such as your server performance and the distance you are from the server also changes connection times.

But this comment caught my eye:


My forum got a 73/100.

And when I ran it on this forum it got 59/100.

VBSeo got a 72/100 also.

As you can see, the results cannot be relied upon.

depends on the skin and how much bloat you run.... 3.2's
https://developers.g...2F&mobile=false 70/100 (and thats with a slider)
https://developers.g...ex&mobile=false 76/100
3.1.4
https://developers.g...2F&mobile=false
it depends on what you have present, and how much you are willing to bother with converting to sprites/axing for css

[Matt]

I'm aware of PageSpeed and its results. I'm not convinced it translates into real world speed. It penalises heavily if you don't use sprites and sprites aren't something that we can really add into a product that allows admin customisation. It would make CSS editing quite dramatic.

There are a few things I've noted down for the next version, though. For example, we're penalised for sending images with large cookies. I have some ideas to fix that which should improve the results.

[tenaki]

72/100 on ipcontent home page

59/100 forums

I need to some work I think

[Matt]

Page Speed also penalises where you send an image that is then scaled by CSS. We could hack it around to please PageSpeed but the end result will not be any faster or slower to the end user.

I'm happy to implement real and proper fixes to improve speed and reduce bandwidth, but I won't be a slave to a figure produced by a calculation based on generic recommendations for websites.

[tenaki]

I agree I do however think I need to look again at the content, how I size images and the like as I think the actual skin is ok. Also I am wondering about feed blocks on the home page and how they make a difference, if at all and lastly ad banners as they are served off site but are sometimes slow to load.

[SECTalk.com]

80/100 on http://www.sectalk.com/

We used to get close to 90 on vBulletin with vBSEO, but that was after a lot of customization. Only been with IPB a month or so.. it will get better as I have more time to look into speed.

One thing severely lacking is the use of image sprites.

Edit: Also meant to add that guests see a lot more ads so this probably takes a couple points off for us.

[RidinHighSpeeds]

80/100 on http://www.sectalk.com/

We used to get close to 90 on vBulletin with vBSEO, but that was after a lot of customization. Only been with IPB a month or so.. it will get better as I have more time to look into speed.

One thing severely lacking is the use of image sprites.

Edit: Also meant to add that guests see a lot more ads so this probably takes a couple points off for us.

What's your score with ads and then without? You can run the tool as a guest; and then run it when logged in (if ads are not visible) to compare the difference.

[RidinHighSpeeds]

I'm aware of PageSpeed and its results. I'm not convinced it translates into real world speed. It penalises heavily if you don't use sprites and sprites aren't something that we can really add into a product that allows admin customisation. It would make CSS editing quite dramatic.

There are a few things I've noted down for the next version, though. For example, we're penalised for sending images with large cookies. I have some ideas to fix that which should improve the results.

That's great to hear Matt! I just figured it couldn't hurt to share this information... wouldn't want your team to spend a lot of time trying to achieve the highest results, but some comments provided in the speed test may be helpful

[maidos]

xenforo got 91 out of 100 though, nice scores

[Matt]

xenforo got 91 out of 100 though, nice scores

It does - but only because the default skin ships with sprites. This means that if you edit or add a new skin, you're no longer using sprites. If you test a xenforo with a simple custom skin that doesn't use sprites, it drops down to 70/100.

[Marcher Technologies]

It does - but only because the default skin ships with sprites. This means that if you edit or add a new skin, you're no longer using sprites. If you test a xenforo with a simple custom skin that doesn't use sprites, it drops down to 70/100.

thats exactly what i was referring to...
now, if you(not you matt explicitly) are willing to css sprite it ALL, axe any sliders/ads, and use CSS3 wherever possible... you can get just as high as that xf stock skin.... but is it worth that 20 points to have a skin that editing can be a nightmare?

[Matt]

That's the key question. I am happy to work on PageSpeed's results but only to a point. I don't see the value in obfuscating the skin editing process purely to get a higher rank from a tool.

[mikesound]

maybe adding a mini profile thumbnail to the setup could help ?

there must be about 50 100x100 profile thumbnails rescaled to 30x30 on each of the various forum indexes

[Enkidu]

It does - but only because the default skin ships with sprites. This means that if you edit or add a new skin, you're no longer using sprites. If you test a xenforo with a simple custom skin that doesn't use sprites, it drops down to 70/100.

Is it? IPB page speed result:



xF pagespeed result

[Marcher Technologies]

Is it? IPB page speed result:



xF pagespeed result

pray-tell, is that a stock xf?
that is irrelevant if so.... a modified xf skin will show around ipb stock range... this is specific to the use of css sprites... additionally, IPS doesnt send keep-alive headers here i believe, most do.

[Enkidu]

thats exactly what i was referring to... now, if you(not you matt explicitly) are willing to css sprite it ALL, axe any sliders/ads, and use CSS3 wherever possible... you can get just as high as that xf stock skin.... but is it worth that 20 points to have a skin that editing can be a nightmare?

XF comes with a feature that let you add images to your sprites.

http://xenforo.com/c...-sprites.20121/

Although that link talks about smilies and I'm not sure if it extends to skin images but you get the meaning*


---------------------------------
* be creative don't complain about limitations

[Marcher Technologies]

but again, we as the user can sprite it if we want, not everyone wants to hassle with sprites at all, this would confound most users(us skinners would obviously be fine), but it would make the script less approachable to the novice IMHO.