I’ve seen a bunch of links to various tweaks intended to make Firefox run faster. Boing Boing probably spread this go-faster tweak for Firefox farther than anyone. In addition, Brian Livingston published a lengthy Secrets of Firefox 1.0 article in his Windows Secrets newsletter last month.
I’ve been writing about various Windows speed-up tricks through the years, many of which are very popular and either misleading or flat-out wrong. Often, someone who follows all the advice in one of these articles winds up with a system that runs slower and is less stable than it was before.
That may well be the case with these Firefox tweaks as well. Brian is a reliable source of information, and I trust his advice. I also believe him when he writes:
The most sought-after performance improvements in any browser will always involve how quickly it downloads and renders Web pages. The good news is that Firefox (which is already pretty fast in its default configuration) includes numerous about:config settings that can improve the downloading and display of content. The bad news is that the optimum settings will differ from machine to machine, and there’s no consensus on what they should be.
After extensive research, I haven’t found a utility or even a well-tested explanation that can guarantee the optimum settings for any particular Windows scenario (Windows 2000 vs. XP, DSL vs. T1, etc.).
There are scores of Web sites that speculate on configuration settings that are said to speed up the browsing experience in Firefox. But these sites largely don’t show that they’ve done adequate testing of the alternatives, much less explain how such tests might have been conducted.
Asa Dotzler of Mozilla has written a cautionary note about some of the speed-up tips going around. He says something very similar:
Just note that what works for one person/system, may not work for another.
Yes, there are tuning change you can make (even at compile time, see Moox’ optimized builds) that will dramatically alter the performance characteristics of Firefox. Feel free to experiment, but remember that most of the defaults are defaults for a reason. If your browser starts misbehaving or web sites look broken, it might be worth going back to default settings.
That seems like a good opportunity to mention what I consider as one of Firefox’s greatest features: You can create and copy profiles anytime so you can test settings and extensions. If you’re trying out some odd tweak or extension, keep a copy of your old profile. If the tweak doesn’t work or the extension causes problems, you can quickly return to your old profile.
Windows XP users can open the Firefox Profiles folder by clicking Start, then Run. In the Open box, type %userprofile%\Application Data\Mozilla\Firefox\Profiles (include the percent signs, which automatically take you to your personal data folders). Make a copy of the profile folder you see there, which consists of a random eight-character string and the name of your default profile. You can then make changes to your current profile; you can undo those changes by closing Firefox and restoring the backed-up folder.
To create a new profile, use the well-hidden Profile Manager. Use the Run dialog box again and type firefox.exe -profilemanager as the command.
You can use the Profile Manager to switch between profiles. I actually keep several profiles – one for everyday use, one for some special-purpose tasks that require extensions I don’t normally use, and one that is completely clean, so I can test pages without fear that an extension is distorting my results.
Ed Bott 
The best performance tweak I ever tried was about a month ago — I bought a brand new computer with a much faster CPU and lots more memory. 🙂
These tweaks are like many of the drug commercials you see on TV: lot’s of information about how wonderful the drugs supposedly are, followed by a long litany of scary and potentially very serious side effects that the drugs might cause. I would love to see much more information about what these tweaks are really supposed to do, why they are supposedly better for your machine than the default settings, whether and how objectively to measure their impact, whether the changes will actually be transparent to the user (they almost never are), and what the risks or downsides are.
The Firefox layout and networking changes are actually reasonable places to experiment. For example, we used to set the intial paint delay at 500ms. After quite a bit of tweaker/tester feedback and the general trend of faster CPUs and higher bandwidth we changed that in Firefox to 250ms and with mostly favorable results (though some people on slower connections, slower CPUs or both, might want to set it at a higher value, like 500ms or even 1000ms). People with really good bandwidth and fast CPUs might have a better experience setting the value even lower because they’ll see something on screen faster (though it may take longer to render everything if the CPU is blowing a lot of cycles trying to recompute the layout as more data comes in.)
Many of these Firefox tweaks are like almost any other optimization, a tradeoff. It’s just important that people know what it is they’re trading when they make these changes.
Turning on experimental features like Pipelining is a bit different. It can make things quite fast and that’s why support (though turned off by default) was added to our products. One major problem is that many servers and proxies have problems with pipelining so we decided that the performance gain, however nice, isn’t worth the occasional ‘unexplained’ failure.
–Asa