Firefox 2.0 'optimized' builds for G4, G5 and Intel Macs (10/26/2006)

"Just a quick note to let you know that I've posted Firefox 2.0 optimized builds for G4, G5, and intel macs. Besides the build optimizations these also have native-looking ("aqua") form widgets to help make Firefox look more mac-like, though there is also G4 and Intel mac versions with the Firefoxy form widgets for those who prefer them.
http://www.beatnikpad.com/firefox
Thanks!
(I haven't updated to v2.0 yet, but asked him about the update breaking 3rd party addons)
It does, but you can use an extension I link to from that page (Nightly Tester Tools) to quickly turn compatibility back on.
-Neil "

I usually only run the standard releases (and still running v1.5.0x). PCworld has a story today on a Minor Bug in New Firefox.

(NOTE: reports below were from older builds)

(added 3/3/2005)
"Mike
Just checking in to report that the new G5 Firefox build would not even boot on my late-model, box stock G5 2x2Ghz. After installing in Applications, it crashed hard on mousedown and would not even open.

I had to pull Firefox 1.0 when I went to G5 in January because it refused to cooperate with the digital optical sound feature (Logitech Z-680). I was getting Firefox crashes that seemed to implicate audio drivers--go figure. No crashes after Firefox removed...
Semper fi Mac,
Richard D.
(I asked Richard for more info like build version and if he was using URL Manager Pro after the author's comments below-Mike)
It was the "release" (1.01) version found on Neil's "Beatnik Pad" site. I do not use URL Manager Pro and have never installed it on this box. I will download FFox again from Neil and see if it runs from the desktop.
Richard "

(From the 3/1/2005 news page) The author of the page with the G5 builds sent some additional notes and links:

"Hi, Mike
I just wanted to drop you a quick email, as I noticed someone passed along a link to the G5-optimized Firefox builds I've been creating.

First of all, if you could update the link to point to the entry on my web site, I'd be most appreciative - this is the best place for updated info on the builds, as well as user comments.
http://www.beatnikpad.com/firefoxG5
For people looking for a G5-optimized build of the Firefox 1.0.1 release version, you can download it from here:
http://www.beatnikpad.com/archives/2005/02/27/firefoxG5

Regarding the crash that "Richard D." mentioned, here are a few things to look into:

1. Was it the nightly or release build that was crashing?
(see Richard's reply above (release build he said). He also noted he did not have URL Manager Pro installed.-Mike)

2. With Firefox quit, try moving the entire Firefox folder from the user Application Support folder to the desktop, and then relaunch the application. This will cause Firefox to create a fresh profile which may eliminate any installed extension conflicts or profile issues. If this does fix the issue, you can then go into the old Firefox folder and move your bookmarks over, etc.

3. As far as I know, the only crashing conflict with the G5 builds is a conflict with URL Manager Pro - is it possible that Richard has this installed? I've spoken with the URL Manager Pro developer and neither of us have any idea why this is happening. My suspicious is that there are bugs in the G5 optimizations, but I'm not sure.
Hope this is helpful. Thanks for the great web site and resource!
Neil"

(added 3/3/2005 from 3/2 mail)
"Hi Mike, I've spent several hours now trying to measure any performance difference between the official 1.01 release, the published G5-optimized build noted in your news, and a build I configured myself after a few hours perusing GCC build options and using Apple-designed compilation options along with use-modeled branch prediction. My testing included:

1) loading three large pages in a row, each scrolled to the bottom, and then using history to navigate back to the first page and then forward to the third page, only clicking back or next after the page had loaded and re-scrolled to the bottom.

