Slashdot Mirror
FireFox 3.1 Leaves IE in the Dust
Anonymous writes "Granted, FireFox 3.1 is just a beta and IE 8 is also in beta, but it looks like Microsoft has some ground to make up when it comes to browser performance. Given that Mozilla appears to be on a much faster cycle than Microsoft with this stuff, it's also possible that it could increase the gap even more before IE 8 is GA, no?"
If you've even remotely been keeping up with FireFox, WebKit and IE progress, it's no surprise that IE8 fares poorly. It fared poorly the day it was released, which was about two months ago. Why are we getting this story now?
As a side note, IE8 does fix the pathologically bad performance IE6/IE7 exhibited on certain SunSpider benchmarks. That alone should improve its overall SunSpider score by an order of magnitude. Its javascript engine will still be 2-3X slower than FireFox and Safari, but it will at least be in the same "ball park".
It's not like IE has not been a slow dog in javascript performance and standards adoption. Yeah, IE 7/8 are supposed to be an improvement, but since IE is years behind and their development cycles seem to be as slow as their javascript engine (probably due to compatibility) it's not like IE 8 or 9 is going to catchup with the rest of the browsers easily.
BTW, those benchmarks in TFA were probably run with the new tracemonkey javascript engine disabled (it need to be enabled manually in about:config). And my firefox nightly version passes 93/100 on the acid 3 test.
It feels like the .NET koolaid is coming even to the IE team. Microsoft's .NET push now borders on maniacal, standardizing on .NET and in places where it should not be standardized. Performance matters, particularly when processors aren't getting any faster, just more parallel. Microsoft's has left C++ to languish, has all but abandoned C, and as such has no real performance tool in their own arsenal.
But IE isn't built on .NET is it? And there are improvements in MSVC in VS2008 for both C and C++ and they've had OpenMP and a much improved STL for two versions now.
For my interest, when have major OSS projects "changed compilers" to respond? I can't think of any examples.
I don't know about the rest of the world, but Firefox 3.0.3 sucks on my three XP machines. Version 3.0.2 worked just fine. I let Firefox upgrade itself to 3.0.3 and it immediately started crashing. It crashed so much that I actually had to use IE to download a copy of 3.0.2 to downgrade Firefox on those machines. And Firefox 3.0.3 crashes on my Ubuntu machine far far more often that earlier versions ever did (although I'm still using 3.0.3 on Ubuntu).
The NSA: The only part of the US government that actually listens.
Huh? Firefox 3 is in production. Firefox 3.1 is in beta. As in real beta, not out yet, in testing.
Reviewing just the first hour of video games.
Last time I checked, Ctrl+Tab switched tabs in Firefox 2, it's just a new flashy display in 3.1
Because in most programming libraries, time is normally expressed in milliseconds.
And some of us installed "Old Location Bar" add-in within 20 minutes of installing FF3 because, as somebody so eloquently put it, "If I'd wanted to check my bookmarks, I would have *opened* my bookmarks".
Well, javascript-wise, maybe not. I've spent the last week doing some extensive testing on pure javascript performance (not DOM-tree manipulation, etc) using a little raytracer I hacked up overnight.
Opera is noticably above average, in this respect. More importantly, however, you might note that the Firefoxes are absolutely, hideously bad at memory management. When rendering a big scene here, Safari will do it in a fraction of the time using 60mb of RAM, whereas Firefox 3.1beta's memory usage spirals out of control and into swap space. And the JIT compiler is way broken still :)
Anyway, here are some figures (only meaningful when comparing different browsers on the same box):
IE 7.0.5730.13 -- 10.1 seconds
Firefox 2.0.0.17 -- 9.9 seconds
Safari (win32) 3.1.1 -- 5.9 seconds
Opera 9.60 -- 3.6 seconds
Firefox 3.1b2pre (no JIT) -- 2.8 seconds
Safari (win32) 2008-10-15 -- 1.0 seconds
Google Chrome 0.2.149.30 -- 0.8 seconds
Firefox 3.1b2pre (JIT) -- anywhere between 0.6-35.0 seconds
Today's weirdness is tomorrow's reason why. -- Hunter S. Thompson
In JS, it's "not identical". It means "don't try to do any implicit casting - not only must their values be the same, but their type must be the same too"
I get pinged on it all the time when I'm running other people's JS through http://www.jslint.com/
Opera used to be faster, but not any longer now that Mozilla, Apple, and Google have been improving performance so much. If you look at the links in the previous post, you can see that Opera beat Firefox in only one speed test. Yes, Opera is faster than IE, but that's only because IE is so slow.
What a fool believes, he sees, no wise man has the power to reason away.
Actually, you're wrong -- Firefox 3.1 comprehensively outperforms Opera in pretty much all tests now.
On the other hand, what does blow the FF 3.1 beta out of the water, are the latest WebKit betas. Here's the stats on my machine:
Sunspider (faster times are better)
FF3.0.3 - 2697.2ms
FF3.1 - 2442.8ms
WebKit - 702.6ms
V8 Benchmark (more runs are better)
FF3.0.3 - 199 runs
FF3.1 - 241 runs
WebKit - 2087 runs
ACID 3
FF3.0.3 - 71 and significant laggyness
FF3.1 - 89 and significant laggyness
WebKit - 100 and passes timing tests
Your code is totally wicked ! Am i the only one impressed by raytracing in javascript ? By the way on my machine midori (webkit) is two times faster than opera 9.60.
No, I don't remember performance being mentioned in the anti-trust cases. And why would they? It would highlight an uncompetitive advantage and weaken their position.
Citation needed.
Well if you're willing to take Wikipedia as a citation, then:
"A number of videotapes were submitted as evidence by Microsoft during the trial, including one that demonstrated that removing Internet Explorer from Microsoft Windows caused slowdowns and malfunctions in Windows." (Emphasis added)
Microsoft later withdrew the claim, but only because the plaintiffs spotted that Microsoft attempted to mislead the court with their initial video demonstration. I'm still disgusted that no one from Microsoft was directly prosecuted for any of that.
Yes but wget still rules
And it works even if all you have is punch cards for input and a dot matrix printer for screen output. ;-)
And it has none of that fancy curses based rendering to slow it down.
It is my understanding that the primary advantages of FF 3.1 are speedups to Javascript and adoption of new HTML tags.
For me (and for much of the web browsing community, such as my cousin, aunt & father, e.g. the 50-84 y.o community) these are USELESS.
1) I am slowly converting my family entirely over to using Firefox with NoScript -- because *anyone* who allows random internet sites to run software on their machine is *nuts* [1].
2) A significant majority of "common" sites will not be using enhanced HTML tags because they have to continue to work with the installed browser base.
This is another example of Mozilla developers getting side-tracked with respect to what is important to *them* rather than what might be important to the community [2].
1. The *real* advantage of Firefox is the selected enabling of Javascript for a few "trusted" relatively non-commercial sites (e.g. gmail, ones bank, ones broker) using NoScript. I will assume the display of pages from such sites is relatively unimpacted by Javascript speedups (since they tend to be network bandwidth or user input consrained). [Though it is worth noting that the gmail javascript appears to be becoming a bit of a pig.]
2. It is worth noting that my cousin, my aunt and my father continue to survive on the internet quite well using dial-up connections (in large part because they live in regions where DSL (or fiber) is unavailable and Cable is too expensive). I presume that G3 service will fall into the $$$ category even when reasonably priced modems that can connect their computers to the net become available.
Right you are. FF3.0.3 appears to give very slightly slower results than 3.1beta with JIT disabled
Today's weirdness is tomorrow's reason why. -- Hunter S. Thompson
Because IIRC they were claiming that by integrating they were helping the customer who had a low memory machine by reducing the memory footprint of the browser. Remember that this was a time when RAM was still pretty expensive and many machines only had 64Mb of RAM from the factory. So their claim was that they were helping the customer,not trying to kill competition.
Of course many of us that wanted speed at that time simply used tools like this to strip IE out of the OS while leaving the HTML rendering component to keep from breaking apps that used the IE engine for help files. This actually would give a nice speed boost to Win98 as it didn't have IE running constantly in the background anymore.Of course if you want to look up the trial trascript go ahead,but I already have a skull thumping headache and can't take legalese ATM,sorry.
ACs don't waste your time replying, your posts are never seen by me.
FF3.1 beta 1 comes with Tracemonkey disabled by default. You can enable it in about:config by setting javascript.options.jit.content to true. Otherwise you are just using the same javascript engine of FF3.0.3 with minor improvements.
source: http://digg.com/software/First_look_Firefox_3_1_beta_1_officially_released#c-text-c19760714