2) loading a tab of bookmarks consisting of 11 computer news-oriented sites and timing:

a) the point at which the first tab became usable
b) the total time for all tabs to finish loading

3) loading Gmail (full page of email on the index page)

4) subjective testing comparing scroll performance on busy pages (anandtech articles used, as they're among the slower-scrolling pages I routinely visit on Mac browsers due to heavy animated advertising use)

First a comment: far in excess of everything else like build settings / browser types, etc, the single biggest performance penalty when surfing is advertising, usually due to the frequent (but not consistent) load delays as the browser waits on remote ads to be served. Removing all advertising (via Adblock, a firefox extension) produces a massive increase in the reliability of website load times. The sites themselves are generally very heavily optimized and run on good networks with plenty of bandwidth. The ads served along with those pages are an entirely different matter, with wildly varying page-load times as the result. It was almost impossible to consistently measure performance in some of the tests, simply due to the variance imposed by third party advertising services used by those sites. It took dozens of iterations of each test to produce results unmarred by an obvious ad-related delay, which would affect one or more sites in one test run and another random set on the next. These delays far outweigh any performance difference I was able to record on the browsers themselves.

With that out of the way, some performance numbers (adblock was used to remove off-server advertising on the entire list of sites in order to minimize inconsistent performance results - note that this did not noticeably improve the best scores - only made scores more consistent). Tested on a Dual G5 2.0, processor performance "highest", 10.3.8:

Tab-load 11 PC-related sites (BEST time to load FIRST page):

• Firefox official 1.01: 4.5
• Firefox "G5-optimized": 4.3
• Safari: 2.9
• Private firefox build: 4.0

Tab-load 11 PC-related sites (BEST time to load ALL pages):

• Firefox official 1.01: 4.9
• Firefox "G5-optimized": 5.0
• Safari: 12.3 (not a strong-point of Safari)
• Private firefox build: 4.5

back/back/fwd/fwd test:

• Firefox official 1.01: 4.0
• Firefox "G5-optimized": 3.8
• Safari: 7.3
• Private firefox build: 3.0

(Note: the ads were removed for all firefox versions, which improved scores by about 1.8-2 seconds - they averaged closer to 5 seconds with ads, in a similar spread).

GMail (first visit after launch):

• Firefox official 1.01: 4.7
• Firefox "G5-optimized": 4.9
• Safari: 4.7
• Private firefox build: 4.7

Subjective test: scrolling Anandtech article pages: This is the most revealing test, in my opinion: These browsers are all terrible at scrolling those pages, apparently due to the animations on them. Nothing was noticeably different about the general browser experience on any of these 4 browsers. The only time things improved noticeably was when Adblock was used to prune the advertisements - and then the change was dramatic.

Conclusion: it's really a wash, from a perceived experience point of view (except with Safari in some cases - and that's mostly because Apple is holding off on the next real Safari upgrade until Tiger). The things the Mac is currently not very good at don't get significantly better without taking a blowtorch to the sites themselves. Ad removal is a far larger performance improvement. Note that I do not suggest to people that they do this -- I'm just pointing out where the real performance issues are (not that that should surprise anyone). People make their living from those ads, and if nobody sees/uses them, many if not most those sites would likely go away. Keep that in mind when deciding what you will do.

I'll see if I can profile the applications with Shark (apples performance-tuning tool) and perhaps at least identify where the main bottlenecks are on image scrolling (not that I'm likely to be able to do anything about them - I'm just curious and it looks like a neat tool).
cheers,
Michael "

The problem in my mind with doing typical web site page loading tests is that the source's web server load (and net traffic) is constantly changing which can affect response times and page loading times. But Michael included other tests also not just those affected by the network/server.

" Hi, Mike -
Just a very quick followup to the G5-optimized Firefox speed tests that someone has been doing (and are very cool, by the way).

My impression with these builds is that you can't really speed up pageload times that radically, even with an optimized version. Where I really notice the differences is in just how the application itself *handles* - switching tabs, for example, is much faster with these builds than with the official ones, as is invoking menus. I've personally found them to be faster at scrolling long pages too, but your mileage may vary.

It's interesting to see actual speed tests, but as your note at the bottom of the reader's report states, a much more accurate test would be to load test pages from within an internal network. I have done a teensy bit of this (to see if I'm just wasting my time with these builds), and there does seem to be an improvement, albeit fairly small.

I do think (hope?) that these builds will be noticeably faster once OS 10.4 and Xcode 2 is released - there are a bunch of changes in the gcc compiler, and as well I seem to remember reading somewhere that the OS itself will have optimizations for the G5. As with anything 10.4-related, we'll have to wait and see.
Neil "